J. Perinat. Med. 2015; 43(3): 283–289
Rinat Gabbay-Benziv, Lauren E. Doyle, Miriam Blitzer and Ahmet A. Baschat*
First trimester prediction of maternal glycemic status Abstract Objective: To predict gestational diabetes mellitus (GDM) or normoglycemic status using first trimester maternal characteristics. Methods: We used data from a prospective cohort study. First trimester maternal characteristics were compared between women with and without GDM. Association of these variables with sugar values at glucose challenge test (GCT) and subsequent GDM was tested to identify key parameters. A predictive algorithm for GDM was developed and receiver operating characteristics (ROC) statistics was used to derive the optimal risk score. We defined normoglycemic state, when GCT and all four sugar values at oral glucose tolerance test, whenever obtained, were normal. Using same statistical approach, we developed an algorithm to predict the normoglycemic state. Results: Maternal age, race, prior GDM, first trimester BMI, and systolic blood pressure (SBP) were all significantly associated with GDM. Age, BMI, and SBP were also associated with GCT values. The logistic regression analysis constructed equation and the calculated risk score yielded sensitivity, specificity, positive predictive value, and negative predictive value of 85%, 62%, 13.8%, and 98.3% for a cut-off value of 0.042, respectively (ROCAUC – area under the curve 0.819, CI – confidence interval 0.769–0.868). The model constructed for normoglycemia prediction demonstrated lower performance (ROC-AUC 0.707, CI 0.668–0.746). Conclusions: GDM prediction can be achieved during the first trimester encounter by integration of maternal
*Corresponding author: Ahmet A. Baschat, Obstetrics, Gynecology and Reproductive Science, University of Maryland Medical Center, Baltimore, MD, USA; and Department of Gynecology and Obstetrics, Johns Hopkins School of Medicine, Baltimore, MD, USA, E-mail: [email protected] Rinat Gabbay-Benziv and Lauren E. Doyle: Obstetrics, Gynecology and Reproductive Science, University of Maryland Medical Center, Baltimore, MD, USA Miriam Blitzer: Department of Pediatrics, University of Maryland Medical Center, Baltimore, MD, USA
characteristics and basic measurements while normoglycemic status prediction is less effective. Keywords: Gestational diabetes mellitus (GDM); normoglycemia state; prediction model. DOI 10.1515/jpm-2014-0149 Received April 30, 2014. Accepted July 15, 2014. Previously published online August 19, 2014.
Introduction Gestational diabetes mellitus (GDM) complicates between 5% and 8% of pregnancies with rates over 30% in highrisk settings and is associated with maternal, fetal, and neonatal adverse outcomes [7, 10, 11, 18, 22, 24]. The Hyperglycemia Adverse Pregnancy Outcomes (HAPO) Study showed that the relationship between adverse outcomes and maternal glucose levels is a continuum that already exists below the threshold considered diagnostic for GDM [8]. The maternal and neonatal risks rise further in women with overt GDM and early identification and rigorous treatment offer the opportunity of improved outcomes [4, 15, 16]. Gestational diabetes mellitus is diagnosed based on the recommendations from the American College of Obstetricians and Gynecologists for universal screening at 24–28 weeks’ of gestation with a 50-g glucose challenge test (GCT) followed by a diagnostic 100-g oral glucose tolerance test (OGTT) for screen-positive women. This approach is limited by both late third trimester diagnosis of GDM, which has the potential to allow too little time to optimize management, as well as universal screening of a large population of women who are at low risk for glycemic complications. The ability to stratify glycemic risks in the first trimester offers potential advantages of earlier diagnostic testing for high-risk women as well as avoids universal screening for women at low risk. While previous studies have shown that first trimester prediction of women at high risk for GDM is feasible [5, 6, 14, 20, 21, 25, 26, 29–31] the stratification of low-risk women requires knowledge of the blood
Brought to you by | Karolinska Institute Authenticated Download Date | 6/3/15 11:00 AM
284 Gabbay-Benziv et al., First trimester prediction of maternal glycemic status sugar values observed after a GCT. It is the aim of our study to develop a first trimester prediction that allows stratification of women at risk for GDM and those at low risk for glycemia-related risks that may not require universal screening.
