Diabetes Care Volume 37, July 2014 1959

Distribution of C-Peptide and Its Ping Xu,1 Xiaoning Qian,1 Desmond A. Schatz,2 Determinants in North American David Cuthbertson,1Jeffrey P. Krischer,1 and the DPT-1 Study Group Children at Risk for Type 1

Diabetes Care 2014;37:1959–1965 | DOI: 10.2337/dc13-2603

OBJECTIVE To determine basal and stimulated C-peptide percentiles in North American chil- dren and adolescents at risk for (T1D) and to examine factors associated with this distribution in the Diabetes Prevention Trial–Type 1 (DPT-1).

RESEARCH DESIGN AND METHODS We included 582 subjects aged 4–18 years at randomization in the DPT-1 trials. A 2-h oral (OGTT) was performed at baseline and every 6 months during the 5-year follow-up period. The percentile values of C-peptide after baseline OGTT were estimated according to age, BMI Z score (BMIZ), and/or sex categories. Conditional quantile regression was used to examine the relation- PATHOPHYSIOLOGY/COMPLICATIONS ship between C-peptide percentiles and various independent variables.

RESULTS The basal and stimulated C-peptide levels increased significantly as age and BMIZ increased (P < 0.05).BothageandBMIZhadastrongerimpactontheupper quartile of C-peptide distributions than the lower quartile. Sex was only signifi- cantly associated with stimulated C-peptide. Higher stimulated C-peptide levels were generally observed in girls compared with boys at the same age and BMIZ (P < 0.05). HLA type and number of positive antibodies and antibody titers (islet cell antibody [ICA], autoantibody, GAD65A, and ICA512A) were not significantly associated with C-peptide distribution after adjustment for age, BMIZ, and sex.

CONCLUSIONS Age-, sex-, and BMIZ-specific C-peptide percentiles can be estimated for North 1 American children and adolescents at risk for T1D. They can be used as an assess- Pediatrics Epidemiology Center, College of Med- icine, University of South Florida, Tampa, FL ment tool that could impact the recommendations in T1D prevention trials. 2Department of Pediatrics, College of Medicine, University of Florida, Gainesville, FL Type 1 diabetes (T1D) is a metabolic disease characterized by elevated blood glucose Corresponding author: Ping Xu, [email protected]. levels due to insufficient insulin production (1–3). It results from an autoimmune edu. process that leads to destruction of pancreatic b-cells. C-peptide and insulin are Received 7 November 2013 and accepted 21 February 2014. released simultaneously from the pancreas with the cleavage of proinsulin. Mea- surement of C-peptide production under a standardized condition provides a sen- This article contains Supplementary Data online b – at http://care.diabetesjournals.org/lookup/ sitive assessment of -cell function (4 8). suppl/doi:10.2337/dc13-2603/-/DC1. b C-peptide levels have been used as a surrogate outcome for preserved -cell © 2014 by the American Diabetes Association. function in intervention trials conducted in new-onset patients. Even earlier assess- See http://creativecommons.org/licenses/by- ment (i.e., in at-risk subjects) of C-peptide and its precipitating or determinant nc-nd/3.0/ for details. 1960 C-Peptide Percentile in Children at T1D Risk Diabetes Care Volume 37, July 2014

factors is critical. Most previous studies their risk of T1D based on genetic, im- A nonparametric approach, condi- have reported only means and SDs as- munologic, and metabolic characteris- tional quantile regression, was used to suming that measurement of C-peptide tics. A total of 711 individuals were examine the relationship between base- production follows a normal distribu- randomized into either a parenteral trial line C-peptide percentiles and various tion (9,10). However, C-peptide profiles or an oral insulin trial according to their independent variables. By extending may follow a nonnormal distribution, risk profiles. These randomized subjects the exclusive focus of the estimation of which could substantially impact inter- were