0021-972X/01/$03.00/0 The Journal of Clinical Endocrinology & Metabolism Copyright © 2001 by The Endocrine Society
Vol. 86, No. 4 Printed in U.S.A.
Comparison of Measured and Estimated Indices of Insulin Sensitivity and  Cell Function: Impact of Ethnicity on Insulin Sensitivity and  Cell Function in Glucose-Tolerant and Normotensive Subjects* KEN C. CHIU, LEE-MING CHUANG,
AND
CAROL YOON
Division of Endocrinology, Diabetes, and Hypertension (K.C.C., C.Y.), Department of Medicine, University of California–Los Angeles, School of Medicine, Los Angeles, California 90095-7097; and Department of Internal Medicine and Graduate Institute of Clinical Medicine (L.-M.C.), National Taiwan University Hospital, 10016 Taipei, Taiwan ABSTRACT Type 2 diabetes mellitus is the result of an imbalance between insulin sensitivity and  cell function. Although the assessment of these 2 parameters is critical for various studies, the current methods are time consuming and labor intensive. Recently, new estimated indices have been proposed. We examined the impact of ethnicity on the indices of insulin sensitivity and  cell function measured from the hyperglycemic clamp and compared the results to the estimated indices, proposed by Matsuda and DeFronzo and Stumvoll et al., from a standard oral glucose tolerance test in 105 healthy, glucose-tolerant, and normotensive subjects from 4 ethnic groups. Among the ethnic groups, differences were noted in the measured insulin sensitivity (P ⫽ 0.0006) and  cell function (P ⫽ 0.006 for the first phase insulin response, P ⫽ 0.0002 for the second phase insulin
T
YPE 2 DIABETES mellitus (or noninsulin-dependent diabetes mellitus, NIDDM) is a heterogeneous syndrome resulting from a combination of insulin resistance and  cell dysfunction (1). Although insulin resistance may not be required for the development of NIDDM (2, 3), defects in insulin sensitivity and  cell function have been demonstrated in most subjects with NIDDM (4, 5). Furthermore, insulin resistance has been demonstrated in nondiabetic offspring of two diabetic parents (6) and decreased  cell function has been detected in glucose-tolerant subjects with family histories of NIDDM (7). In addition, both insulin sensitivity and  cell function have been demonstrated as being inherited traits (8 –11). Therefore, assessment of insulin sensitivity and  cell function is essential if one plans to study the genetic influences of these traits in NIDDM. Although various methods have been developed for the measurement of insulin sensitivity and  cell function, they are not only Received August 81, 2000. Revision received October 23, 2000. Rerevision received December 6, 2000. Accepted January 4, 2001 Address all correspondence and requests for reprints to: Ken C. Chiu, M.D., F.A.C.E., 675 Charles E. Young Drive South, 4629 MacDonald Research Laboratories, Los Angeles, California 90095-7097. E-mail:
[email protected]. * This work was supported in part by grants from USPHS (MO1RR00865; University of California, Los Angeles-General Clinical Research Center), NIH/NIDDK (RO1DK52337-01; to K.C.C.), Diabetes Action Research and Education Foundation (to K.C.C.), and American Diabetes Association (to K.C.C.).
