Comparisons of home blood glucose testing and glycated protein measurements

Diabetes Research and Clinical Practice, 16 (1992) (3 1992 Elsevier

Science Publishers

53-62

B.V. All rights reserved 016%8227/92/$05.00

53

DIABET 00609

Comparisons

of home blood glucose testing and glycated protein measurements

Phillip D.K. Lee ‘, Lori D. Sherman ‘, Margaret R. O’Day ‘, Cheryl L. Rognerud * and Ching-Nan Ou * Departments of ‘Pediatrics and ‘Pathology. Baylor College

of Medicine.Houston, Texas, U.S.A.

(Received 26 August 1991) (Revision accepted 29 October 199 1)

Summary

We examined the relationships between 4 glycated protein assays and home blood glucose monitoring (HBGM) in 26 children with poorly-controlled insulin-dependent diabetes mellitus (IDDM) during a period of improved management. At 2 week intervals for 6 visits (12 weeks in total), HBGM records were collected and a blood sample was obtained for measurement of glycated proteins and glucose. Assays included glycated hemoglobin (GHb) and glycated serum proteins (GP) by boronate affinity chromatography, hemoglobin A,C by PolyCAT A high performance liquid chromatography (HAC) and fructosamine (FA). All 4 glycated protein levels declined significantly over the 12 week period. Significant correlations between the glycated proteins and HBGM were observed over 2 week intervals. None of the 4 assays were affected by the glucose level in the sample. Changes in mean HBGM readings over 2 week intervals were correlated with both FA and GP with wide prediction intervals. Over cumulative 2 week intervals, which may more accurately reflect longitudinal trends, all 4 glycated proteins were correlated with mean HBGM readings. At each cumulative interval, GHb and GP showed the largest variation with MBG, while FA showed the least variation with MBG. Our data indicate that of the 4 assays tested, FA has limited clinical value as compared to other glycated protein assays, whereas assays based on boronate alhnity chromatography (GHb and GP) provide the most useful clinical indicators of short-term changes in glycemic control. The clinical utility of a new HPLC method for determination of glycated hemoglobins is also demonstrated. Key words: Glycohemoglobin; --__

Fructosamine;

Glycated

Correspondence to: Phillip D.K. Lee, Diabetes Care Center, Texas Children’s Hospital Clinical Care Center, Suite 770, 6621 Flannin Street, Houston, TX 77030, U.S.A. Portions of this study were presented in abstract form at the 14th International Diabetes Federation Conaress. Washington D.C., 1991.

protein;

Home blood glucose monitoring

Introduction

Over the past decade, treatment of insulindependent diabetes mellitus (IDDM) has increasingly emphasized near-normalization of blood

54

glucose levels [ l] ; a goal based on evidence that hyperglycemia is the major risk factor for diabetic vascular complications. To achieve this goal, current standards of care for IDDM include daily home blood glucose monitoring (HBGM) and periodic measurement of glycated hemoglobin to assess integrated glycemic control over a 3-4 month period [ 1,2]. Other glycated protein measurements, including fructosamine and total glycated protein or albumin have been recently advocated as indicators of shorter-term integrated glycemia. Despite these clinical practice recommendations, limited data exists correlating changes in glycated protein assays with one another or, more importantly, with HBGM testing records in a pediatric population. To help develop guidelines for our own clinic, we studied changes in HBGM and glycated protein levels at 2 week intervals in children and adolescents with IDDM.

Materials and Methods This study was approved by the institutional review boards for human subjects of Baylor College of Medicine and Texas Children’s Hospital. Twenty-six children and adolescents with IDDM, aged 7-21 yr, 14 female, were selected from the routine outpatient population of the Diabetes Care Center at Texas Children’s Hospital. Entry criteria included an elevated glycohemoglobin (> 117; by boronate affinity chromatography) and personal motivation for improved glycemic control. HBGM records and peripheral venous blood samples were obtained at 2 week intervals for a total 12 week period. Blood samples were stored at - 70°C until assay. All patients used a OneToucbrM glucose meter for HBGM. The internal memory capacity of this meter permits verification of patient records. Patients were encouraged to obtain preprandial readings 2 to 4 times daily; before major meals and before the bedtime snack. Insulin doses were adjusted and blood testing technique was evaluated at each visit as needed (MRO).