Methods This analysis was based on data from a previously published prospective cohort study designed to develop a first trimester screening model for placental dysfunction [1]. The study was conducted in the Baltimore metropolitan area between 2007 and 2010 and was approved by the Institutional Review Board of all participating centers. Women presenting with a singleton intrauterine pregnancy between 11 and 14 weeks’ gestation were offered participation. After obtaining written informed consent, a standardized questionnaire ascertaining general health and prepregnancy measures was completed, and maternal physical examination, ultrasound examination, and blood sampling were performed as previously described [12]. For this subanalysis we included women who had prenatal care and subsequent GDM screening at the study centers. Maternal demographics, medical and obstetric history, prepregnancy maternal weight, and smoking status were recorded. Maternal height, weight, and blood pressure were obtained. Body mass index (BMI) was calculated as weight (kg)/height (m)2. Peripheral blood samples were obtained and transmitted on a filter paper to the determining laboratory (NTD Labs, Perkin Elmer, Melville, NY, USA) for analysis. Absolute measurements of pregnancy-associated plasma protein-A, free β-human chorionic gonadotropin, placental growth factor, and alpha-fetoprotein were converted to multiples of the median utilizing reference ranges of a population with documented normal outcome [13]. All participating study centers implement universal screening for GDM. A screening 50-g GCT was performed at 24–28 weeks’ gestation for all patients. Women with a 1-h value ≥ 135 mg/dL were scheduled for a 100-g 3-h diagnostic OGTT. Women with a 1-h value lower than 135 mg/dL had OGTT test in cases of abnormal fetal ultrasound findings (macrosomia or polyhydramnios) depending on the managing physician. A single value > 2 00 mg/dL on the GCT, or two abnormal values on the OGTT, based on Carpenter-Coustan thresholds were diagnostic for GDM [2]. Monitoring and therapy of glycemic control was at the discretion of the managing physician. Women were considered normoglycemic, at low risk for GDM, when the 1-h GCT maternal glucose level was lower than 135 mg/ dL and all 4 glucose levels at OGTT, whenever obtained, were normal. Therefore, even women with one abnormal value at OGTT were not considered normoglycemic (as we targeted for the lowest risk group). Source documentation for pregnancy outcome and delivery details was retrieved by research personnel and transmitted to the research office. Study data were collected, validated, and entered into a dedicated study database by trained personnel. Large for gestational age (LGA) was defined as birth weight > 90% percentile for gestational age according to local growth charts and macrosomia was defined as birth weight > 4000 g.
Statistical analysis Univariate analysis was performed to study the distribution of characteristics stratified by subsequent GDM or normoglycemia. We used χ2-test for categorical variables and Mann-Whitney U-test for continuous variables. Linear regression was utilized to identify among the continuous variables the primary determinants of the 1-h glucose levels following the GCT. Utilizing the significant independent variables, a prediction model for GDM was developed using multivariable logistic regression with backward stepwise elimination. Odds ratio (OR) was derived as the exponential function of the regression coefficient for each variable. Patient risk for GDM was then constructed from the coefficient of each variable using the formula: GDM probability = 1/(1+e–(t)) where e is the base of the natural logarithm (about 2.718) and t is the combination of explanatory variables, derived from the logistic regression analysis. The same approach was followed to predict normoglycemia. The predictive performance and optimal risk scoring of each model along with their respective sensitivity, specificity, positive predictive value (PPV), and negative predictive value (NPV) were determined by receiver operating characteristics (ROC) curve analysis and calculation of the area under the curve (AUC). The goodness-of-fit of the models was evaluated with the Hosmer-Lemeshow statistical test. Low Hosmer-Lemeshow P values ( 92 mg/dL; 1 h > 180 mg/dL; 2 h > 153 mg/dL) as these values were correlated with diabetes-associated adverse outcomes [9]. Women in our study were screened for GDM using the two-step approach and 100-g GTT, thus containment of the HAPO thresholds on our data was inaccurate. Be that as it may, predicting normoglycemia based on the HAPO thresholds in our population (given the limitations) yielded similar predicting performance (ROC-AUC 0.736, 95% CI 0.689–0.782), available upon request. Both our prediction models used prior GDM as one of their variables. Considering that 46.2% of our cohort were nullipara, we wanted to be sure that the algorithms constructed fit all. We, therefore, performed two additional analyses. First, including only the nulliparous or the multiparous women using same variables and coefficients in the logistic regression. This yielded AUC 0.805 (95% CI 0.722–0.889) and 0.821 (95% CI 0.764–0.879), respectively. Second, substituting the variable “prior GDM” in the regression by two variables: “multiparous with prior GDM” and “multiparous without prior GDM”. In this case, ROC curve analysis yielded AUC 0.812 (95% CI 0.762– 0.861). These additional analyses showed no significant difference in the prediction performance of the model. Our study has several limitations. Only women with an abnormal screening GCT received an OGTT. Accordingly we may have under-diagnosed GDM given the false negative of GCT. Our prepregnancy maternal characteristics were based on questionnaires and they open up the possibility of ascertainment bias. Moreover, given our cohort size we did not use a validation cohort. It would
be interesting to examine our algorithm performance in a parallel cohort with different basic characteristics compared to our population. We did not determine any additional serum analytes that may improve our predictive accuracy of normal and abnormal glycemic status and finally, as stated before, we feel that prediction of normoglycemia would be more appropriate analyzing population using the IADPSG criteria for diagnosis of GDM. In contrast, our strength derives from development of the model by identifying factors through their relationship with absolute glucose levels as well as the diagnostic cutoffs for GDM. This ameliorates some of the statistical risk of collinearity and over fitting of the logistic model. Accordingly, our model may have wider applicability because it is less likely to be influenced by disease prevalence.