response). Although the estimated indices correlated with the measured indices (r2 ⫽ 0.5184 – 0.3014), the estimated indices barely detected the differences among the ethnic groups. Multivariate analysis confirmed that ethnicity had an independent impact for the measured indices, but had only a modest impact on the estimated insulin sensitivity indices and had no impact on the estimated indices of  cell function. We conclude that although the estimated indices of insulin sensitivity and  cell function from the oral glucose tolerance test correlated with the measured ones in a wide spectrum of healthy, glucosetolerant, and normotensive subjects, they were much less likely to detect the differences than measured ones among the ethnic groups. (J Clin Endocrinol Metab 86: 1620 –1625, 2001)
time consuming, but also labor intensive. Thus, it is desirable to develop more practical methods for the estimation of insulin sensitivity and  cell function. The oral glucose tolerance test (OGTT) was originally developed as a research tool and was subsequently adapted and standardized as a diagnostic tool for diabetes (12). It would be ideal and time efficient if one could classify the state of glucose tolerance and estimate insulin sensitivity and  cell function from this single test. In addition to numerous attempts in the past, two groups recently presented their equations based on relatively large-scale studies (13, 14). Matsuda and DeFronzo (13) developed an estimated insulin sensitivity index (ISIM) obtained from the OGTT through comparison with the euglycemic clamp. Stumvoll et al. (14) used the results of the OGTT to calculate an estimated insulin sensitivity index (ISIS) by comparing with the euglycemic clamp. They also estimated the first-phase insulin release (1stPHS) and second-phase insulin release (2ndPHS) by comparison with the hyperglycemic clamp (14). Both groups showed excellent correlation with the measured indices (13, 14). These surrogate measures are easier, less invasive, and cheaper to employ and, therefore, can be applied more readily to a large number of subjects. However, the correlation between these surrogates and the direct measurements of insulin action and  cell capacity are less than perfect (13, 14). It remains to be seen whether these indices with the loss of information as the results from their less-than-perfect cor-
1620
INSULIN SENSITIVITY AND  CELL FUNCTION
relation with direct measures of these indices can have adequate statistical power for the studies desired. In the present study, we recruited 105 glucose-tolerant subjects for the assessment of insulin sensitivity and  cell function using a hyperglycemic clamp. To examine the performance of the estimated indices, we first examined the impact of ethnicity on the measured indices of insulin sensitivity and  cell function and compared these results against the estimated indices. Subjects and Methods Study protocol To minimize confounding factors, only healthy subjects who received no regular medical treatments were invited to undergo a screening test as outpatients after an overnight fast. It included a standard OGTT with 75 g glucose and a brief physical examination as previously described (15). To exclude the secondary influence of insulin sensitivity and  cell function from abnormal glucose tolerance (16), only those who were noted to be glucose tolerant (fasting plasma glucose (FPG) less than 6.1 mm, interval plasma glucose less than 11.1 mm, and 2-h plasma glucose less than 7.7 mm) were invited back for the assessment of  cell function and insulin sensitivity using a hyperglycemic clamp technique. Because hypertension has been shown to be associated with insulin resistance (17), only normotensive (less than 140/90 mm Hg) subjects were enrolled in the study. For safety reasons, anemic (hemoglobin less than 11.0 g/dL) subjects were also excluded from the hyperglycemic clamp of the study. To minimize the effect of smoking, they were asked to refrain from smoking for at least 12 h before the study. Briefly, after fasting overnight and resting in the General Clinical Research Center of the University of California, Los Angeles (Los Angeles, CA), participants received a bolus of 50% dextrose solution based on their body surface area (11.4 g/m2) at 0 min. Continuous infusion of 30% dextrose solution was commenced at 15 min at variable rates, which were adjusted every 5 min based on the prevailing plasma glucose levels, to maintain a plasma glucose level around 10 mm toward 180 min using the negative feedback principle (18). The insulin sensitivity index (ISI) was calculated by dividing the average glucose infusion rate during the last 60 min of the clamp by the average plasma insulin level. Glucose clearance (GCl) was calculated as ISI divided by the plasma glucose concentration (7). The coefficient of variation for steady-state plasma glucose levels was 5.6 ⫾ 0.2%. The first-phase insulin response (1stIR) was the sum of plasma insulin levels during the first 10 min (2.5, 5, 7.5, and 10 min) and the second-phase insulin response (2ndIR) was the average of plasma insulin levels at 130, 140, 150, 160, 170, and 180 min. FPG and fasting plasma insulin (FPI) concentrations were the average of 3 samples before the OGTT. Plasma glucose, insulin, and lipid were assayed as previously described (19). The demographic features of the studied subjects are shown in Table 1. There were 28 Asian-, 11 African-, 46 Caucasian-, and 20 Mexican-Americans. Because birth control pills have been shown to affect insulin sensitivity (20), gender was categorically classified as male, female taking birth control pills, and female without birth control pills. Ethnicity was defined as the reported ethnicity by each participant. The
1621
study was approved by the Human Subject Protection Committee of University of California, Los Angeles. Written informed consent was obtained from all of the participants before entering the study. We confirm that the study has complied with the recommendations of the Declaration of Helsinki.