ASSU_YS

All assays were performed in the Texas Children’s Hospital Clinical Pathology Laboratory. A random plasma glucose was measured at each visit using a glucose oxidase method (Y SI Inc, Yellow Springs, OH). Glycated hemoglobin was measured by 2 different methods: boronate affinity chromatography (GHb) and hemoglobin A,C (HAC) by high-performance liquid chromatography (HPLC); and glycated serum proteins were measured by boronate afhnity chromatography (GP) and using the fructosamine (FA) assay. All assays were performed on a single venous blood sample at each visit. The GHb assay relies on the affinity of the cis-diol group, present in stable glycated hemoglobin, for boronate anion [4]. For this assay, 0.05 ml well-mixed EDTA blood was added to 1.O ml water and vortexed for 15 s. After removal of cell debris by centrifugation (3000 rpm, 5 min), 100 ~1 of the hemolysate was applied to a preequilibrated 2 ml boronate affinity column (Endocrine Sciences Laboratory, Calabasas Hills, CA). Nonglycated hemoglobin was eluted with 2.5 volumes (5 ml total) of wash buffer, and the collected fraction was then diluted 1: 1 with an additional 5 ml wash buffer. The glycated fraction was then eluted with one volume (2 ml) of elution buffer containing the counter-ligand, sorbitol. Hemoglobin-specific spectrophotometric absorbance in each fraction was then determined at 414 nm against a water blank and the percent GHb was calculated as lOOa/(5b + a), where a and b are absorbance units for the glycated and nonglycated fractions, respectively. Inter- and intraassay coefficients of variation were 2.8 and 1.3% at GHb = 17 “4. The manufacturer’s nondiabetic range for children (2-20 yr) is 4.5-6.096, while the adult range is 4.2-6.8 “//,. Forty-one nondiabetic children and adolescents tested in our laboratory had a mean level of 5.3 _+0.1 %, range 3.0-6.74,. The assay is not affected by hemoglobin variants. HAC was measured by high-performance liquid chromatography (HPLC) using a method developed at this institution for rapid (30 min)

55

high-resolution hemoglobin separation [ 51. This assay depends on ionic charge changes which occur during nonenzymatic glycation, resulting in altered affinity for the weak cation exchange matrix. In brief, packed erythrocytes were mixed with 2 vol double-distilled water for 30 s, and 50 ~1 of the hemolysate was added to 1 ml 40 mmol/l BisTris, pH 6.5, 4 mmol/l KCN (mobile phase A), and 20 ~1 was then applied to a 20 x 0.46 cm 5 pm microparticulate poly(aspartic acid)-silica (PolyCAT A) column (Custom LC, Houston, TX) pre-equilibrated with 22% mobile phase B (40 mmol/l Bis-Tris, pH 6.8, 4 mmol/l KCN, 0.2 M/l NaCl). The sample was eluted at a flow rate of 1 ml/min, with the concentration of mobile phase B increased linearly to 56% and 100% at 16 and 22 min, respectively, then decreased to 22yb at 24 min. The HPLC apparatus includes a Varian Model 5060 ternary HPLC with Vista 402 Data System interface (Varian Instruments, Palo Alto, CA), an automatic sample injector (WISP 712, Waters Assoc., Milford, MA) and a Waters Model 440 UV detector set at 405 nm wavelength. HAC was calculated as the a ratio of the areas under the curve: lOO(A,.)/(A,. + A). Nondiabetic samples (n = 100) had an HAC level of 4.5 + 1.0 SD, range 2.5-6.5, coefficient of variation = 2.3 %. The sample chromatogram in Fig. 1 illustrates the excellent resolution of hemoglobin F, which interferes with most other cationexchange HAC methods [6]. Serum fructosamine was measured using the RoTAGTM kit (Roche Diagnostic Systems, Nutley, NJ) on a COBAS BIO centrifugal analyzer [ 7,8]. This assay measures the ability of the ketoamine bond in glycated proteins to reduce nitroblue tetrazolium. In brief, 20 1.11serum was mixed with 50 ~1 diluent and 200 ~1 nitroblue tetrazolium, 0.25 mmol/l in 100 mmol/l sodium carbonate, pH 10.35. Absorbance at 550 nm wavelength was measured at 10 and 15 min, and the change was compared to glycated protein calibrators standardized against l-deoxy-l-morpholino-D-fructose (DMF). The nondiabetic reference range is 1.5-2.7 mmol/l DMF equivalents; inter- and intra-assay coefficients of var-