followed until T1D onset or up to conditional mean functions in tradi- pretation. When traditional statistical 5 years. Previous analyses showed that tional regression models, the quantile linear regression techniques are de- the subjects failed to reach the primary regression approach allows examining ployed such as ordinary least square end point of preventing diabetes. We of the entire distribution of the variable and general linear model, a departure analyzed 582 subjects aged 4–18 years of interest (C-peptide here) rather from normality can result in inaccurate at randomization. Among them, 224 than a single measure of the central ten- estimates of C-peptide. Accurate risk (38.5%) subjects were 4–8yearsold dency of its distribution. In addition, the characterization is critical in the design and 358 subjects (61.5%) were 9–18 quantile regression approach has its ad- of prevention trials. years old. There were 350 boys (60.1%) vantages over traditional regression A number of environmental and ge- and 232 (39.9%) girls. models for its flexibility to allow the co- netic factors are known to be associated variates to have different impacts at dif- with T1D. Environmental factors, such Laboratory Measures ferent percentiles of the distribution as as exposure to enteroviral infections All assays including antibody assays well as its robustness with respect to and cow’s milk, have been identified as were performed as previously described departures from normality and skewed potential triggers of T1D in epidemiolog- (16). For HLA-DQ typing, DNA was ex- tails because it does not put any distri- ical and immunological studies (11–13). tracted from the buffy coats of periph- butional assumption beforehand (18– Previous studies also established that eral blood leukocytes, and HLA-DQA1 20). The effect of potential covariates, factors, such as age, sex, BMI, relation- and DQB1alleles were amplified by PCR including age categories ($4and#8 ship to proband, HLA type, and antibody with the use of sequence-specificprobes. vs. $9and#18 years), sex (female vs. titer levels, were related to progression A high-risk HLA genotype was defined as male), BMI Z score (BMIZ) percentile of clinical T1D onset in children and ado- having one of the following haplotype categories (,85.0 vs. $85.0), relation- lescents at risk (14,15). These factors combinations: DQA1*0301-DQB1*0302, ship to proband (offspring, sibling, sec- contribute to the risk for T1D indepen- DQA1*0501-DQB1*0201, DQA1*04- ond degree), HLA type (high vs. low risk), dently or interactively at different pre- DQB1*0201, or DQA1*0301-DQB*0201. number of positive antibodies, and an- diabetes stages. The distribution of No HLA DQB*0602s were included. tibody titer levels, were examined uni- C-peptide may or may not depend on At baseline and every 6 months dur- variately first. Prepubertal subjects one or more of these factors that are ing the follow-up period, an oral glucose were defined as age ,9 years (21–23). linked to clinical disease onset. Taking tolerance test was done after an over- The significant covariates in the univar- into account the above observations night fast and blood samples were iate model were then selected as the and considering the importance of early drawn at 210 and 0 min. An oral glucose predictors in a multivariate model to detection of low C-peptide or preserva- load was then administered (1.75 g/kg, account for the possible variations tion of C-peptide in subjects at risk for maximum 75 g). Blood samples were in determining the distribution of T1D (as the best strategy for the preven- drawn at 30, 60, 90, and 120 min after C-peptide. Lastly, the estimated per- tion of T1D), we aimed in this study to glucose consumption. C-peptide levels centile values of C-peptide were recon- determine percentiles for basal and were measured by radioimmunoassay structed and refined by the covariates stimulated C-peptide in children and in the DPT-1 b-cell function core labora- that were both univariately and multi- adolescents at risk for T1D and to exam- tory (Seattle, WA). The basal C-peptide variately related to C-peptide distribu- ine factors associated with the distribu- level in the current analysis was calcu- tion profile. tion of C-peptide using the data from lated as the mean of C-peptide levels at All tests of significanceweretwo one of the largest T1D prevention trials. 