Epidemiology/Health Services/Psychosocial Research

O R I G I N A L A R T I C L E

Defining the Relationship Between

Plasma Glucose and HbA1c
Analysis of glucose profiles and HbA1c in the Diabetes Control and
Complications Trial
CURT L. ROHLFING, BES JACK D. ENGLAND patients and their health care providers to
HSIAO-MEI WIEDMEYER, MS ALETHEA TENNILL, MS set appropriate daily PG testing goals to
RANDIE R. LITTLE, PHD DAVID E. GOLDSTEIN, MD achieve HbA1c levels representing low
risks for adverse outcomes.
Several previous studies have ana-
lyzed the relationship between blood glu-
cose (BG) and HbA1c. Svendson et al. (3)
OBJECTIVE To define the relationship between HbA1c and plasma glucose (PG) levels in assessed 15 subjects with type 1 diabetes
patients with type 1 diabetes using data from the Diabetes Control and Complications Trial who collected seven-point BG profiles
(DCCT).
over a 5-week period (three profiles per
RESEARCH DESIGN AND METHODS The DCCT was a multicenter, randomized week) and used a curvilinear equation to
clinical trial designed to compare intensive and conventional therapies and their relative effects correlate BG and HbA1c. Nathan et al. (4)
on the development and progression of diabetic complications in patients with type 1 diabetes. obtained repeated preprandial and post-
Quarterly HbA1c and corresponding seven-point capillary blood glucose profiles (premeal, post- prandial BG samples from 21 subjects
meal, and bedtime) obtained in the DCCT were analyzed to define the relationship between with type 1 diabetes over an 8-week pe-
HbA1c and PG. Only data from complete profiles with corresponding HbA1c were used (n riod and used a linear regression equation
26,056). Of the 1,441 subjects who participated in the study, 2 were excluded due to missing to describe the relationship between BG
data. Mean plasma glucose (MPG) was estimated by multiplying capillary blood glucose by 1.11. and HbA1c. In the DCCT, the correlation
Linear regression analysis weighted by the number of observations per subject was used to between HbA1c and mean BG was initially
correlate MPG and HbA1c.
determined in a limited number of pa-
RESULTS Linear regression analysis, using MPG and HbA1c summarized by patient (n tients (n 278) for the feasibility study
1,439), produced a relationship of MPG (mmol/l) (1.98 HbA1c) 4.29 or MPG (mg/dl) (5). However, a comprehensive analysis
(35.6 HbA1c) 77.3, r 0.82). Among individual time points, afternoon and evening PG of the relationship of BG and HbA1c, ex-
(postlunch, predinner, postdinner, and bedtime) showed higher correlations with HbA1c than amining BG at different time points and
the morning time points (prebreakfast, postbreakfast, and prelunch). using the entire data set, was never per-
formed. Here, we examine, in detail, the
CONCLUSIONS We have defined the relationship between HbA1c and PG as assessed in relationship between BG (converted to
the DCCT. Knowing this relationship can help patients with diabetes and their healthcare PG) and HbA1c, using data obtained from
providers set day-to-day targets for PG to achieve specific HbA1c goals. the entire DCCT data set to better define
Diabetes Care 25:275278, 2002
this relationship.

RESEARCH DESIGN AND

METHODS The DCCT data set was

T
he results of the Diabetes Control results of the DCCT, the American Diabe- provided by the National Institutes of Di-
and Complications Trial (DCCT), tes Association (ADA) has published rec- abetes, Digestive, and Kidney Diseases of
published in 1993, and the U.K. ommendations for HbA1c and plasma the National Institutes of Health and was
Prospective Diabetes Study, published in glucose (PG) levels that are widely used prepared by the Data Coordinating Cen-
1998, established the relationship be- (1,2). However, it is important that the ter at George Washington University. The
tween HbA1c levels and risks for diabetic relationship between daily patient- DCCT was a multicenter, randomized
complications in patients with type 1 and monitored blood glucose determinations clinical trial designed to compare inten-
type 2 diabetes, respectively. Based on the and HbA1c be clearly defined to enable sive and conventional therapies and their
relative effects on the development and
progression of diabetic complications in
From the University of Missouri School of Medicine, Columbia, Missouri.
Address correspondence and reprint requests to Curt L. Rohlfing, University of MissouriColumbia, patients with type 1 diabetes (1). The
Department of Child Health, 1 Hospital Drive M772, Columbia, MO 65203. E-mail: rohlfingc@ study population consisted of 1,441 pa-
health.missouri.edu. tients with type 1 diabetes recruited by 29
Received for publication 20 April 2001 and accepted in revised form 18 October 2001. centers located throughout the U.S. and
Abbreviations: ADA, American Diabetes Association; BG, blood glucose; DCCT, Diabetes Control and
Complications Trial; MPG, mean plasma glucose; PG, plasma glucose.
Canada. Patients were between 13 and 39
A table elsewhere in this issue shows conventional and Systeme International (SI) units and conversion years of age and did not show evidence of
factors for many substances. severe diabetic complications at the time