L’ 20

’

3b

Fig. 1. Hemoglobin chromatogram using PolyCAT A HPLC. The sample was from a child with IDDM and persistent hemoglobin F. The bottom axis shows retention time (min). Hemoglobin peaks are labelled as A,,, F, A, and A?. The peaks under the area labelled ‘1’ are A,, and A,,. Peak 2 is acetyl F and peak 3 is an incompletely identified glycated A subfraction. For this patient, areas under the curve for A ,c and A were 489894 and 3 316 880 counts and HAC was 12.9”/,.

iation for the COBAS BIO are 3.67, and 1.2% at both 3.2 mmol/l and 5.8 mmol/l. Total glycated serum proteins (GP) were measured using a kit assay (Glyc-AffinTM GP, IsoLab Inc, Akron, OH), which is based on the same principle as the GHb assay described above [ 9,101. In brief, 50 ~1 of serum was mixed with 100 ~1 of an asparagine buffer and applied to a

56

pre-equilibrated 2 ml agarose-immobilized boronic acid column. After 10 min, the nonglycated proteins were eluted with 3 ml of the same buffer, and the eluate was further diluted to 10 ml with distilled water. The glycated proteins were then eluted with 2 ml of buffer containing sorbitol. The protein in each of the 2 fractions was measured by spectrophotometry after reaction with pyragallol red. The percentage GP was then calculated as lOOa/(5b + a), where a and b are absorbance units for the glycated and non-glycated fractions, respectively. The manufacturer’s recommended assay ranges for plasma are 9-17:; and 14-25”” for nondiabetic and diabetic samples, respectively. Inter- and intra-assay coefficients of variation are stated as 2.02-3.057, and 1.44-2.65%, respectively, across the range of the samples. In our laboratory, nondiabetic serum values were 14.0 + 0.9 (SD) (n = 6, cv = 6.5”,) and 14.1 -t_0.4 (n = 5, cv = 3.0?,,) for previouslyfrozen and fresh samples, respectively. Statistical ana!ysis

Descriptive data are expressed as the arithmetic mean and standard error (SE). Mean blood glucose (MBG) values are calculated from the HBGM records over the specified time periods. Correlations between variables were analyzed by Pearson’s r, and relationships between variables were analyzed by regression analysis with calculation of 95 y0 prediction intervals [ 111. Differences between groups of data were analyzed by paired t-test with a one-tail probability distribution. Significance was assigned at P < 0.05. Data were analyzed using Microstat-II, rel. 2.5 (Ecosoft Inc., Indianapolis, IN) on a personal computer.

First visit 12 week visit

GHb

HAC

FA

GP

17.6 t 0.8

12.2 k 0.6

5.6 f 0.3

27.0 + 1.0

13.2 + 0.6

9.6 f 0.4

4.6 + 0.2

22.4 i 0.8

All differences were significant at P < 0.0001 by paired r-test.

these, 14 sets of 2 week interval data were excluded due either to nonavailability (e.g. at the entry visit) or inconsistencies between the meter memory and the written values. Mean + SE number of HBGM readings for each 2 week period was 35 k 2. All data sets had complete GHb, HAC and GP determinations; one FA measurement was not available. Average glycated protein levels for the first and last visits are shown in Table 1 for the 22 patients who completed all 6 visits. All 4 glycated proteins were significantly lower (P < 0.0001 by paired r-test) at the 12 week visit. A correlation matrix of R values for 2 week interval GHb, HAC, FA, GP, MBG and the random plasma glucose (GLU) is shown in Table 2. All cross correlations of the 4 glycated proteins and 2 week MBG were significant at P < lo- ‘O. The random glucose in the venous blood sample was apparently correlated with GHb (P = 0.003) HAC (P = 0.0004), FA (P = 0.02) and MBG (P = 0.003); but not with GP (P = 0.11). However, multiple regression analysis with the glyTABLE 2 Correlation intervals

matrix (R values) for study variables at 2 week

GHb

HAC

FA

GP

MBG

GHb

1.00

-

-

_

-

HAC FA GP MBG GLU

0.93 0.79 0.80 0.69 0.35

1.00 0.74 0.76 0.65 0.39

1.00 0.75 0.61 0.23

1.00 0.61 0.17

1.00 0.28

Results

Of the 26 subjects, 23 completed all 6 visits and 2 completed 3 consecutive visits. One patient was noncompliant after the first visit and this data is included only as appropriate. HBGM data were considered valid for 22 of the 26 patients: comprising 126 potential 2 week intervals. Of

1

TABLE

Comparisons of initial and final glycated protein measurements for the 22 patients completing all 6 study visits

See text for discussion

of statistics.