Estimation of insulin sensitivity and  cell function The estimated ISI proposed by Matsuda and DeFronzo (13) was calculated based on the following formula: ISIM ⫽ 10,000 ⫻ (FPG ⫻ FPI ⫻ mean OGTT glucose concentration ⫻ mean OGTT insulin concentration)0.5. Plasma glucose concentration was in milligrams per dl and plasma insulin concentration was in microunits per ml for this estimate by Matsuda and DeFronzo (13). The estimated ISI proposed by Stumvoll et al. (14) was calculated based on the following formula: ISIS ⫽ 0.226 ⫺ (0.0032 ⫻ body mass index (BMI)) ⫺ (0.0000645 ⫻ plasma insulin concentration at 120 min) ⫺ (0.0037 ⫻ plasma glucose concentration at 90 min). The estimates of  cell function proposed by Stumvoll et al. were calculated using the following two formulae: 1stPHS ⫽ 1283 ⫹ (1.829 ⫻ plasma insulin concentration at 30 min) ⫺ (138.7 ⫻ plasma glucose concentrations at 30 min) ⫹ (3.772 ⫻ FPI) for the 1stIR; and 2ndPHS ⫽ 287 ⫹ (0.4164 ⫻ plasma insulin concentration at 30 min) ⫺ (26.07 ⫻ plasma glucose concentration at 30 min) ⫹ (0.9226 ⫻ FPI) for the 2ndIR. These estimations were based on plasma glucose concentrations in milimoles per L and plasma insulin concentrations in picomoles per L (14).
Statistical analysis The continuous variables that failed the Normality test were logarithmically transformed before analysis. The variables transformed were age, BMI, waist to hip ratio, plasma insulin levels, 1stIR, 2ndIR, ISI, GCl, ISIM, ISIS, 1stPHS, and 2ndPHS. Differences in continuous variables between groups of subjects were tested with either one-way ANOVA or Student’s t test when appropriate. Differences in proportions were evaluated by a 2 test. The relationships between variables were determined by using a simple regression analysis. To examine the influence of confounding variables, a stepwise regression analysis was used. Backward stepwise with ␣-to-enter of 0.10 and ␣-to-remove of 0.10 was employed to exclude variables that had little or no influence on the parameter under analysis. SYSTAT 8.0 for Windows package from SPSS, Inc. (Chicago, IL) was used for the statistical analysis. Data were presented as arithmetic means with 95% confidence intervals, unless otherwise specified. A P value of less than 0.05 (two-tailed) was considered significant.
Results Measured insulin sensitivity and  cell function
This study included 105 healthy subjects with normal glucose tolerance and normal blood pressure as shown in Table 1. During the hyperglycemic clamp, a steady-state plasma glucose concentration was achieved with a mean of 10.03 mm (range, 9.33–11.03 mm). There were wide variations in ISI
TABLE 1. Demographic features of studied subjects No.
No. of subjects Gender (female/male) Age (yr) BMI (kg/m2) Waist-hip ratio (cm/cm) Systolic blood pressure (mm Hg) Diastolic blood pressure (mm Hg) OGTT Fasting plasma glucose (mM) Plasma glucose at 30 min (mM) Plasma glucose at 60 min (mM) Plasma glucose at 90 min (mM) Plasma glucose at 120 min (mM)
105 64/41
Arithmetic mean ⫾
Minimum
Maximum
26 ⫾ 1 24.79 ⫾ 0.44 0.793 ⫾ 0.007 115 ⫾ 1 67 ⫾ 1
18 17.58 0.649 92 48
40 37.57 1.033 138 89
4.67 ⫾ 0.04 7.38 ⫾ 0.11 7.15 ⫾ 0.13 6.36 ⫾ 0.13 5.88 ⫾ 0.10
3.30 5.07 4.44 3.62 3.39
5.56 10.10 10.20 9.69 7.72
SE
1622
JCE & M • 2001 Vol. 86 • No. 4
CHIU ET AL.
(geometric mean, 5.2456; range, 1.3632–17.9944 m/m2/ min/pm), 1stIR (geometric mean, 1,637; range, 465–7,415 pm), and 2ndIR (geometric mean, 443; range, 104 –1,567 pm). However, ISI correlated with 1stIR and 2ndIR very well (r2 ⫽ 0.3289 and r2 ⫽ 0.5548, respectively). These results indicate that a reciprocal change occurred between insulin sensitivity and  cell function to maintain glucose homeostasis in these subjects. Simple regression analysis of the measured and estimated insulin sensitivity and  cell function
In the univariate analyses, the measured ISI, 1stIR, and 2ndIR were mutually interrelated as shown in Table 2. Similarly, there were very strong correlations between the measured and estimated insulin sensitivity indices (ISI, ISIM, and ISIS). The relatively large portion of variation in the measured ISI can be explained by ISIM (52%) and ISIS (30%). There was a very close relationship between 1stIR and 1stPHS and one explained 45% of the variation of the other. A very close relationship was also noted between 2ndIR and 2ndPHS, explaining 32% of the variation in each other.