Conclusion Maternal age, prior history of GDM, ethnicity, and prepregnancy BMI can identify the majority of women than develop GDM. Measurement of maternal blood pressure, height, and weight at the time of first trimester screening improves prediction to over 80%. In contrast, normoglycemia is less predictable and accordingly first trimester screening with the aforementioned factors does not obviate universal screening.
References [1] Baschat AA, Magder LS, Doyle LE, Atlas RO, Jenkins CB, Blitzer MG. Prediction of preeclampsia utilizing the first trimester screening examination. Am J Obstet Gynecol. 2014. [Epub ahead of print]. [2] Carpenter MW, Coustan DR. Criteria for screening tests for gestational diabetes. Am J Obstet Gynecol. 1982;144:768–73. [3] Catalano PM, Nizielski SE, Shao J, Preston L, Qiao L, Friedman JE. Down regulated IRS-1 and PPARc in obese women with gestational diabetes: relationship to FFA during pregnancy. Am J Physiol Endocrinol Metab. 2002;282:E522–33. [4] Crowther CA, Hiller JE, Moss JR, McPhee AJ, Jeffries WS, Robinson JS, et al. Effect of treatment of gestational diabetes mellitus on pregnancy outcomes. N Engl J Med. 2005;352:2477–86. [5] D’Anna R, Baviera G, Corrado F, Giordano D, Recupero S, Di Benedetto A. First trimester serum neutrophil gelatinaseassociated lipocalin in gestational diabetes. Diabet Med. 2009;26:1293–5. [6] Georgiou HM, Lappas M, Georgiou GM, Marita A, Bryant VJ, Hiscock R, et al. Screening for biomarkers predictive of gestational diabetes mellitus. Acta Diabetol. 2008;45:157–65. [7] Hadar E, Yogev Y. Translating the HAPO Study into new diagnostic criteria for GDM? From HAPO to IADPSG and back to O’Sullivan. Clin Obstet Gynecol. 2013;56:758–73.
Brought to you by | Karolinska Institute Authenticated Download Date | 6/3/15 11:00 AM
Gabbay-Benziv et al., First trimester prediction of maternal glycemic status 289 [8] HAPO Study Cooperative Research Group, Metzger BE, Lowe LP, Dyer AR, Trimble ER, Chaovarindr U, et al. Hyperglycemia and adverse pregnancy outcomes. N Engl J Med. 2008;358:1991–2002. [9] International Association of Diabetes and Pregnancy Study Groups. International Association of Diabetes and Pregnancy Study Groups recommendations on the diagnosis and classification of hyperglycemia in pregnancy. Diabetes Care. 2010;33:676–82. [10] Jenum AK, Morkrid K, Sletner L, Vangen S, Torper JL, Nakstad B, et al. Impact of ethnicity on gestational diabetes identified with the WHO and the modified International Association of Diabetes and Pregnancy Study Groups criteria: a populationbased cohort study. Eur J Endocrinol. 2012;166:317–24. [11] Jiwani A, Marseille E, Lohse N, Damm P, Hod M, Kahn JG. Gestational diabetes mellitus: results from a survey of country prevalence and practices. J Matern Fetal Neonatal Med. 2012;25:600–10. [12] Kasdaglis T, Aberdeen G, Turan O, Kopelman J, Atlas R, Jenkins C, et al. Placental growth factor in the first trimester: relationship with maternal factors and placental Doppler studies. Ultrasound Obstet Gynecol. 2010;35:280–5. [13] Krantz D, Hallahan T, Ravens R, He K, Cuckle H, Sherwin J, et al. First trimester Down syndrome screening with dried blood spots using a dual analyte free beta hCG and PAPP-A immunofluorometric assay. Prenat Diagn. 2011;31:869–74. [14] Lain KY, Daftary AR, Ness RB, Roberts JM. First trimester adipocytokine concentrations and risk of developing gestational diabetes later in pregnancy. Clin Endocrinol (Oxf). 2008;69:407–11. [15] Langer O, Rodriguez DA, Xenakis EM, McFarland MB, Berkus MD, Arrendondo F. Intensified versus conventional management of gestational diabetes. Am J Obstet Gynecol. 1994;170;1036–47. [16] Langer O, Yogev Y, Most O, Xenakis MJ. Gestational diabetes: the consequences of not treating. Am J Obstet Gynecol. 2005;192:989–97. [17] Miller JL, de Veciana M, Turan S, Kush M, Manogura A, Harman CR, et al. First trimester detection of fetal anomalies in pregestational diabetes using nuchal translucency, ductus venosus Doppler and maternal glycosylated hemoglobin. Am J Obstet Gynecol. 2013;218:385, e1–8. [18] Moses RG, Morris GJ, Petocz P, San Gil F, Garg D. The impact of potential new diagnostic criteria on the prevalence of gestational diabetes mellitus in Australia. Med J Aust. 2011;194:338–40. [19] Nanda S, Savvidou M, Syngelaki A, Akolekar R, Nicolaides KH. Prediction of gestational diabetes mellitus by maternal factors and biomarkers at 11 to 13 weeks. Prenat Diagn. 2011;31:135–41. [20] Naylor CD, Sermer M, Chen E, Farine D. Selective screening for gestational diabetes mellitus. Toronto Trihospital Gestational Diabetes Project Investigators. N Engl J Med. 1997;337:1591–6.