DIABETES CARE, VOLUME 25, NUMBER 2, FEBRUARY 2002 275

HbA1c and plasma glucose

of admission into the study. Intensive

therapy consisted of three or more insulin
injections daily or use of an insulin pump
with the intent of achieving BG values as
close to the normal range as possible.
Conventional therapy consisted of one or
two insulin injections per day. Mean du-
ration of participation was 6.5 years
(range 39 years).
Quarterly HbA1c measurements (n
37,058) and corresponding BG profiles
were obtained from 1,441 subjects. After
exclusions due to incomplete profiles,
there were 26,056 HbA1c values with cor-
responding seven-point profiles from
1,439 subjects (an average of 18 HbA1c
values and corresponding profiles per
patient).
For the seven-point BG profiles, cap-
illary blood hemolysates were collected
before meals, 90 min after meals, and at
bedtime by patients in the home (6). BG
was measured in a central laboratory us-
ing a hexokinase enzymatic method (7).
Blood for HbA1c analysis was collected by Figure 1MPG versus HbA1c: n 1,439; r 0.82; PG (mmol/l) (1.98 HbA1c) 4.29. The
venipuncture. HbA1c was measured in a dashed line indicates the regression line.
central laboratory using an ion-exchange
high-performance liquid chromatogra-
phy method (8,9). 6 9% HbA1c. Within-subject (intraindi- HbA1c to facilitate clinical interpretation
Statistical analysis was performed us- vidual) variation in HbA1c was much and use of these data.
ing SAS and SPSS (Chicago, IL). Mean BG lower than for seven-point PG (mean in- Results of regression analyses corre-
was determined using area-under-the- traindividual coefficient of variation lating HbA1c with individual premeal and
curve analysis (10). For each profile, the 9.7 vs. 29.8%, respectively). postmeal PG are summarized in Figs. 2
seven time points were connected by MPG at increasing levels of HbA1c is and 3. All individual time points showed
straight lines over time for a 24-h period, shown in Table 1. Along with regression- lower correlations than the seven-point
and then the trapezoidal areas under each estimated MPG, the table shows approxi- profiles. Prelunch and earlier PG time
curve were determined, added together, mate MPG based on increments of 2 points showed lower correlations with
and divided by time. A constant BG level mmol/l or 35 mg/dl per 1% change in HbA1c than postlunch and later PG time
between bedtime and the following points.
morning was assumed. Mean plasma glu-
cose (MPG) was estimated by adding 11% Table 1MPG as estimated from the regres-
to mean BG estimates (11). Mean MPG sion line and approximate MPG (based on CONCLUSIONS The increasing
and HbA1c were calculated for each sub- MPG change of 35 mg/dl or 2 mmol/l per 1% use of HbA1c to monitor long-term glyce-
ject and used to perform least-squares lin- change in HbA1c) at different HbA1c levels mic control in diabetic patients is largely
ear regression analysis. Due to variation in the result of data from the DCCT and the
the number of observations per subject, U.K. Prospective Diabetes Study showing
the regression analysis was weighted to Regression- Approximate that HbA1c is strongly correlated with ad-
account for this. The relationships be- estimated MPG for clinical verse outcome risks. For patients and
tween individual PG time points and MPG use health care providers, a clear understand-
HbA1c
HbA1c were also examined. (%) mmol/l mg/dl mmol/l mg/dl ing of the relationship between PG and
HbA1c is necessary for setting appropriate
RESULTS The results of linear re- 4 3.6 65 3.5 65 day-to-day PG testing goals with the
gression analysis are summarized in Fig. 5 5.6 101 5.5 100 expectation of achieving specific HbA1c
1. The Pearson correlation coefficient (r) 6 7.6 137 7.5 135 targets.
was 0.82; change in MPG per increase of 7 9.6 172 9.5 170 The relationship between HbA1c and
1% HbA1c was 1.98 mmol/l (35.6 mg/dl). 8 11.5 208 11.5 205 PG is complex. Many studies have shown
The 95% prediction interval for a subject 9 13.5 244 13.5 240 that HbA1c is an index of MPG over the
with 18 observations (the average num- 10 15.5 279 15.5 275 preceding weeks to months. Erythrocyte
ber of profiles per patient in this study) 11 17.5 315 17.5 310 life span averages 120 days. The level of
was 3.81 mmol/l (69 mg/dl) at levels of 12 19.5 350 19.5 345 HbA1c at any point in time is contributed