Univariate analyses of the impact of ethnicity on the measured and estimated indices
This study included four ethnic groups of healthy subjects as shown in Table 3. Although minor differences were noted in the clinical features, only differences in age and BMI reached the statistical level (P ⫽ 0.023 for both). During the hyperglycemic clamp, all four ethnic groups achieved similar steady-state plasma glucose concentrations (P ⫽ 0.30), as shown in Table 3. The coefficient of variation for steady-state plasma glucose levels was 5.2 ⫾ 0.5% for Asian-Americans, 5.9 ⫾ 1.0% for African-Americans, 5.8 ⫾ 0.3% for CaucasianAmericans, and 5.1 ⫾ 0.5% for Mexican-Americans. Univariate analysis revealed that there were significant differences in ISI, 1stIR, and 2ndIR among the four ethnic groups (P ⫽ 0.0006, P ⫽ 0.006, and P ⫽ 0.0002, respectively). Asian-Americans had the lowest ISI with the highest 2ndIR and the second highest 1stIR, whereas Caucasian-Americans were most sensitive to insulin with the lowest 1stIR and 2ndIR. Although the other estimated indices showed only marginal differences, ISIS was the only one that reached the statistically significant level (P ⫽ 0.024). However, a discrep-
TABLE 2. Pearson correlation matrix
ISI GCl 1stIR 2ndIR ISIM ISIS 1stPHS 2ndPHS
ISI
GCl
1stIR
2ndIR
ISIM
ISIS
1stPHS
2ndPHS
1.0000 0.9909 – 0.5735 – 0.7449 0.7200 0.5490 – 0.5545 – 0.5586
1.0000 – 0.5774 – 0.7422 0.6865 0.5542 – 0.5635 – 0.5679
1.0000 0.6904 – 0.4869 – 0.2736a 0.6675 0.6609
1.0000 – 0.5273 – 0.3658b 0.5659 0.5671
1.0000 0.6378 – 0.6261 – 0.6594
1.0000 – 0.3152a – 0.3232c
1.0000 0.9966
1.0000
P ⬍ 0.0001, except for where noted. a P ⬍ 0.005. b P ⫽ 0.0001. c P ⬍ 0.001. TABLE 3. Clinical features, insulin sensitivity, and  cell function by ethnicity
No. Gender (female) Agea (yr) Maternal history of diabetes (yes) Paternal history of diabetes (yes) BMIa (kg/m2) Waist-hip ratioa (cm/cm) Systolic blood pressure (mm Hg) Diastolic blood pressure (mm Hg) Hyperglycemic clamp Steady-state plasma glucose (mM) ISIa (M/m2/min/pM) GCla (L/m2/min/pM) 1stIRa (pM) 2ndIRa (pM) Estimated indices ISIMa ISISa 1stPHsa 2ndPHsa Mean (95% CI) or no. (%). a Geometric mean (95% CI).
Asian-American
African-American
Caucasian-American
Mexican-American
28 20 (71%) 23 (21, 25) 2 (7%) 5 (18%) 23.58 (22.36, 24.87) 0.76 (0.74, 0.77) 112 (108, 117)
11 6 (55%) 25 (21, 29) 1 (9%) 1 (9%) 25.08 (21.32, 29.52) 0.80 (0.75, 0.85) 114 (107, 122)
46 25 (54%) 27 (25, 29) 4 (9%) 5 (11%) 24.02 (22.88, 25.22) 0.80 (0.77, 0.82) 116 (112, 119)
20 13 (65%) 25 (23, 28) 1 (5%) 5 (25%) 26.24 (24.14, 28.52) 0.82 (0.79, 0.85) 115 (110, 120)
64 (61, 67)
63 (59, 67)
68 (66, 71)
67 (63, 71)
P
NS 0.023 NS NS NS 0.023 NS NS
10.06 (9.93, 10.19)
9.84 (9.59, 10.09)
10.05 (9.95, 10.15)
10.02 (9.87, 10.17)
NS
4.03 (3.22, 5.04) 0.411 (0.328, 0.514) 1980 (1498, 2617) 569 (463, 698)
5.06 (3.29, 7.76) 0.514 (0.335, 0.789) 1744 (1218, 2496) 494 (339, 718)
6.76 (5.83, 7.83) 0.673 (0.582, 0.779) 1320 (1137, 1533) 343 (296, 397)
4.20 (3.16, 5.60) 0.420 (0.316, 0.558) 1986 (1690, 2335) 533 (431, 659)
⬍0.001 ⬍0.001 0.006 ⬍0.001
4.701 (4.162, 5.310) 0.105 (0.097, 0.112) 1282 (1068, 1539) 337 (286, 396)
4.947 (4.006, 6.109) 0.095 (0.079, 0.114) 1268 (984, 1634) 330 (266, 409)
5.607 (5.007, 6.280) 0.108 (0.102, 0.115) 1091 (999, 1192) 289 (268, 312)
4.629 (3.873, 5.533) 0.091 (0.080, 0.104) 1355 (1161, 1581) 348 (302, 400)
NS 0.024 NS NS
INSULIN SENSITIVITY AND  CELL FUNCTION
1623
TABLE 4. Multivariate analysis of insulin sensitivity and glucose clearance Dependent variable