[21] Nicolaides KH. A model for a new pyramid of prenatal care based on the 11 to 13 weeks’ assessment. Prenat Diagn. 2011;31:3–6. [22] O’Sullivan EP, Avalos G, O’Reilly M, Dennedy MC, Gaffney G, Dunne F, et al. Atlantic diabetes in pregnancy (DIP): the prevalence and outcomes of gestational diabetes mellitus using new diagnostic criteria. Diabetologia. 2011;54:1670–5. [23] Phaloprakarn C, Tangjitgamol S, Manusirivithaya S. A risk score for selective screening for gestational diabetes mellitus. Eur J Obstet Gynecol Reprod Biol. 2009;145:71–5. [24] Reyes-Muñoz E, Parra A, Castillo-Mora A, Ortega-González C. Effect of the diagnostic criteria of the International Association of Diabetes and Pregnancy Study Groups on the prevalence of gestational diabetes mellitus in urban Mexican women: a cross-sectional study. Endocr Pract. 2012;18:146–51. [25] Savvidou M, Nelson SM, Makgoba M, Messow CM, Sattar N, Nicolaides K. First-trimester prediction of gestational diabetes mellitus: examining the potential of combining maternal characteristics and laboratory measures. Diabetes. 2010;59:3017–22. [26] Smirnakis KV, Plati A, Wolf M, Thadhani R, Ecker JL. Predicting gestational diabetes: choosing the optimal early serum marker. Am J Obstet Gynecol. 2007;196:410. [27] Syngelaki A, Bredaki FE, Vaikousi E, Maiz N, Nicolaides KH. Body mass index at 11-13 weeks’ gestation and pregnancy complications. Fetal Diagn Ther. 2011;30:250–65. [28] Teede HJ, Harrison CL, Teh WT, Paul E, Allan CA. Gestational diabetes: development of an early risk prediction tool to facilitate opportunities for prevention. Aust NZ J Obstet Gynaecol. 2011;51:499–504. [29] Tul N, Pusenjak S, Osredkar J, Spencer K, Novak-Antolic Z. Predicting complications of pregnancy with first-trimester maternal serum free-betahCG, PAPP-A and inhibin-A. Prenat Diagn. 2003;23:990–6. [30] Van Leeuwen M, Opmeer BC, Zweers EJ, van Ballegooie E, ter Brugge HG, de Valk HW, et al. Estimating the risk of gestational diabetes mellitus: a clinical prediction model based on patient characteristics and medical history. Br J Obstet Gynaecol. 2010;117:69–75. [31] Wolf M, Sandler L, Hsu K, Vossen-Smirnakis K, Ecker JL, Thadhani R. First-trimester C-reactive protein and subsequent gestational diabetes. Diabetes Care. 2003;26:819–24. [32] Zhang S, Folsom AR, Flack JM, Liu K. Body fat distribution before pregnancy and gestational diabetes: findings from Coronary Artery Risk Development in Young Adults (CARDIA) Study. Br Med J. 1995;311:1139–40. The authors stated that there are no conflicts of interest regarding the publication of this article. Supplemental Material: The online version of this article (DOI: 10.1515/jpm-2014-0149) offers supplementary material, available to authorized users.