210 and 0 min. The stimulated C-peptide tailed. P # 0.05 was considered statisti- was analyzed as the peak C-peptide level cally significant. Statistical analyses RESEARCH DESIGN AND METHODS during a 2-h OGTT test. C-peptide was were performed with SAS (version 9.2; Subjects measured in nanograms per milliliter SAS Institute, Cary, NC). The Diabetes Prevention Trial–Type 1 (1 ng/mL is equal to 0.333 nmol/L). (DPT-1) was a multicenter randomized, RESULTS controlled clinical study in North Amer- Statistical Methods Table 1 demonstrates the subjects’ de- ica designed to determine whether it is The descriptive statistics of both ob- mographics and clinical and laboratory possible to delay or prevent the clinical served basal and stimulated C-peptide characteristics. The majority of subjects onset of T1D in individuals with autoim- levels at baseline, such as their mean, were white (92.71%), had high-risk HLA munity. More than 100,000 nondiabetic minimum, median, maximum values, types (84.54%), and had three or four relativesofsubjectswithT1Dwere SDs, skewness, and kurtosis, were re- positive antibodies (84.88%) at baseline. screened to detect the presence of islet ported. A Kolmogorov-Smirnov good- Approximately 23% of subjects in this cell antibodies (ICAs). Individuals found ness-of-fit test was used to test for the population were overweight or obese to have ICAs were staged to determine normality of its distribution (17). basedontheirBMIZ($85.0 percentile). care.diabetesjournals.org Xu and Associates 1961

fi Table 1—Subjects’ demographics and clinical characteristics (N = 582) were signi cantly associated with both Age, years, mean (SD) 9.98 (3.82) basal and stimulated C-peptide distribu- 4–8, years, n (%) 224 (38.49) tion (P , 0.001), with no evidence of 9–18, years, n (%) 358 (61.51) significant effect modification by either Sex, n (%) age or sex. As indicated in Fig. 1, both Female 232 (39.86) age and BMIZ had a stronger impact on Male 350 (60.14) the upper quartile of C-peptide distribu- Race, n (%) tion than the lower quartile. HLA type White 529 (92.71) and IAA titer were no longer associated Nonwhite 41 (5.23) Unknown 12 (2.06) with the percentiles of basal or stimu- lated C-peptide after adjustment for Relationship to proband, n (%) Sibling 367 (63.06) age, sex, and BMIZ percentile. Further Offspring 156 (26.80) analysis demonstrated that age was sig- Parent 1 (0.17) nificantly inversely related to IAA titer Non–first degree relative 58 (9.97) (data not shown). IAA titers were higher BMIZ percentile and age, mean (SD) 57.80 (29.38) in the younger age-group (P , 0.001) ,85.0, n (%) 416 (71.48) compared with the older group, which $ 85.0, n (%) 132 (22.68) indicates that IAA might be a confound- Unknown 34 (5.84) ing factor of C-peptide distribution HLA risk, n (%) rather than a determinant factor. High risk 492 (84.54) Low risk 88 (15.12) Supplementary Table 1 shows the es- Unknown 2 (0.34) timated percentile values of basal Number of positive antibodies, n (%) C-peptide in children and adolescents 133(5.67)with T1D ICA autoimmunity according 255(9.45)to their age categories and BMIZ cate- 3 205 (35.22) gories, respectively. Supplementary Ta- 4 289 (49.66) bles 2 and 3 show the percentile values ICA titer, median (Q1–Q3) 160.00 (40.00–320.00) for boys and girls of all age and BMIZ IAA titer, median (Q1–Q3) 185.90 (69.20–450.30) groups combined. Percentile values GAD65A titer, median (Q1–Q3) 0.260 (0.049–0.744) were calculated for each group, even ICA512A titer, median (Q1–Q3) 0.105 (0.013–0.714) when the statistical tests