Covariate entered
Covariate removed
ISI
r2
P
0.4366 Diastolic blood pressure Ethnicity Gender Age Waist-hip ratio BMI
0.0001 0.0004 0.0004 0.0009 0.0113 0.0754 0.6547
Systolic blood pressure GCl
0.4280 Diastolic blood pressure Gender Ethnicity Age Waist-hip ratio BMI
0.0002 0.0003 0.0008 0.0016 0.0149 0.0543 0.7893
Systolic blood pressure ISIM
0.3684 Diastolic blood pressure Waist-hip ratio Gender Age Ethnicity BMI
0.0001 0.0043 0.0148 0.0250 0.0319 0.0358 0.3844
Systolic blood pressure ISIS
0.6614 BMI Gender Diastolic blood pressure Ethnicity Waist-hip ratio Age Systolic blood pressure
ancy was noted between the measured and estimated indices. The order of insulin resistance based on ISI was Caucasian-, African-, Mexican-, and Asian-Americans (being the most resistant); however, the order changed to Caucasian-, Asian-, African-, and Mexican-Americans, when based on ISIS. Multivariate analysis of the measured and estimated indices
A multivariate analysis was used to exclude the influence of confounding factors on insulin sensitivity and  cell function. Ethnicity had an independent impact on ISI (P ⫽ 0.0004) and along with diastolic blood pressure, gender, age, waisthip ratio, and BMI, accounted for 44% of the variation in ISI. Although ethnicity had no impact on ISIM during the univariate analysis, multivariate analysis revealed that ethnicity was a weak but independent determinant for ISIM (P ⫽ 0.0319). Systolic blood pressure was the only analyzed factor that had no impact on both ISI and ISIM. Despite that, up to 66% of the variation in ISIS could be explained by BMI, gender, diastolic blood pressure, and ethnicity. Ethnicity only had a very marginal impact on it (P ⫽ 0.0652). Because ISIS included BMI as one of its parameters in the estimation, BMI by itself accounted for 60% of the variation in ISIS (P ⬍ 0.000001). In contrast, waist-hip ratio, age, and systolic blood pressure had no influence on ISIS. Multivariate analysis revealed that ethnicity (P ⫽ 0.0079)
⬍0.0001 0.0155 0.0617 0.0652 0.1236 0.4292 0.4527
was an independent determinant for 1stIR; ethnicity with BMI accounted for 17% of the variation in 1stIR. Because 2ndIR was a denominator of ISI, the same confounding factors, including ethnicity (P ⫽ 0.0011) determined 36% of the variation of 2ndIR. In contrast, ethnicity had no influence on both 1stPHS and 2ndPHS. Because both of them were derived from FPI and plasma glucose and insulin concentrations at 30 min, they had the same set of covariates with highly similar P values. Both ethnicity and systolic blood pressure were excluded from the multivariate analysis and had very similar insignificant P values (see Table 5). Discussion
The purpose of this study is to examine the abilities of the estimated indices to detect differences compared with the measured indices, because the estimated indices were reported to correlate very well with the measured indices (13, 14). We found significant differences in the measured insulin sensitivity (P ⫽ 0.0006 for ISI) among the four ethnic groups with a reciprocal change in the measured  cell function (P ⫽ 0.006 for 1stIR, P ⫽ 0.0002 for 2ndIR). We confirmed that the estimated indices correlated with the measured indices (r2 ⫽ 0.3014 – 0.5184), although the method of assessment used in the present study (hyperglycemic clamp) was different from those methods (euglycemic and hyperglycemic clamps) used previously by the other two groups (13, 14). However, univariate analyses revealed that the estimated indices barely
1624
JCE & M • 2001 Vol. 86 • No. 4
CHIU ET AL.