Brought to you by | Karolinska Institute Authenticated Download Date | 6/3/15 11:00 AM
Rinat Gabbay-Benziv, Lauren E. Doyle, Miriam Blitzer and Ahmet A. Baschat*
First trimester prediction of maternal glycemic status Abstract Objective: To predict gestational diabetes mellitus (GDM) or normoglycemic status using first trimester maternal characteristics. Methods: We used data from a prospective cohort study. First trimester maternal characteristics were compared between women with and without GDM. Association of these variables with sugar values at glucose challenge test (GCT) and subsequent GDM was tested to identify key parameters. A predictive algorithm for GDM was developed and receiver operating characteristics (ROC) statistics was used to derive the optimal risk score. We defined normoglycemic state, when GCT and all four sugar values at oral glucose tolerance test, whenever obtained, were normal. Using same statistical approach, we developed an algorithm to predict the normoglycemic state. Results: Maternal age, race, prior GDM, first trimester BMI, and systolic blood pressure (SBP) were all significantly associated with GDM. Age, BMI, and SBP were also associated with GCT values. The logistic regression analysis constructed equation and the calculated risk score yielded sensitivity, specificity, positive predictive value, and negative predictive value of 85%, 62%, 13.8%, and 98.3% for a cut-off value of 0.042, respectively (ROCAUC – area under the curve 0.819, CI – confidence interval 0.769–0.868). The model constructed for normoglycemia prediction demonstrated lower performance (ROC-AUC 0.707, CI 0.668–0.746). Conclusions: GDM prediction can be achieved during the first trimester encounter by integration of maternal
*Corresponding author: Ahmet A. Baschat, Obstetrics, Gynecology and Reproductive Science, University of Maryland Medical Center, Baltimore, MD, USA; and Department of Gynecology and Obstetrics, Johns Hopkins School of Medicine, Baltimore, MD, USA, E-mail: [email protected] Rinat Gabbay-Benziv and Lauren E. Doyle: Obstetrics, Gynecology and Reproductive Science, University of Maryland Medical Center, Baltimore, MD, USA Miriam Blitzer: Department of Pediatrics, University of Maryland Medical Center, Baltimore, MD, USA
characteristics and basic measurements while normoglycemic status prediction is less effective. Keywords: Gestational diabetes mellitus (GDM); normoglycemia state; prediction model. DOI 10.1515/jpm-2014-0149 Received April 30, 2014. Accepted July 15, 2014. Previously published online August 19, 2014.
Introduction Gestational diabetes mellitus (GDM) complicates between 5% and 8% of pregnancies with rates over 30% in highrisk settings and is associated with maternal, fetal, and neonatal adverse outcomes [7, 10, 11, 18, 22, 24]. The Hyperglycemia Adverse Pregnancy Outcomes (HAPO) Study showed that the relationship between adverse outcomes and maternal glucose levels is a continuum that already exists below the threshold considered diagnostic for GDM [8]. The maternal and neonatal risks rise further in women with overt GDM and early identification and rigorous treatment offer the opportunity of improved outcomes [4, 15, 16]. Gestational diabetes mellitus is diagnosed based on the recommendations from the American College of Obstetricians and Gynecologists for universal screening at 24–28 weeks’ of gestation with a 50-g glucose challenge test (GCT) followed by a diagnostic 100-g oral glucose tolerance test (OGTT) for screen-positive women. This approach is limited by both late third trimester diagnosis of GDM, which has the potential to allow too little time to optimize management, as well as universal screening of a large population of women who are at low risk for glycemic complications. The ability to stratify glycemic risks in the first trimester offers potential advantages of earlier diagnostic testing for high-risk women as well as avoids universal screening for women at low risk. While previous studies have shown that first trimester prediction of women at high risk for GDM is feasible [5, 6, 14, 20, 21, 25, 26, 29–31] the stratification of low-risk women requires knowledge of the blood
Brought to you by | Karolinska Institute Authenticated Download Date | 6/3/15 11:00 AM
284 Gabbay-Benziv et al., First trimester prediction of maternal glycemic status sugar values observed after a GCT. It is the aim of our study to develop a first trimester prediction that allows stratification of women at risk for GDM and those at low risk for glycemia-related risks that may not require universal screening.