GLU

I .oo

57

cated protein (GHb, HAC or FA) as the dependent variable and MBG and GLU as the dependent variables revealed that virtually all of the correlation could be accounted for by MBG. The GLU value was not included in the MBG calculation, therefore the relationship between GLU and MBG is probably due to their independent correlations with glycemic control. Fig. 2 shows the linear regression lines, 95% prediction intervals and equations for MBG against the 4 glycated protein measurements. With this cross-sectional data analysis, all 4 glycated proteins were significantly correlated with 2 week MBG. The slopes of the regression lines were steepest for GHb and GP. To examine 2 week trends, changes (A) during each 2 week interval were examined for each of the study variables and compared to 2 week changes in MBG. As shown in Fig. 3, AFA and AGP were correlated with AMBG, but the correlation of AGHb and AHAC with AMBG did not reach statistical significance and 95 yO prediction intervals were large for all 4 glycated proteins.

40

20

g

;;;&

E

5 . 0

,---, 5

15

Mean

20

25

. 0

Blood Glucose

Despite the failure of AGHb and AHAC to significantly correlate with AMBG, changes in the 4 glycated proteins were correlated with one another over 2 week intervals, as is shown in Table 3. To estimate the effect of duration of glycemia on glycated protein levels, MBG data were recalculated to give cumulative averages in 2 week