TABLE 5. Multivariate analysis of  cell function Dependent variable
Covariate entered
Covariate removed
1stIR
r2
P
0.1663 Ethnicity BMI
0.0079 0.0141 0.1878 0.2460 0.5706 0.7356 0.7724
Diastolic blood pressure Gender Waist-hip ratio Age Systolic blood pressure 2ndIR
0.3623 Age Ethnicity Gender Waist-hip ratio BMI Diastolic blood pressure
0.0008 0.0011 0.0311 0.0500 0.0917 0.0944 0.5129
Systolic blood pressure 1stPHS
0.2883 Age Gender Waist-hip ratio Diastolic blood pressure BMI
0.0001 0.0032 0.0355 0.0389 0.0404 0.3791 0.8730
Ethnicity Systolic blood pressure 2ndPHS
0.2825 Age Gender Waist-hip ratio Diastolic blood pressure BMI Ethnicity Systolic blood pressure
detected the observed differences in the measured indices (P ⫽ 0.1087 for ISIM, P ⫽ 0.02416 for ISIS, P ⫽ 0.0983 for 1stPHS, and P ⫽ 0.0967 for 2ndPHS). Multivariate analyses confirmed the independent impact of ethnicity on the measured indices (P ⫽ 0.0004 for ISI, P ⫽ 0.0079 for 1stIR, and P ⫽ 0.0011 for 2ndIR) and also noted marginal impacts of ethnicity on ISIM (P ⫽ 0.0358) and ISIS (P ⫽ 0.0652). However, ethnicity had no impact on 1stPHS (P ⫽ 0.3791) and 2ndPHS (P ⫽ 0.3100) in multivariate analyses. Although excellent correlation of insulin sensitivity measured by euglycemic and hyperglycemic clamps have been reported before (7, 18, 21), the hyperglycemic clamp is more dependent on insulin-independent uptake than the euglycemic clamp for maintaining plasma glucose concentrations at a higher level. To correct the mass effect of plasma glucose on glucose uptake, we also calculated GCl. Because the steady-state plasma glucose concentrations were about the same for all four ethnic groups, the differences among the four ethnic groups were about the same for ISI and GCl (P ⬍ 0.001; Table 3). Multivariate analyses revealed the same set of covariates for both ISI and GCl at very similar P values for each covariate (Table 4). Ethnicity is an independent determinant for ISI (P ⫽ 0.0004) and GCl (P ⫽ 0.0008). Various factors could affect insulin sensitivity and  cell function. However, it is almost impossible to match all subjects with every factor for all ethnic groups. As noted in Table 3, although differences in clinical features were noted in age
0.0001 0.0046 0.0278 0.0383 0.0443 0.3100 0.8211
and waist-hip ratio among the ethnic groups, minor differences were also noted in diastolic blood pressure (P ⫽ 0.10) and BMI (P ⫽ 0.14). Multivariate analysis confirmed that ethnicity was an independent determinant for ISI (P ⫽ 0.0004; Table 4), 1stIR, and 2ndIR (P ⫽ 0.0079 and P ⫽ 0.0011, respectively; Table 5). The fundamental issue was whether ethnicity had an independent influence on both insulin sensitivity and  cell function or whether ethnicity had a primary effect on one and a secondary effect on the others. In glucose-tolerant subjects, it is essential to maintain plasma glucose levels in a relatively narrow physiological range and  cells have to respond to prevailing insulin resistance (sensitivity). If  cells fail to compensate for insulin resistance, abnormal glucose tolerance (either impaired glucose tolerance or diabetes) will develop. In contrast, if  cells respond too strongly (overresponse) to insulin resistance, hypoglycemia will develop. Therefore, in glucose-tolerant subjects, there is a dynamic balance between insulin sensitivity and  cell function. Any interruption of the balance will lead to a change in glucose homeostasis and a pathological state will develop. The subject will no longer be glucose tolerant. Therefore, it is impossible, at least in the present time, to disentangle the independent effects of ethnicity on insulin sensitivity and  cell function in glucose-tolerant subjects. We attempted to resolve this issue by using multivariate analysis. First, we assumed that the primary influence of ethnicity was on ISI and