Methods This analysis was based on data from a previously published prospective cohort study designed to develop a first trimester screening model for placental dysfunction [1]. The study was conducted in the Baltimore metropolitan area between 2007 and 2010 and was approved by the Institutional Review Board of all participating centers. Women presenting with a singleton intrauterine pregnancy between 11 and 14 weeks’ gestation were offered participation. After obtaining written informed consent, a standardized questionnaire ascertaining general health and prepregnancy measures was completed, and maternal physical examination, ultrasound examination, and blood sampling were performed as previously described [12]. For this subanalysis we included women who had prenatal care and subsequent GDM screening at the study centers. Maternal demographics, medical and obstetric history, prepregnancy maternal weight, and smoking status were recorded. Maternal height, weight, and blood pressure were obtained. Body mass index (BMI) was calculated as weight (kg)/height (m)2. Peripheral blood samples were obtained and transmitted on a filter paper to the determining laboratory (NTD Labs, Perkin Elmer, Melville, NY, USA) for analysis. Absolute measurements of pregnancy-associated plasma protein-A, free β-human chorionic gonadotropin, placental growth factor, and alpha-fetoprotein were converted to multiples of the median utilizing reference ranges of a population with documented normal outcome [13]. All participating study centers implement universal screening for GDM. A screening 50-g GCT was performed at 24–28 weeks’ gestation for all patients. Women with a 1-h value ≥ 135 mg/dL were scheduled for a 100-g 3-h diagnostic OGTT. Women with a 1-h value lower than 135 mg/dL had OGTT test in cases of abnormal fetal ultrasound findings (macrosomia or polyhydramnios) depending on the managing physician. A single value > 2 00 mg/dL on the GCT, or two abnormal values on the OGTT, based on Carpenter-Coustan thresholds were diagnostic for GDM [2]. Monitoring and therapy of glycemic control was at the discretion of the managing physician. Women were considered normoglycemic, at low risk for GDM, when the 1-h GCT maternal glucose level was lower than 135 mg/ dL and all 4 glucose levels at OGTT, whenever obtained, were normal. Therefore, even women with one abnormal value at OGTT were not considered normoglycemic (as we targeted for the lowest risk group). Source documentation for pregnancy outcome and delivery details was retrieved by research personnel and transmitted to the research office. Study data were collected, validated, and entered into a dedicated study database by trained personnel. Large for gestational age (LGA) was defined as birth weight > 90% percentile for gestational age according to local growth charts and macrosomia was defined as birth weight > 4000 g.
Statistical analysis Univariate analysis was performed to study the distribution of characteristics stratified by subsequent GDM or normoglycemia. We used χ2-test for categorical variables and Mann-Whitney U-test for continuous variables. Linear regression was utilized to identify among the continuous variables the primary determinants of the 1-h glucose levels following the GCT. Utilizing the significant independent variables, a prediction model for GDM was developed using multivariable logistic regression with backward stepwise elimination. Odds ratio (OR) was derived as the exponential function of the regression coefficient for each variable. Patient risk for GDM was then constructed from the coefficient of each variable using the formula: GDM probability = 1/(1+e–(t)) where e is the base of the natural logarithm (about 2.718) and t is the combination of explanatory variables, derived from the logistic regression analysis. The same approach was followed to predict normoglycemia. The predictive performance and optimal risk scoring of each model along with their respective sensitivity, specificity, positive predictive value (PPV), and negative predictive value (NPV) were determined by receiver operating characteristics (ROC) curve analysis and calculation of the area under the curve (AUC). The goodness-of-fit of the models was evaluated with the Hosmer-Lemeshow statistical test. Low Hosmer-Lemeshow P values ( 92 mg/dL; 1 h > 180 mg/dL; 2 h > 153 mg/dL) as these values were correlated with diabetes-associated adverse outcomes [9]. Women in our study were screened for GDM using the two-step approach and 100-g GTT, thus containment of the HAPO thresholds on our data was inaccurate. Be that as it may, predicting normoglycemia based on the HAPO thresholds in our population (given the limitations) yielded similar predicting performance (ROC-AUC 0.736, 95% CI 0.689–0.782), available upon request. Both our prediction models used prior GDM as one of their variables. Considering that 46.2% of our cohort were nullipara, we wanted to be sure that the algorithms constructed fit all. We, therefore, performed two additional analyses. First, including only the nulliparous or the multiparous women using same variables and coefficients in the logistic regression. This yielded AUC 0.805 (95% CI 0.722–0.889) and 0.821 (95% CI 0.764–0.879), respectively. Second, substituting the variable “prior GDM” in the regression by two variables: “multiparous with prior GDM” and “multiparous without prior GDM”. In this case, ROC curve analysis yielded AUC 0.812 (95% CI 0.762– 0.861). These additional analyses showed no significant difference in the prediction performance of the model. Our study has several limitations. Only women with an abnormal screening GCT received an OGTT. Accordingly we may have under-diagnosed GDM given the false negative of GCT. Our prepregnancy maternal characteristics were based on questionnaires and they open up the possibility of ascertainment bias. Moreover, given our cohort size we did not use a validation cohort. It would
be interesting to examine our algorithm performance in a parallel cohort with different basic characteristics compared to our population. We did not determine any additional serum analytes that may improve our predictive accuracy of normal and abnormal glycemic status and finally, as stated before, we feel that prediction of normoglycemia would be more appropriate analyzing population using the IADPSG criteria for diagnosis of GDM. In contrast, our strength derives from development of the model by identifying factors through their relationship with absolute glucose levels as well as the diagnostic cutoffs for GDM. This ameliorates some of the statistical risk of collinearity and over fitting of the logistic model. Accordingly, our model may have wider applicability because it is less likely to be influenced by disease prevalence.