TABLE

3

Correlation matrix (R values) for changes in study variables over 2 week intervals

~~~~

,. 10

AMean

Fig. 3. Correlations of changes in MBG (AMBG) with changes in glycated proteins over 2 week intervals. Regression lines (solid) and 955,,, prediction interval boundaries (interrupted lines) are shown. Statistical analyses are as follows: graph A: AGHb = 0.10 (AMBG) - 0.69, R = 0.18, P = 0.09; graph B: AHAC = 0.09 (AMBG) - 0.40, R = 0.21, P = 0.05; graph C: AFA = 0.054 (AMBG) - 0.12, R = 0.25, P = 0.02; graph D: AGP = 0.24 (AMBG) - 0.060. R = 0.32, P = 0.002.

,,:’ 5

10

15

20

AGHb

AHAC

1.00 -

-

AFA

AGP

AMBG

25

Blood Glucose (mmol/L)

Fig. 2. Correlations of MBG with glycated proteins: crosssectional 2 week interval data. Regression lines (solid) and 95:,, prediction interval boundaries (interrupted lines) are shown. Statistical analyses are as follows: graph A: GHb = 0.59 (MBG) + 7.43, R = 0.47. P < lo- ‘“; graph B: HAC = 0.39 (MBG) + 5.69, R = 0.43, P = 5.3 x lo- “; FA = 0.13 graph C: (MBG) + 3.19, R = 0.37, P = 9.8 x IO- “; graph D: GP = 0.68 (MBG) + 15.12, R = 0.37. P = 9.0 x lo- I’.

AGHb AHAC AFA AGP AMBG

See text

R P R P R P R P R P

0.48 2x10 0.37 4x10

1.00

_

’

1.00 -

-

’

0.24 0.02

_

0.34 0.0003

0.49 1x10”

1.00 -

-

0.50 4x 10 1’ 0.18

0.21

0.25

0.32

1.oo

0.09

0.053

0.02

0.002

-

for discussion

of statistics.

58 TABLE 4 Correlations

of cumulative MBG in 2 week increments

and glycated protein measurements _.

Cumulative time period (weeks)

N GHb

HAC

FA

GP

R P A B R P A B R P A B R P A B

2

4

6

8

10

22 0.70 0.0003 0.53 9.3 0.65 0.00 1 0.37 6.6 0.71 0.0002 0.14 3.3 0.65 0.001 0.63 17.0

21 0.70 0.0004 0.42 9.6 0.71 0.0003 0.31 6.9 0.44 0.04 0.07 3.9 0.52 0.02 0.42 18.5

21 0.66 0.00 1 0.51 7.8 0.61 0.003 0.32 6.2 0.40 0.07 0.07 3.8 0.45 0.04 0.48 16.7

20 0.60 0.005 0.60 6.5 0.62 0.003 0.43 4.6 0.33 0.16 0.08 3.5 0.45 0.05 0.62 14.6

19 0.64 0.003 0.75 4.6 0.53 0.02 0.43 4.4 0.65 0.02 0.19 2.2 0.54 0.02 0.81 12.4

12 9 0.46 0.22 0.53 6.6 0.64 0.07 0.50 3.9 0.58 0.10 0.20 2.6 0.43 0.30 0.67 14.6

Table shows number ofpatients at each cumulative period, and R and P values for the linear regression equations, for the equation [glycated protein] = A{MBG} + B.

increments. Data were re-aligned to start with the first available MBG for each patient. The cumulative average was then compared with the glycated protein measurements at the end of each cumulative period. Results of this analysis are shown in Table 4. All 4 glycated proteins were significantly correlated with MBG over each cumulative incremental period up to 10 weeks. At 12 weeks, the correlation probabilities were not significant due to the decreased number of observations. Fig. 4 compares the slopes of the cumulative data correlations at 2, 4 and 10 week intervals. At each interval, GHb and GP showed a larger change (e.g. steeper slope) with MBG, while the slope of MBG vs FA was relatively shallow.

Discussion

Monitoring of glycemic control in IDDM has been facilitated by the nearly concurrent develop-

ment of techniques for HBGM and glycated protein assays. Early studies of glycated hemoglobin levels showed a significant correlation with laboratory blood glucose in diabetes mellitus [ 12,131; supporting the assumption that non-enzymatic glycation of hemoglobin reflects blood glucose levels. However, the clinical application of glycated protein levels and their relationship to HBGM is not well defined, although these tests are routinely used for assessment of glycemic control. Our current investigations rely upon HBGM readings obtained using a OneTouchTM glucose meter, a device which shows excellent correlation with laboratory methods and, perhaps more importantly, a low user-error rate [3,14, unpublished data]. Therefore, the accuracy of the HBGM data is typical for a reasonably compliant and motivated pediatric population; however, the absolute numerical correlations may be limited to values obtained with this meter.

59

5

_________--

_______--

0

2

FA

4

6

6

10

25

6 wk 20

GP

0

20

2

4

6

6

10

10wk

Fig. 4. Regression lines for glycated protein measurements and 2 week incremental MBG at 2,6 and 10 week cumulative intervals. Glycated protein levels are plotted against the y-axis. Statistical information for these graphs is presented in Table 4.