INSULIN SENSITIVITY AND  CELL FUNCTION
we considered ISI as a covariate for 1stIR in the multivariate analysis. We found that ISI was the only determinant for 1stIR (r2 ⫽ 0.3289, P ⬍ 0.000001), accounting for 0.5903 of the variation of 1stIR, whereas ethnicity had no impact on 1stIR (P ⫽ 0.5213). Conversely, we also considered 1stIR as a covariate for ISI. We found that ethnicity (P ⫽ 0.0200) was an independent determinant for ISI and ethnicity, along with 1stIR (P ⬍ 0.000001), diastolic blood pressure (P ⫽ 0.0002), gender (P ⫽ 0.0003), age (P ⫽ 0.0008), and waist-hip ratio (P ⫽ 0.0010) accounted for 54% of the variation of ISI. Because 2ndIR was a denominator for ISI, they were very tightly related to each other statistically. Ethnicity (P ⫽ 0.5342) was removed from the multivariate analysis when ISI was considered as a covariate for 2ndIR. Similarly, when 2ndIR was considered as a covariate for ISI, ethnicity (P ⫽ 0.2867) was excluded from the analysis as expected. Therefore, it was impossible to disentangle the relationship between ISI and 2ndIR using a multivariate analytical approach. Nonetheless, the multivariate analyses of ISI and 1stIR suggested that the impact of ethnicity was primarily on insulin sensitivity, and we observed a compensatory change in  cell function. This is consistent with the notion that insulin levels are increased in insulin-resistant subjects (22), which results from both increased secretion and reduced clearance of insulin (23). This compensatory hypersecretion of insulin not only reflects the expansion of  cell mass (24, 25), but also altered expression of key enzymes of glucose metabolism in  cells as observed in Zucker fatty rats (26). Among the four ethnic groups, Asian-Americans were the most insulin resistant with the lowest adjusted ISI (3.7334 m/ m2/min/pm), followed by Mexican-Americans (4.7025 m/ m2/min/pm) and African-Americans (4.9027 m/m2/min/ pm). Caucasian-Americans were the most insulin sensitive (6.7810 m/m2/min/pm). A reciprocal change was noted in their adjusted  cell function (2,035 pm in Asian-Americans, 1,877 pm in Mexican-Americans, 1,707 pm in African-Americans, and 1,337 pm in Caucasian-Americans for the adjusted 1stIR; 582 pm in Asian-Americans, 487 pm in both Mexican- and African-Americans, and 353 pm in Caucasian-Americans for the adjusted 2ndIR). However, ethnicity had very little impact on the estimated indices. It remains to be seen whether the enhanced statistical power obtained by using the estimated indices that can be more readily applied to relatively large samples outweighs the loss of information resulting from their lessthan-perfect correlations with more definitive measures of insulin sensitivity and  cell function. Because the estimated indices reflected the variation of the measured indices poorly, it is essential to measure insulin sensitivity and  cell function in a thorough manner if one plans to compare the impact of various factors, such as ethnicity, on these indices. Acknowledgments We thank Mohammad F. Saad, M.D., for the measurement of plasma insulin concentration and the staff of the General Clinical Research Center at the University of California, Los Angeles for continued support. We also thank Audrey Chu, George P. Tsai, Jennifer M. Ryu,
1625
Jennifer L. McGullam, and Jennifer E. McCarthy for laboratory assistance.