Conclusion Maternal age, prior history of GDM, ethnicity, and prepregnancy BMI can identify the majority of women than develop GDM. Measurement of maternal blood pressure, height, and weight at the time of first trimester screening improves prediction to over 80%. In contrast, normoglycemia is less predictable and accordingly first trimester screening with the aforementioned factors does not obviate universal screening.
References [1] Baschat AA, Magder LS, Doyle LE, Atlas RO, Jenkins CB, Blitzer MG. Prediction of preeclampsia utilizing the first trimester screening examination. Am J Obstet Gynecol. 2014. [Epub ahead of print]. [2] Carpenter MW, Coustan DR. Criteria for screening tests for gestational diabetes. Am J Obstet Gynecol. 1982;144:768–73. [3] Catalano PM, Nizielski SE, Shao J, Preston L, Qiao L, Friedman JE. Down regulated IRS-1 and PPARc in obese women with gestational diabetes: relationship to FFA during pregnancy. Am J Physiol Endocrinol Metab. 2002;282:E522–33. [4] Crowther CA, Hiller JE, Moss JR, McPhee AJ, Jeffries WS, Robinson JS, et al. Effect of treatment of gestational diabetes mellitus on pregnancy outcomes. N Engl J Med. 2005;352:2477–86. [5] D’Anna R, Baviera G, Corrado F, Giordano D, Recupero S, Di Benedetto A. First trimester serum neutrophil gelatinaseassociated lipocalin in gestational diabetes. Diabet Med. 2009;26:1293–5. [6] Georgiou HM, Lappas M, Georgiou GM, Marita A, Bryant VJ, Hiscock R, et al. Screening for biomarkers predictive of gestational diabetes mellitus. Acta Diabetol. 2008;45:157–65. [7] Hadar E, Yogev Y. Translating the HAPO Study into new diagnostic criteria for GDM? From HAPO to IADPSG and back to O’Sullivan. Clin Obstet Gynecol. 2013;56:758–73.
Brought to you by | Karolinska Institute Authenticated Download Date | 6/3/15 11:00 AM
Gabbay-Benziv et al., First trimester prediction of maternal glycemic status 289 [8] HAPO Study Cooperative Research Group, Metzger BE, Lowe LP, Dyer AR, Trimble ER, Chaovarindr U, et al. Hyperglycemia and adverse pregnancy outcomes. N Engl J Med. 2008;358:1991–2002. [9] International Association of Diabetes and Pregnancy Study Groups. International Association of Diabetes and Pregnancy Study Groups recommendations on the diagnosis and classification of hyperglycemia in pregnancy. Diabetes Care. 2010;33:676–82. [10] Jenum AK, Morkrid K, Sletner L, Vangen S, Torper JL, Nakstad B, et al. Impact of ethnicity on gestational diabetes identified with the WHO and the modified International Association of Diabetes and Pregnancy Study Groups criteria: a populationbased cohort study. Eur J Endocrinol. 2012;166:317–24. [11] Jiwani A, Marseille E, Lohse N, Damm P, Hod M, Kahn JG. Gestational diabetes mellitus: results from a survey of country prevalence and practices. J Matern Fetal Neonatal Med. 2012;25:600–10. [12] Kasdaglis T, Aberdeen G, Turan O, Kopelman J, Atlas R, Jenkins C, et al. Placental growth factor in the first trimester: relationship with maternal factors and placental Doppler studies. Ultrasound Obstet Gynecol. 2010;35:280–5. [13] Krantz D, Hallahan T, Ravens R, He K, Cuckle H, Sherwin J, et al. First trimester Down syndrome screening with dried blood spots using a dual analyte free beta hCG and PAPP-A immunofluorometric assay. Prenat Diagn. 2011;31:869–74. [14] Lain KY, Daftary AR, Ness RB, Roberts JM. First trimester adipocytokine concentrations and risk of developing gestational diabetes later in pregnancy. Clin Endocrinol (Oxf). 2008;69:407–11. [15] Langer O, Rodriguez DA, Xenakis EM, McFarland MB, Berkus MD, Arrendondo F. Intensified versus conventional management of gestational diabetes. Am J Obstet Gynecol. 1994;170;1036–47. [16] Langer O, Yogev Y, Most O, Xenakis MJ. Gestational diabetes: the consequences of not treating. Am J Obstet Gynecol. 2005;192:989–97. [17] Miller JL, de Veciana M, Turan S, Kush M, Manogura A, Harman CR, et al. First trimester detection of fetal anomalies in pregestational diabetes using nuchal translucency, ductus venosus Doppler and maternal glycosylated hemoglobin. Am J Obstet Gynecol. 2013;218:385, e1–8. [18] Moses RG, Morris GJ, Petocz P, San Gil F, Garg D. The impact of potential new diagnostic criteria on the prevalence of gestational diabetes mellitus in Australia. Med J Aust. 2011;194:338–40. [19] Nanda S, Savvidou M, Syngelaki A, Akolekar R, Nicolaides KH. Prediction of gestational diabetes mellitus by maternal factors and biomarkers at 11 to 13 weeks. Prenat Diagn. 2011;31:135–41. [20] Naylor CD, Sermer M, Chen E, Farine D. Selective screening for gestational diabetes mellitus. Toronto Trihospital Gestational Diabetes Project Investigators. N Engl J Med. 1997;337:1591–6.