indicated lack Q1, the first quantile; Q3, the third quantile. of differences between certain groups in lower percentiles. For all percentiles, stimulated C-peptide in lower-BMIZ girls was significantly different (P , 0.03) The mean basal and stimulated proband, number of positive antibodies, from that in high-BMIZ girls at both C-peptide levels were 1.0 and 4.9 ICA titer, GAD65A titer, and ICA512A titer age periods. For boys, the same pattern ng/mL, respectively. The median values were not associated with the C-peptide was observed. were 0.6 ng/mL for basal and 3.5 ng/mL percentiles. Age, BMIZ percentile, HLA for stimulated C-peptide. The median type, and insulin autoantibody (IAA) an- CONCLUSIONS values were less than the (arithmetic) tibody titer were significantly related to Previous findings from the DPT-1 study mean values for both basal and stimu- the distribution of C-peptide, and these have shown that C-peptide is a good bio- lated C-peptides, which indicated that covariates were included in the multivar- marker in predicting T1D onset in chil- the distribution of C-peptide was right iate models to account for all possible dren at risk, with a level of prediction skewed. The kurtosis was higher for the variations of C-peptide distribution. Sex ability similar to that of glucose level basal C-peptide (4.6) than for the stim- was not associated with basal C-peptide (24,25). Longitudinal studies have ulated C-peptide (1.6). Higher value of but was significantly associated with shown that individuals at risk have a kurtosis indicates a higher and sharper stimulated C-peptide in univariate anal- prolonged and gradual loss of C-peptide peak. These observations suggest that yses. Therefore, sex was included in the with the persistence of substantial the C-peptide distributions do not con- multivariate model for the stimulated b-cell function until at least 6 months form to a normal distribution, and a for- C-peptide only. before the onset of clinical disease mal Kolmogorov-Smirnov goodness-of- Table 3 reveals the results from the (9,26,27). We believe that this is the first fit test for normality based on skewness multivariate model. An age-related sig- study to examine factors associated and kurtosis rejected the hypothesis nificant increase in C-peptide distribu- with C-peptide production in children that the basal C-peptide or stimulated tion was detected (P , 0.001). Sex was and adolescents at risk for T1D. We C-peptide is normally distributed (P , associated with stimulated C-peptide. found that ICA, IAA, GAD65A, and 0.01 for both basal and stimulated Higher stimulated C-peptide percentiles ICA512A antibody titers were not signif- C-peptide). were generally observed in girls com- icantly associated with C-peptide distri- Table 2 presents univariate quantile pared with boys at the same age and bution after adjustment for age, BMIZ, regression analyses. Relationship to BMIZ scores (P , 0.05). BMIZ scores and sex in the samples from DPT-1. This 1962 C-Peptide Percentile in Children at T1D Risk Diabetes Care Volume 37, July 2014

Table 2—Univariate quantile regression model for predicting C-peptide percentiles Basal C-peptide Stimulated C-peptide 25th percentile 50th percentile 75th percentile 25th percentile 50th percentile 75th percentile Sibling 0.25 0.11 0.54 0.18 1.00 0.46 Offspring 1.00 0.50 0.55 0.60 0.61 0.44 Second degree 0.08 0.40 0.12 1.00 0.15 0.84 Number of positive antibodies 1 vs.4 0.06 0.10 0.11 1.00 0.27 0.85 2 vs. 4 0.49 1.00 0.57 0.32 0.51 0.19 3 vs. 4 0.33 0.09 1.00 1.00 0.65 0.30 IAA titer 0.13 0.05 0.29 0.16 ,0.001 0.02 ICA titer 0.62 1.00 0.84 0.25 0.66 0.21 GAD65A titer 1.00 1.00 0.61 0.63 1.00 0.54 ICA512 titer 0.47 0.53 0.60 1.00 0.63 0.71 HLA (high vs. low risk) 0.05 0.05 0.09 0.54 0.02 0.31 Sex (female vs. male) 0.23 0.10 0.54 0.11 ,0.001 0.01 BMIZ percentile (,85.0 vs. $85.0) 0.01 ,0.001 ,0.001 ,0.001 ,0.001 ,0.001 Age (4–8 vs. 9–18 years) ,0.001 ,0.001 ,0.001 ,0.001 ,0.001 ,0.001 Data are P values.