References 1. Kahn BB. 1998 Type 2 diabetes: when insulin secretion fails to compensate for insulin resistance. Cell. 92:593–596. 2. Gerich JE. 1998 The genetic basis of type 2 diabetes mellitus: impaired insulin secretion versus impaired insulin sensitivity. Endocr Rev. 19:491–503. 3. Gerich JE. 2000 Insulin resistance is not necessarily an essential component of type 2 diabetes. J Clin Endocrinol Metab. 85:2113–2115. 4. Polonsky KS, Given BD, Hirsch LJ, et al. 1988 Abnormal patterns of insulin secretion in non-insulin-dependent diabetes mellitus. N Engl J Med. 318:1231–1239. 5. DeFronzo RA. 1992 Pathogenesis of type 2 (non-insulin dependent) diabetes mellitus: a balanced overview. Diabetologia. 35:389 –397. 6. Gulli G, Ferrannini E, Stern M, Haffner S, DeFronzo RA. 1992 The metabolic profile of NIDDM is fully established in glucose-tolerant offspring of two Mexican-American NIDDM parents. Diabetes. 41:1575–1586. 7. Pimenta W, Korytkowski M, Mitrakou A, et al. 1995 Pancreatic -cell dysfunction as the primary genetic lesion in NIDDM. Evidence from studies in normal glucose-tolerant individuals with a first-degree NIDDM relative. JAMA. 273:1855–1861. 8. Watanabe RM, Valle T, Hauser ER, et al. 1999 Familiarity of quantitative metabolic traits in Finnish families with non-insulin-dependent diabetes mellitus. Finland-United States investigation of NIDDM genetics (FUSION) study investigators. Hum Hered. 49:159 –168. 9. Mitchell BD, Kammerer CM, Hixson JE, et al. 1995 Evidence for a major gene affecting postchallenge insulin levels in Mexican-Americans. Diabetes. 44:284 –289. 10. Elbein SC, Hasstedt SJ, Wegner K, Kahn SE. 1999 Heritability of pancreatic -cell function among nondiabetic members of Caucasian familial type 2 diabetic kindreds. J Clin Endocrinol Metab. 84:1398 –1403. 11. Schumacher MC, Hasstedt SJ, Hunt SC, Williams RR, Elbein SC. 1992 Major gene effect for insulin levels in familial NIDDM pedigrees. Diabetes. 41:416 – 423. 12. National Diabetes Data Group. 1979 Classification and diagnosis of diabetes mellitus and other categories of glucose intolerance. Diabetes. 28:1039 –1057. 13. Matsuda M, DeFronzo RA. 1999 Insulin sensitivity indices obtained from oral glucose tolerance testing: comparison with the euglycemic insulin clamp. Diabetes Care. 22:1462–1470. 14. Stumvoll M, Mitrakou A, Pimenta W, et al. 2000 Use of the oral glucose tolerance test to assess insulin release and insulin sensitivity. Diabetes Care. V23:295–301. 15. Raffel LJ, Robbins DC, Norris JM, et al. 1996 The GENNID study. A resource for mapping the genes that cause NIDDM. Diabetes Care. 19:864 – 872. 16. Rossetti L, Giaccari A, DeFronzo RA. 1990 Glucose toxicity. Diabetes Care. 13:610 – 630. 17. Ferrannini E, Buzzigoli G, Bonadonna R, et al. 1987 Insulin resistance in essential hypertension. N Engl J Med. 317:350 –357. 18. DeFronzo RA, Tobin JD, Andres R. 1979 Glucose clamp technique: a method for quantifying insulin secretion and resistance. Am J Physiol. 237:E214 –E223. 19. Chiu KC, McCarthy JE. 1997 The insertion allele at the angiotensin I-converting enzyme gene locus is associated with insulin resistance. Metabolism. 46:395–399. 20. Clausen JO, Borch-Johnsen K, Ibsen H, et al. 1996 Insulin sensitivity index, acute insulin response, and glucose effectiveness in a population-based sample of 380 young healthy Caucasians. Analysis of the impact of gender, body fat, physical fitness, and lifestyle factors. J Clin Invest. 98:1195–1209. 21. Mitrakou A, Vuorinen-Markkola H, Raptis G, et al. 1992 Simultaneous assessment of insulin secretion and insulin sensitivity using a hyperglycemia clamp. J Clin Endocrinol Metab. 75:379 –382. 22. Jones CN, Abbasi F, Carantoni M, Polonsky KS, Reaven GM. 2000 Roles of insulin resistance and obesity in regulation of plasma insulin concentrations. Am J Physiol Endocrinol Metab. 278:E501–E508. 23. Polonsky KS, Given BD, Hirsch L, et al. 1988 Quantitative study of insulin secretion and clearance in normal and obese subjects. J Clin Invest. 81:435– 441. 24. Pick A, Clark J, Kubstrup C, et al. 1998 Role of apoptosis in failure of -cell mass compensation for insulin resistance and -cell defects in the male Zucker diabetic fatty rat. Diabetes. 47:358 –364. 25. Zhou YP, Cockburn BN, Pugh W, Polonsky KS. 1999 Basal insulin hypersecretion in insulin-resistant Zucker diabetic and Zucker fatty rats: role of enhanced fuel metabolism. Metabolism. 48:857– 864. 26. Tokuyama Y, Sturis J, DePaoli AM, et al. 1995 Evolution of -cell dysfunction in the male Zucker diabetic fatty rat. Diabetes. 44:1447–1457.