276 DIABETES CARE, VOLUME 25, NUMBER 2, FEBRUARY 2002

 Rohlfing and Associates

PG levels from 90 120 days earlier con-

tribute only 10% (12,13). This explains
why the level of HbA1c can increase or
decrease relatively quickly with large
changes in PG; it does not take 120 days
to detect a clinically meaningful change in
HbA1c after a change in MPG.
Another factor that complicates ef-
forts to describe an accurate and precise
relationship between PG and HbA1c is
that, for practical reasons, previous stud-
ies and our present study have attempted
to define this relationship using a limited
number of PG levels measured over a lim-
ited time period (in this case, 1 day every
3 months) to estimate HbA1c. Short-term
PG levels can fluctuate markedly, partic-
ularly in patients with type 1 diabetes; this
can result in significant discrepancies
when attempting to estimate HbA1c based
on a single PG measurement or even a
series of measurements on a single day. In
this study, the time between sampling
also contributes to intraindividual varia-
Figure 2Premeal MPG and r at different testing times. , Prebreakfast; -----, prelunch; tion, especially for PG. However, we have
- , predinner; , seven-point. achieved greater certainty in our estimates
of the relationship between PG and HbA1c
than was possible in previous studies by
to by all circulating erythrocytes, from the past PG levels (i.e., 3 4 months earlier). using a considerably larger number of pa-
oldest (120 days old) to the youngest. Therefore, HbA1c is a weighted average tients and observations obtained over a
However, recent PG levels (i.e., 3 4 of BG levels during the preceding 120 longer period of time. The resulting
weeks earlier) contribute considerably days; PG levels in the preceding 30 days strong correlation suggests that, although
more to the level of HbA1c than do long- contribute 50% to the final result, and a single PG measurement or a single daily
profile may not reliably predict HbA1c,
PG levels measured over time can provide
a reasonably accurate estimation of
HbA1c.
Several studies have suggested that,
although intraindividual variation in
HbA1c is minimal, there is evidence of
wide fluctuations in HbA1c between indi-
viduals that are unrelated to glycemic sta-
tus, suggesting that there are low
glycators and high glycators (14 16).
However, a recent study showed that
when multiple observations per patient
are used to minimize the effects of assay
variation, the interindividual range of
HbA1c results in nondiabetic individuals
is actually quite narrow, 1% HbA1c
(17). Therefore, for any individual pa-
tient, a consistent discrepancy between
patient-monitored PG determinations
and estimated HbA1c should be investi-
gated; there may be other factors causing
this discrepancy, such as improper meter
use, laboratory error, a physical condition
that alters red cell life span, or a variant
Figure 3 Postmeal MPG and r at different testing times. , Postbreakfast; -----, Postlunch; hemoglobin interfering with the HbA1c
- , postdinner; -- , bedtime; , seven-point. assay method. With the advent of new