is consistent with previous studies in years old at baseline were progressing T1D among boys may be slower com- children with newly diagnosed T1D through puberty, at which time there paredwithgirls.Overallb-cell function (28–30). The relationships to proband is a known increase in insulin produc- may need to be further reduced in boys or HLA type were also not found to be tion, and a total of 16% of subjects than in girls to progress to T1D (32,33). significant predictors of C-peptide. who had lower BMIZ percentile at base- In addition, girls may have less insulin These data clearly demonstrate that line had increased BMIZ at the time of sensitivity owing to higher BMIZ than the distributions of basal and stimulated diagnosis. Thus, the decrease in C-peptide boys at the same age. C-peptide in a sample of children and production at diagnosis may be consid- The estimated values in Supplemen- adolescents with autoimmunity depend ered not just a loss in absolute terms but tary Tables 4–6 provide a powerful tool on BMIZ and age. Subjects with higher also a failure to increase with age or for the interpretation of C-peptide in BMIZ have higher C-peptide at the con- BMIZ (31). A future longitudinal analysis children at different BMIZ categories sidered percentiles than the subjects to study the percentile changes before and age-groups. Based on these values, with lower BMIZ. Likewise, older chil- disease onset is warranted. careful attention to children with dren have higher C-peptide than chil- Our results also demonstrate that C-peptide values that fall on the 10th dren at a younger age. Interestingly, boys have lower stimulated C-peptide and 25th percentiles, according to their .30% (N = 68) of subjects who pro- compared with girls in all age-groups. BMIZ classification and age, becomes gressed to overt T1D at the end of study However, basal C-peptide was not sex important in the identification of sub- did not have an absolute decrease in dependent. This may suggest that boys groups of children progressing to clinical stimulated C-peptide at the time of di- with autoimmunity are less likely to T1D. For example, at the 25th percentile agnosis from baseline. Approximately progress to overt disease than compara- of the basal distribution, the children 75% of these subjects who were ,9 ble girls and that the pathogenesis of between 9 and 18 years of age exceed

Table 3—Multivariate quantile regression model for predicting C-peptide percentiles (coefficient and its 95% CI) 25th percentile 50th percentile 75th percentile Estimate 95% CI Estimate 95% CI Estimate 95% CI Basal C-peptide HLA (low vs. high risk) 0.00 20.15 0.15 0.10 20.02 0.22 0.10 20.13 0.33 IAA titer 0.00 0.00 0.00 0.00 0.00 0.00 0.00 0.00 0.00 BMIZ percentile (,85.0 vs. $85.0) 20.24 20.33 20.16 20.31 20.43 20.19 20.55 20.72 20.38 Age (4–8 vs. 9–18 years) 20.34 20.43 20.25 20.40 20.48 20.33 20.60 20.72 20.48 Stimulated C-peptide Sex (female vs. male) 0.58 0.27 0.89 0.50 0.10 0.90 0.79 0.37 1.20 HLA (low vs. high risk) 0.26 20.09 0.60 0.26 20.29 0.81 0.27 20.66 1.20 IAA titer 0.00 0.00 0.00 0.00 0.00 0.00 0.00 0.00 0.00 BMIZ percentile (,85.0 vs. $85.0) 21.00 21.32 20.68 21.05 21.54 20.56 21.85 22.32 21.39 Age (4–8 vs. 9–18 years) 21.28 21.52 21.04 21.50 21.89 21.10 21.90 22.29 21.52 care.diabetesjournals.org Xu and Associates 1963