The glycated protein assays were chosen to give theoretical measures of both short (FA, GP) and longer term (GHb, HAC) glycemic control. Since the turnover of hemoglobin in the peripheral circulation is - 120 d (Ti,* N 60 d) [ 151, glycated hemoglobin measurements are thought to be most useful when obtained at 2-4 month intervals. For shorter-term measurement of glycemic control, glycated serum protein assays have been developed. The major serum protein is albumin, which constitutes 50-70 y0 of total serum protein and has a T,,* of 20 d. The other 30-50% includes a variety of proteins with T,,,‘s ranging from < 1 to > 20 days [ 161. The FA and GP

assays both measure total glycated serum proteins. Previous studies have shown a good but variable correlation of glycated serum protein assays with one another and with blood glucose levels, and a few studies have addressed the utility of glycated serum protein assays for monitoring short-term glycemic control [ 17-251. A 6 week study of 100 IDDM patients found that changes in fasting blood glucose were best correlated with glycated serum albumin, and were poorly correlated with HAC and FA [26]. This study had several limitations, including the reliance on fasting blood glucose, which may not reflect overall glycemic control in IDDM, and measurement of HAC only at 0 and 6 weeks, while other assays were monitored at 0, 2,4 and 6 weeks. It is interesting to note, however, that we also found a better correlation of AMBG with AGP than with AFA (the GP and glycated serum albumin assays use the same method). Cefalu et al. [27] studied 97 pregnant women, 13 with gestational diabetes mellitus, and found correlations of FA with both fasting blood glucose and 2 week MBG. HAC, measured by open-column cation exchange chromatography, did not correlate with either glucose measurement. Neither HAC nor FA provided significant discrimination between diabetic and nondiabetic groups. Our composite data (Table 2 and Fig. 2) also show good correlation of glycated proteins with one another and MBG. Over 2 week intervals (Table 3 and Fig. 3) AFA and AGP showed relatively good correlation with AMBG. However, calculations based on this data indicate that FA has limited clinical utility in detecting changes in MBG. For a 2 week AMBG = 5.6 mmol/l ( 100 mg/dl), the predicted A FA would be 0.18 ; or 3.6% for an initial FA = 5.0 mmol/l. This degree of change approximates the interassay coefficient of variation for the FA assay. On the other hand, AGP would be predicted at 1.28, or 6% for an initial GP = 20%, a change which is twice the interassay coefficient of variation. observed for individual Actual changes patients were greater than predicted from the

60

composite data. We postulated that this might be due to the cross-sectional nature of the data sampling, and that 2 week incremental cumulative data (Table 4) may be more indicative of longitudinal trends. Using the 6 week cumulative interval, which approximates albumin turnover, the predicted FA value would decrease from 4.54 to 4.17, or 8.27,, for a MBG change from 11.2 to 5.6 mmol/l. Corresponding predicted changes in other glycated proteins are HAC: 18.3%, GHb: 20.79/, , GP: 12.2%. The predicted changes represent significant changes in the individual assay values, and are in agreement with what was actually observed for individual patients. For instance, over five 2 week intervals, patient # 6 had cumulative MBG values of 17.3, 15.1, 12.5, 11.1 and 9.9 mmol/l. Corresponding glycated protein FA(7.0, 5.9, 4.7, 3.8 and values were: 3.6mmol/l}, GP(33.2, 27.7, 21.8, 19.4 and 18.2Y;j, HACi13.2, 12.8, 11.3, 8.6 and 7.4”z;j and GHb{ 18.7, 17.7, 14.5, 11.3 and 9.49,). Over each incremental interval, predicted changes for a given change in MBG were greatest for GP and GHb. If near normoglycemia is a goal of IDDM therapy, then a reasonable clinical goal might be to normalize glycated protein levels, which represent integrated glucose concentrations. However, since typical HBGM monitoring patterns measure only preprandial glucose levels and postprandial glucose excursions may be quite wide in IDDM, this goal carries a risk of frequent hypoglycemia. From our current data, in order to achieve a GHb = 7% (=HAC 6%), the MBG should approximate 3.2 mmol/l based on the regression equation for the 10 week cumulative incremental data (Table 4). Over shorter cumulative periods, which may be more typical of the time period reviewed in clinical practice, the MBG would need to approach 0 mmol/l to achieve a GHb in the upper nondiabetic range. The inability of absolute glycated protein values to predict absolute HBGM levels is further indicated by the fact that none of the glycated protein correlations extrapolate to 0 at MBG = 0. Our data are limited by the short time period, which