[21] Nicolaides KH. A model for a new pyramid of prenatal care based on the 11 to 13 weeks’ assessment. Prenat Diagn. 2011;31:3–6. [22] O’Sullivan EP, Avalos G, O’Reilly M, Dennedy MC, Gaffney G, Dunne F, et al. Atlantic diabetes in pregnancy (DIP): the prevalence and outcomes of gestational diabetes mellitus using new diagnostic criteria. Diabetologia. 2011;54:1670–5. [23] Phaloprakarn C, Tangjitgamol S, Manusirivithaya S. A risk score for selective screening for gestational diabetes mellitus. Eur J Obstet Gynecol Reprod Biol. 2009;145:71–5. [24] Reyes-Muñoz E, Parra A, Castillo-Mora A, Ortega-González C. Effect of the diagnostic criteria of the International Association of Diabetes and Pregnancy Study Groups on the prevalence of gestational diabetes mellitus in urban Mexican women: a cross-sectional study. Endocr Pract. 2012;18:146–51. [25] Savvidou M, Nelson SM, Makgoba M, Messow CM, Sattar N, Nicolaides K. First-trimester prediction of gestational diabetes mellitus: examining the potential of combining maternal characteristics and laboratory measures. Diabetes. 2010;59:3017–22. [26] Smirnakis KV, Plati A, Wolf M, Thadhani R, Ecker JL. Predicting gestational diabetes: choosing the optimal early serum marker. Am J Obstet Gynecol. 2007;196:410. [27] Syngelaki A, Bredaki FE, Vaikousi E, Maiz N, Nicolaides KH. Body mass index at 11-13 weeks’ gestation and pregnancy complications. Fetal Diagn Ther. 2011;30:250–65. [28] Teede HJ, Harrison CL, Teh WT, Paul E, Allan CA. Gestational diabetes: development of an early risk prediction tool to facilitate opportunities for prevention. Aust NZ J Obstet Gynaecol. 2011;51:499–504. [29] Tul N, Pusenjak S, Osredkar J, Spencer K, Novak-Antolic Z. Predicting complications of pregnancy with first-trimester maternal serum free-betahCG, PAPP-A and inhibin-A. Prenat Diagn. 2003;23:990–6. [30] Van Leeuwen M, Opmeer BC, Zweers EJ, van Ballegooie E, ter Brugge HG, de Valk HW, et al. Estimating the risk of gestational diabetes mellitus: a clinical prediction model based on patient characteristics and medical history. Br J Obstet Gynaecol. 2010;117:69–75. [31] Wolf M, Sandler L, Hsu K, Vossen-Smirnakis K, Ecker JL, Thadhani R. First-trimester C-reactive protein and subsequent gestational diabetes. Diabetes Care. 2003;26:819–24. [32] Zhang S, Folsom AR, Flack JM, Liu K. Body fat distribution before pregnancy and gestational diabetes: findings from Coronary Artery Risk Development in Young Adults (CARDIA) Study. Br Med J. 1995;311:1139–40. The authors stated that there are no conflicts of interest regarding the publication of this article. Supplemental Material: The online version of this article (DOI: 10.1515/jpm-2014-0149) offers supplementary material, available to authorized users.
Brought to you by | Karolinska Institute Authenticated Download Date | 6/3/15 11:00 AM