Figure 1—Effect of age, BMIZ, and/or sex on C-peptide by quantile. A: Effect on basal C-peptide (age 4–8 vs. 9–18 years, BMIZ ,85.0 vs. $85.0 percentile) (estimated parameter by quantile for baseline fasting C-peptide with 95% CI). B: Effect on stimulated C-peptide (sex female vs. male; age 4–8 vs. 9–18 years; BMIZ ,85.0 vs. $85.0 percentile) (estimated parameter by quantile for baseline peak C-peptide with 95% CI). the basal C-peptide values of 1.00 ng/mL, Standard linear regression models, models are sensitive to outliers and can identified as the cutoff point for increased such as ordinary least square, are exten- lead to unrealistic models if outliers are risk of T1D among higher-BMIZ children. sively used in statistical analyses. Despite present in the data set. This is especially a Similarly, at ages 9–18 years, the children their popularity, these conditional mean problem if the sample size is moderately with lower BMIZ achieved the 0.75 ng/mL models have several limitations. When small and the error distribution is heavy cutoff point in the 25th percentile of the interest is in the percentiles of the con- tailed (17,18). Quantile regression over- basal C-peptide distribution and 0.50 ditional distribution rather than the comes these limitations of standard linear ng/mL cutoff point in the 10th percentile. mean, standard regression models may regression. It is robust in handling ex- Therefore, the cutoff point for C-peptide fail to provide the desired information treme value points and outliers for the to classify subjects as having a loss of because the assumption of normally dis- outcome of interest. More importantly, b-cell function may be different depend- tributed residuals with constant variance it provides a more complete understand- ing on their age, BMIZ, and/or sex. may not be justified. Standard regression ing of the impact of covariates on the 1964 C-Peptide Percentile in Children at T1D Risk Diabetes Care Volume 37, July 2014

dependent variable across the whole dis- Duality of Interest. No potential conflicts of 13. van der Werf N, Kroese FGM, Rozing J, tribution, especially toward the lower or interest relevant to this article were reported. Hillebrands JL. Viral infections as potential trig- P.X. performed the upper tail of the distribution. The ob- Author Contributions. gers of type 1 diabetes. Diabetes Metab Res Rev statistical analysis, wrote the manuscript, and 2007;23:169–183 served data in this sample show that the read and approved the final manuscript. X.Q. 14. Diabetes Prevention Trial–Type 1 Diabetes distribution of C-peptide is skewed to the researched data, helped draft the manuscript, Study Group. Effects of insulin in relatives of right. The statistical inference based on and read and approved the final manuscript. patients with type 1 diabetes mellitus. N Engl J the standard linear regression is inaccu- D.A.S., D.C., and J.P.K. contributed to discussion, Med 2002;346:1685–1691 reviewed and edited the manuscript, and read rate when the distributions are skewed 15. Sosenko JM, Krischer JP, Palmer JP, et al.; and approved the final manuscript. P.X. and J.P.K. Diabetes Prevention Trial-Type 1 Study Group. and/or when the quantity of interest is are the guarantors of this work and, as such, had A risk score for type 1 diabetes derived from the upper or lower tail of the distribu- full access to all the data in the study and take autoantibody-positive participants in the diabe- tions. This further promotes the quantile responsibility for the integrity of the data and the tes prevention trial-type 1. Diabetes Care 2008; regression method for estimating percen- accuracy of the data analysis. 31:528–533 tiles of C-peptide and examining its rela- 16. Smirnov NV. Tables for estimating the References goodness of fit of empirical distributions. Ann tionship with the covariates. 1. Rewers M, Norris J, Dabelea D. Epidemiology Math Stat 1948;19:279 The results of this study are based of diabetes. In Immunology of Type 1 Diabetes. 