probably did not allow achievement of steady state levels of glycated proteins in relation to MBG. A rigorous investigation of this relationship in IDDM has not been published, and would require a duration encompassing the turnover time of the measured proteins. However, our results are in agreement with another report [ 191 that HBGM values cannot be used to predict absolute levels of glycated proteins. Our data and other studies [ 281 also emphasize that glycated protein assays are not directly comparable. For instance, several different methods for glycated hemoglobin measurement are now available. We did not study the open-column cation exchange or calorimetric techniques, both of which have poor specificity and give relatively poor discrimination between diabetic and nondiabetic samples, probably due to the presence of interfering substances which are more prominent at lower glycated hemoglobin levels [4,29-321. Even for more reliable assays, such as those used for this study, the correlation of glycated hemoglobin assays with one another and the slopes of variance with MBG need to be considered. For instance, a GHb of 19% would approximate an HAC of 14 % . Furthermore, from the slopes of the regression lines, GHb would be expected to show a 507, greater change than HAC for a given change in MBG (see Table 3). Therefore, when comparing glycated hemoglobin levels, it is important to know the methods used for the assays. Other assay limitations also need to be considered. None of the 4 methods tested were independently influenced by the ambient glucose concentration in the sample, indicating that labile glycation intermediates were not measured. The boronate al%nity chromatography method used for GP and GHb assays has been shown to be very specific for measurement of stable glycated protein. However, the FA method is not specific for glycated proteins, and as much as 2/3 of estimated serum FA is due to substances other than glycated proteins [33-361. This may account for the poor results for FA reported in some studies. Furthermore, since both the FA and GP assays measure several different proteins, results may

61

depend on the relative concentrations of the major serum proteins. HAC measurements are highly dependent on the chromatographic resolution of the hemoglobin subfractions and many commercial methods perform poorly in the presence of labile glycation intermediates and hemoglobin variants [ 29,301. This is the first report using the PolyCAT A method for HAC determination. This method appears to have several advantages over other cation-exchange techniaues. including excellent discrimination of hemoglobin variants, and may be a good option for laboratories requiring an automated procedure. In conclusion, we have compared HBGM and 4 glycated protein measurements at 2 week intervals over 12 weeks in patients with IDDM. Changes in all 4 glycated proteins show good correlation with short-term changes in HBGM. However, with usual preprandial HBGM testing patterns, the predicted absolute MBG levels needed to normalize glycated protein levels may be in the hypoglycemic range with very wide prediction intervals; a finding which requires confirmation in longer-term studies. Our results indicate that of the 4 glycated protein assays studied, the 2 methods based on boronate-affinity chromatography, GHb and GP, may provide the most clinically useful indicators of changes in glycemic control. Furthermore, when compared to the other assays, FA appears to have little clinical value. Clinicians should be aware of the differences and limitations of the assays, particularly when defining patient treatment goals.

References American Diabetes Association (1989) Position statement on standards of medical care for patients with diabetes mellitus. Diabetes Care 12, 365-368. American Diabetes Association (1987) Consensus statement on self-monitoring of blood glucose. Diabetes Care 10,95-99. Blood glucose monitors (1988). Health Care Devices 17, 253-271.

4 Glycosylated Hemoglobin (1983). Endocrine Sciences, Tarzana, CA. 5 Ou, C.N., Buffone, G.J. and Reimer. G.L. (1983) Highperformance liquid chromatography of human hemoglobins on a new cation exchanger. J. Chromatogr. 266, 197-205. 6 Goldstein, D.E., Little, R.R., Wiedmeyer, H.M., England, J.D. and McKenzie, E.M. (1986) Glycated hemoglobin: methodologies and clinical applications. Clin. Chem. 32, B64-B70. 7 Johnson, R.N., Metcalf, P.A. and Baker, J.R. (1982) Fructosamine: a new approach to the estimation of serum glycosylprotein. An index of diabetic control. Clin. Chim. Acta 127, 87-95. 8 RoTAGTM fructosamine (glycated protein) assay. Package insert (1989). Roche Diagnostic Systems, Nutley, NJ. 9 Glyc-Affin TM GP. Package insert (1990) IsoLab Inc., Akron, OH. 10 Yatscoff, R.W., Tevaarwerk, G.J.M. and MacDonald, J.C. (1984) Quantification of nonenzymatically glycated albumin and total serum protein by affinity chromatography. Clin. Chem. 30, 446-449. 