DIABETES CARE, VOLUME 25, NUMBER 2, FEBRUARY 2002 277

HbA1c and plasma glucose

technologies that are capable of moni- reference method, as recommended by globins with a new high-performance
toring PG on a 24-h basis (18), it will be the ADA (20). Fasting PG should be used liquid chromatography analyzer. Clin
interesting to see how our estimate of the with caution as a surrogate measure of Chem 32:202203, 1986
relationship between PG and HbA1c com- MPG because it may significantly under- 10. Tai MM: A mathematical model for the
pares with estimates obtained using these estimate HbA1c and, therefore, risks for determination of total area under glucose
tolerance and other metabolic curves. Di-
technologies. complications at increasing HbA1c levels.
abetes Care 17:152154, 1994
Our data indicate that fasting PG 11. Fogh-Andersen N, DOrazio P: Proposal
alone should be used with caution as a for standardizing direct-reading biosen-
measure of long-term glycemia. Fasting Acknowledgments We thank the DCCT
study group and the Data Coordinating Center sors for blood glucose. Clin Chem 44:655
PG tended to progressively underestimate at George Washington University for provid- 659, 1998
HbA1c (and seven-point MPG) at increas- ing the data set as well as the patient volunteers 12. Tahara Y, Shima K: The response of GHb
ing PG levels. The data also suggest that who participated in the DCCT. to stepwise plasma glucose change over
postmeal PG contributes appreciably to time in diabetic patients. Diabetes Care
HbA1c; however, all postmeal times are 16:13131314, 1993
not equal in their contribution. We found References 13. Goldstein DE, Little RR, Wiedmeyer HM,
that compared with the seven-point pro- 1. The Diabetes Control and Complications England JD, Rohlfing CL: Glycohemoglo-
files, postbreakfast levels markedly over- Trial Research Group: The effect of inten- bin testing in diabetes mellitus: assay
estimate HbA1c, whereas postlunch levels sive treatment of diabetes on the develop- methods and clinical interpretation. In
ment and progression of long term Drugs in Development. Vol. 1. Vasselli JR,
show a relationship to HbA1c that is very complications in the diabetes control in Maggio CA, Scriabine A, Eds. Branford,
similar to that of MPG. A previous study insulin dependent diabetes mellitus. CT, Neva Press, 1993, p. 253267
of patients with type 2 diabetes also found N Engl J Med 329:977986, 1993 14. Yudkin JS, Forrest RD, Jackson CA, Ryle
that postlunch PG is a better indicator of 2. American Diabetes Association: Stan- AJ, Davie S, Gould BJ: Unexplained vari-
glycemic control than fasting PG (19). dards of medical care for patients with di- ability of glycated hemoglobin in non-di-
However, that study did not examine abetes mellitus (Position statement). abetic subjects not related to glycemia.
bedtime PG, which we found also shows a Diabetes Care 24 (Suppl. 1):S33S43, Diabetologia 33:208 215, 1990
relationship to HbA1c that is very similar 2001 15. Kilpatrick ES, Maylor PW, Keevil BG: Bi-
to that of MPG. 3. Svendson PA, Lauritzen T, Soegaard U, ological variation of glycated hemoglobin:
Nerup J: Glycosylated haemoglobin and implications for diabetes screening and
The ADA currently recommends that
steady-state mean blood glucose concen- monitoring. Diabetes Care 21:261264,
patients with diabetes attempt to achieve tration in type 1 (insulin-dependent) dia-
average preprandial PG levels of 5.0 7.2 1998
betes. Diabetologia 23:403 405, 1982 16. Hudson PR, Child DF, Jones H, Williams
mmol/l (90 130 mg/dl) and average bed- 4. Nathan DM, Singer DE, Hurxthal K,
time PG levels of 6.1 8.3 mmol/l (110 CP: Differences in rates of glycation (gly-
Goodson JD: The clinical informational
cation index) may significantly affect in-
150 mg/dl) as well as HbA1c 7% (2). value of the glycosylated hemoglobin as-
dividual HbA1c results in type 1 diabetes.
Our results show estimated average pre- say. N Engl J Med 310:341346, 1984
Ann Clin Biochem 36:451 459, 1999
prandial PG and bedtime PG levels of 8.7 5. The Diabetes Control and Complications
17. Wiedmeyer HM, Rohlfing CL, Little R,
and 9.2 mmol/l (157 and 166 mg/dl), re- Trial Research Group: Diabetes Control
and Complications Trial (DCCT): results Grotz VL, Tennill A, Goldstein D: Do bi-
spectively, at 7% HbA1c. These data sug- ological factors other than changes in gly-
of feasibility study. Diabetes Care 10:1
gest that patients who consistently cemic status affect glycohemoglobin
19, 1987
achieve ADA-recommended BG and PG 6. Schlebusch H, Sorger M, Munz E, Kessler results? (Abstract) Diabetes 49 (Suppl. 1):
targets will also achieve an HbA1c level A, Zwez WP: Glucosebestimmung in A96, 2000
7%. hamolysierten blutproben. J Clin Chem 18. Bode BW, Sabbah H, Davidson PC:
In summary, there is a predictable re- Clin Biochem 18:885 891, 1980 Whats ahead in glucose monitoring? New
lationship between PG and HbA1c. Un- 7. Neeley WE: Simply automated determi- techniques hold promise for improved
derstanding this relationship will allow nation of serum or plasma glucose by a ease and accuracy. Postgrad Med 109:41
patients with diabetes and their health- hexokinase/glucose-6-phosphate dehy- 49, 2001
drogenase method. Clin Chem 18:509 19. Avignon A, Radauceanu A, Monnier L:
care providers set appropriate day-to-day
515, 1972 Nonfasting plasma glucose is a better
PG targets based on HbA1c goals. It is im- marker of diabetic control than fasting
8. Dunn PJ, Cole RA, Soeldner JS: Further
portant to note that the relationship be- plasma glucose in type 2 diabetes. Diabe-
development and automation of a high-
tween PG and HbA1c defined in this study pressure liquid chromatography method tes Care 20:18221826, 1997
only applies when HbA1c is measured us- for the determination of glycosylated he- 20. American Diabetes Association: Tests of
ing assay methods that are certified by the moglobins. Metabolism 28:777779, 1979 glycemia in diabetes (Position statement).
National Glycohemoglobin Standardiza- 9. Mosca A, Carpinelli A, Bonini P: Auto- Diabetes Care 24 (Suppl. 1):S80 S82,
tion Program as traceable to the DCCT mated determination of glycated hemo- 2001

278 DIABETES CARE, VOLUME 25, NUMBER 2, FEBRUARY 2002