17. Koenker R, Hallock K. Quantile regression: an on a representative sample of children 2nd ed. Eisenbarth GS, Ed. New York, Springer, introduction. J Econ Perspect 2001;15:143–156 and adolescents with pre-T1D autoim- 2004, p. 221–233 18.HaoL,NaimanDQ.Quantile Regression. munity in North America. Our findings 2. Palmer JP, Asplin CM, Clemons P, et al. In- Thousand Oaks, CA, Sage Publications, 2007 19. SAS Institute Inc. The QUANTREG Docu- thus demonstrate the importance of sulin antibodies in insulin-dependent diabetics before insulin treatment. Science 1983;222: mentation. In SAS/STAT 9.2 User’sGuide. Cary, age, BMIZ, and/or sex groups in C-peptide 1337–1339 NC, SAS Institute Inc., 2008 production in this population. Age-, sex-, 3. Eisenbarth GS, Moriyama H, Robles DT, et al. 20. Krarup T, Regeur L, Faber OK, Binder C. In- and BMIZ-specific C-peptide percentiles Insulin autoimmunity: prediction/precipitation/ sulin secretory reserve in insulin dependent pa- tients at time of diagnosis and the first 180 days are estimated for North American chil- prevention type 1A diabetes. Autoimmun Rev 2002;1:139–145 of insulin treatment. Acta Endocrinol (Copenh) dren and adolescents at risk for T1D 4. Polonsky KS, Rubenstein AH. C-peptide as a 1980;95:359–363 and can be used as an assessment tool measure of the secretion and hepatic extraction 21. Lee PA, Guo SS, Kulin HE. Age of puberty: that could impact the recommendations of insulin. Pitfalls and limitations. Diabetes data from the United States of America. APMIS in T1D prevention trials. Ideally, the use 1984;33:486–494 2001;109:81–88 fi 5. Shapiro ET, Tillil H, Rubenstein AH, Polonsky 22. Herman-Giddens ME, Wang L, Koch G. Sec- of age-, sex- and BMIZ-speci c reference ondary sexual characteristics in boys: estimates fi KS. Peripheral insulin parallels changes in insulin percentiles should use values speci c secretion more closely than C-peptide after bo- from the national health and nutrition examina- to a normal population. Therefore, the lus intravenous glucose administration. J Clin tion survey III, 1988-1994. Arch Pediatr Adolesc estimated values at different percentiles Endocrinol Metab 1988;67:1094–1099 Med 2001;155:1022–1028 describe an at-risk population and do not 6. Eaton RP, Allen RC, Schade DS, Erickson KM, 23. Anderson SE, Dallal GE, Must A. Relative weight and race influence average age at men- establish norms for healthy children. The Standefer J. Prehepatic insulin production in man: kinetic analysis using peripheral connect- arche: results from two nationally representa- majority of subjects in our sample (92%) ing peptide behavior. J Clin Endocrinol Metab tive surveys of US girls studied 25 years apart. were Caucasian, as is the case with T1D 1980;51:520–528 Pediatrics 2003;111:844–850 in North America. Thus, we also have lim- 7. Van Cauter E, Mestrez F, Sturis J, Polonsky 24. Xu P, Beam CA, Cuthbertson D, Sosenko JM, ited data to assess the distribution of KS. Estimation of insulin secretion rates from Skyler JS, Krischer JP; DPT-1 Study Group. Prog- nostic accuracy of immunologic and metabolic C-peptide production in minority groups. C-peptide levels. Comparison of individual and standard kinetic parameters for C-peptide clear- markers for type 1 diabetes in a high-risk pop- It remains unknown whether there is a ance. Diabetes 1992;41:368–377 ulation: receiver operating characteristic analy- difference in C-peptide production among 8. Palmer JP, Fleming GA, Greenbaum CJ, et al. sis. Diabetes Care 2012;35:1975–1980 different race or ethnicity