11 Altman, D.G. and Gardner, M.J. (1988) Calculating confidence intervals for regression and correlation. Br. Med. J. 296, 1238-1242. 12 Koenig, R.J., Peterson, CM., Jones, R.L., Saudek, C., Lehrman, M. and Cerami, A. (1976) Correlation of glucose regulation and hemoglobin A,, in diabetes mellitus. N. Engl. J. Med. 295,417-420. 13 Paulsen, E.P. and Koury, M. (1976) Hemoglobin A,, levels and insulin-dependent diabetes mellitus. Diabetes 25, 890-896. 14 Jovanovic-Peterson, L., Peterson, C.M., Dudley, J.D., Kilo, C. and Ellis, B. (1988) Identifying sources of error in self-monitoring of blood glucose. Diabetes Care 11, 79 l-794. 15 Bunn, H.F. (1972) Erythrocyte destruction and hemoglobin catabolism. Semin. Hematol. 9, 3-17. 16 Lentner, C. (Ed.) 1984 Geigy Scientific Tables, Volume 3, Ciba-Geigy Corp., West Caldwell, NJ, pp. 135-162. 17 Rendell, M., Brannan, C., Nierenberg, J., Rasbold, K. and Hestorff, T. (1987) Fingerstick glycosylated hemoglobin, plasma protein, and albumin. Diabetes Care 10.629-632. 18 Guillausseau, P.J., Charles, M.A., Paolaggi, F. et al. (1990) Comparison of HbA, and fructosamine in diagnosis of glucose-tolerance abnormalities. Diabetes Care 13, 898-900. 19 Service, F.J., O’Brien, P.C. and Rizza, R.A.. (1987) Measurements of glucose control. Diabetes Care 10, 225-237. 20 Sobel, D.O. and Abbasi, A. (1991) Use of fructosamine test in diabetic children. Diabetes Care 14, 478-581. 21 Shima, K., Abe, F., Chikakiyo, H. and Ito. N. (1989) The relative value of glycated albumin, hemoglobin A,, and fructosamine when screening for diabetes mellitus. Diabetes Res. Clin. Pratt. 7, 243-250.

62 22 Mullis, P., Viilkle, H., Sigrist, V. and Zuppinger, K. (1988) Comparison of fructosamine and glycated haemoglobin in children with Type I (insulin-dependent) diabetes mellitus. Diabetologia 3 1, 35 l-354. 23 Baker, J.R., Johnson, R.N. and Scott, D.J. (1984) Serum fructosamine concentrations in patients with type II (non-insulin-dependent) diabetes mellitus during changes in management. Br. Med. J. 288, 1484-1486. 24 Guillauseau, P.J., Charles, M.A., Godard, V. et al. (1990) Comparison of fructosamine with glycated hemoglobin as an index ofglycemic control in diabetic patients. Diabetes Res. 13, 127-131. 25 Smart, L.M., Howie, A.F., Young, R.J., Walker, S.W.. Clarke, B.F. and Smith, A.F. (1988) Comparison of fructosamine with glycosylated hemoglobin and plasma proteins as measures of glycemic control. Diabetes Care 11, 433-438. 26 Winocour, P.H., Bhatnagar, D., Kalsi, P., Hillier, V.F. and Anderson, D.C. (1989) Relative clinical usefulness of glycosylated serum albumin and fructosamine during short-term changes in glycemic control in IDDM. Diabetes Care 12, 665-672. 27 Cefalu, W.T., Prather, K.L., Chester, D.L., Wheeler, C.J., Biswas, M. and Pernoll, M.L. (1990) Total serum glycosylated proteins in detection and monitoring of gestational diabetes. Diabetes Care 13, 872-875. 28 Forrest, R.D., Jackson, C.A., Gould, B.J., Casburn-Budd, M., Taylor, J.E., and Yudkin, J.S. (1988) Four assay methods for glycated hemoglobin compared as screening tests for diabetes mellitus: the Islington diabetic survey. Clin. Chem. 34. 145-148.

29 Allen, L.C. (1985) Effects of hemoglobin variants on Hb A,, determinations using Bio-Rad columns. Clin. Biochem. 18, 173-175. 30 Papadea, C., Austin, G.E. and Mullins, R.E. The effect of storage conditions on ion exchange and affinity chromatographic assays for glycated hemoglobin. Clin. Biochem. 17, 296-301. 31 Pecoraro, R.E., Graf, R.J., Halter, J.B., Beiter, H. and Porte, D., Jr. (1979) Comparison of a calorimetric assay for glycosylated hemoglobin with ion-exchange chromatography. Diabetes 28, 1120-l 125. 32 Klenk, DC., Hermanson, G.T., Krohn, R.I. et al. (1982) Determination of glycosylated hemoglobin by affinity chromatography: comparison with calorimetric and ionexchange methods, and effects of common interferences. Clin. Chem. 28, 2088-2094. 33 Desjarlais, F., Comtois, R., Beauregard, H., Nguyen, M. and Lettelier, G. (1988) Technical and clinical evaluation of fructosamine determination in serum. Clin. Biochem. 22, 329-335. 34 Fliickiger, R., Woodtli, T. and Berger, W. (1987) Evaluation of the fructosamine test for the measurement of plasma protein glycosylation. Diabetologia 30, 648-652. 35 Schleicher, E. and Wieland, O.H. (1989) Protein glycation: measurement and clinical relevance. J. Clin. Chem. Clin. Biochem. 27, 577-587. 36 Rosenthal, M.A. and Gallop, P.M. (1990) Elimination of nonglycation-derived interferences in the fructosamine assay. Clin. Chem. 36, 1114.