groups. C-peptide is the appropriate outcome measure 25. Xu P, Wu Y, Zhu Y, et al.; Diabetes Preven- Our study also suggests the possibil- for type 1 diabetes clinical trials to preserve tion Trial-Type 1 (DPT-1) Study Group. Prognos- b-cell function: report of an ADA workshop, tic performance of metabolic indexes in ity of developing age-, sex-, and BMIZ- – predicting onset of type 1 diabetes. Diabetes fi fi 21-22 October 2001. Diabetes 2004;53:250 speci c cutoff values for identi cation 264 Care 2010;33:2508–2513 of subjects with increased risk in pre- 9. Sosenko JM, Palmer JP, Rafkin LE, et al.; Di- 26. Sosenko JM, Palmer JP, Greenbaum CJ, vention trials that target preserving abetes Prevention Trial-Type 1 Study Group. et al. Patterns of metabolic progression to b-cell function. It is important to estab- Trends of earlier and later responses of C-peptide type 1 diabetes in the Diabetes Prevention Trial- Type 1. Diabetes Care 2006;29:643–649 lish that such specific cutoffs require to oral glucose challenges with progression to type 1 diabetes in diabetes prevention trial-type 1 par- 27. Sosenko JM, Palmer JP, Rafkin-Mervis L, demonstration of differential predic- ticipants. Diabetes Care 2010;33:620–625 et al. Glucose and C-peptide changes in the peri- tive ability and not simply demonstra- 10. Schatz D, Cuthbertson D, Atkinson M, et al. onset period of type 1 diabetes in the Diabetes tion of marginal distributions of the Preservation of C-peptide secretion in subjects Prevention Trial-Type 1. Diabetes Care 2008;31: biomarker. at high risk of developing type 1 diabetes 2188–2192 mellitus–a new surrogate measure of non- 28. Jaeger C, Allendorfer¨ J, Hatziagelaki E, et al. progression? Pediatr Diabetes 2004;5:72–79 Persistent GAD 65 antibodies in longstanding Funding. DPT-1 was sponsored by the National 11. Verge CF, Howard NJ, Irwig L, Simpson JM, IDDM are not associated with residual beta- Institute of Diabetes and Digestive and Kidney Mackerras D, Silink M. Environmental factors cell function, neuropathy or HLA-DR status. Diseases, the National Institute of Allergy and in childhood IDDM. A population-based, case- Horm Metab Res 1997;29:510–515 Infectious Diseases, the Eunice Kennedy Shriver control study. Diabetes Care 1994;17:1381–1389 29. Wang L, Lovejoy NF, Faustman DL. Persis- National Institute of Child Health and Human 12. Akerblom HK, Vaarala O, Hyoty¨ H, Ilonen J, tence of prolonged C-peptide production in Development, the National Center for Research Knip M. Environmental factors in the etiology of type 1 diabetes as measured with an ultrasen- Resources, the American Diabetes Association, type 1 diabetes. Am J Med Genet 2002;115:18– sitive C-peptide assay. Diabetes Care 2012;35: and JDRF. 29 465–470 care.diabetesjournals.org Xu and Associates 1965

30. Nicole AS, Emily BT, Kevan CH. Natural his- diabetes in relation to autoimmunity, in not type 1 diabetes. Diabetes Care 2004;27: tory of b-cell function in type 1 diabetes. Dia- adults with newly diagnosed diabetes in Kro- 1985–1990 betes 2005;54(Suppl. 2):S32–S39 noberg, Sweden. Eur J Endocrinol 2012;166: 33. Weets I, Truyen I, Verschraegen I, et al.; 31. Thunander M, TornC,PeterssonC,¨ 1021–1029 Belgian Diabetes Registry. Sex- and season- Ossiansson B, Fornander J, Landin-Olsson M. 32. Krischer JP, Cuthbertson DD, Greenbaum C; dependent differences in C-peptide levels at Levels of C-peptide, body mass index and age, Diabetes Prevention Trial-Type 1 Study Group. diagnosis of immune-mediated type 1 diabetes. and their usefulness in classification of Male sex increases the risk of autoimmunity but Diabetologia 2006;49:1158–1162