Kirsten Grabowska - UBC Department of Obstetrics & Gynaecology

Jacobson et al 119Conclusion: In a large managed care organization, glyburide was at least as effective as insulin inachieving glycemic control and similar birth weights in women with GDM who failed diettherapy. The increased risk of preeclampsia and phototherapy in the glyburide group warrantfurther study.Ó 2005 Mosby, Inc. All rights reserved.Gestational diabetes mellitus (GDM) complicates 2%to 5% of pregnancies and is associated with significantmaternal and neonatal morbidity and mortality. 1 In theUnited States, traditional therapy for women withGDM who fail diet therapy has been insulin. In awell-controlled randomized trial, Langer et al 2 demonstratedthat glyburide was a clinically effective alternativeto insulin in women with GDM. Despite limitedclinical studies, recent expert opinion has suggestedglyburide as an alternative first choice for GDM, anda survey in 2003 found that 13% of obstetriciangynecologistsbegin with glyburide for patients requiringtherapy beyond diet. 3,4 The aim of this study is to reportour experience after the introduction of a protocol forglyburide treatment of GDM in a large managed careorganization.Material and methodsWe performed a retrospective study in women diagnosedwith GDM requiring medication in the 2-yearperiod before the introduction of glyburide to womenwho received a diagnosis of GDM in the 2 years after itsintroduction at a large managed care organization. Thesetting for this study was the Kaiser Permanente NorthernCalifornia Medical Care Program (KPNC). KPNCis a prepaid group-model managed care organizationthat provides comprehensive medical services to morethan 3 million members in a 14-county region, servingapproximately 30% of the population in the area. TheKPNC membership approximates the sociodemographicsof the general population in the region, exceptfor the extreme tails of the socioeconomic distribution.In February 2001, the KPNC Perinatology Groupintroduced guidelines for the use of glyburide for thetreatment of GDM.Eligibility for the study was limited to women withsingleton pregnancies who had GDM diagnosed between12 and 34 weeks between January 1999 andDecember 2002. Additional inclusion criteria includedan abnormal 50 g glucose load test (GLT) of 140 mg/dLor greater and a diagnostic 3-hour glucose tolerance test(GTT), using National Diabetes Data Group (NDDG)criteria (fasting R105, 1 hour R190, 2 hours R165, 3hours R145 mg/dL). 5 Patients were excluded if they hada fasting plasma glucose greater than 140 mg/dL onGTT. Patients delivered at 9 hospitals: 5 with neonatalintensive care units (NICUs) and 3 with level II nurseries.First- and/or second-trimester ultrasounds arestandard practice in KPNC and were used for datingmost patients. Two primary groups were defined: theinsulin group with GDM diagnosed in 1999 through2000 and the glyburide group with GDM diagnosed in2001 through 2002. We also report on data collected onpatients with GDM diagnosed in 2001 through 2002who met study criteria but were treated with insulin notglyburide (ING group).General management of patients diagnosed withGDM included: nutritional counseling on a diabeticdiet for pregnancy with 3 meals and 3 snacks; andinstruction on self-monitoring capillary blood glucose(BS) with a glucometer. Patients were advised to monitorBS 4 times a day: fasting and either 1 or 2 hourspostprandials (per individual provider preference) forbreakfast, lunch, and dinner. Specific goals for plasmafasting, 1 hour, and 2 hours postprandials were 100, 155,and 130 mg/dL, respectively. Patients were required tohave capillary BS evaluation for a minimum of 3 dayson an appropriate diet before initiating glyburide. Glyburidewas begun with an initial daily dose of 2.5 mgwith the morning meal. If glycemic control goals werenot met, as defined by the individual provider, the dosewas increased by 2.5 mg initially and thereafter by 5 mgweekly. If the dose exceeded 10 mg daily, twice dailydosing was considered. If glycemic goals were not met,as defined by the individual provider, on a maximumdaily dose of 20 mg, patients were switched to insulin.Patients started on glyburide were referred to theRegional Perinatal Nursing Services (RPNS) for entryinto a database for outcome assessment. At the discretionof the referring provider, a patient could befollowed by the regional Diabetes Home ManagementProgram (DHMP) or by their local GDM program. TheDHMP involved regular telephone contact by nursesskilled in the management of GDM, as well as entry ofhome capillary BS values into the Perinatal ManagementSystem database (PERMS) for outcome assessment.Reviews of clinical information databases and chartswere performed to obtain demographic data. Body massindex (BMI) was calculated by using first documentedweight during pregnancy. Preterm delivery (PTD) wasdefined as delivery less than 37 weeks’ gestationalage. According to the gestational age-specific weight

120 Jacobson et alTable IComparison of demographics and background characteristics of insulin and glyburide groupsInsulinn = 268Glyburiden = 236Age (y)* 32.1 G 5.2 32.8 G 5.4 .14Nulliparity no. (%) 91 (34) 78 (33) .83BMI (kg/m 2 )* 31.9 G 6.8 (n = 244) 30.6 G 7.0 (n = 220) .04 yBMI R 30 kg/m 2 (%) 134/244 (55) 104/220 (47) .10BMI R 40 kg/m 2 (%) 32/244 (13) 27/220 (12) .79Ethnicity no. (%) .001White 116 (43) 65 (28)Hispanic 66 (25) 56 (24)Black 10 (4) 9 (4)Asian 64 (24) 88 (37)Family history diabetes no. (%) 148 (55) 137 (58) .52History GDM no. (%) 52 (19) 42 (18) .64Previous macrosomic delivery no. (%) 49 (18) 44 (19) .92GLT: 1-h value (mg/dL)* 179.1 G 26.1 175.9 G 26.4 .11 yGestational age (wk)* 23.3 G 6.5 22.8 G 6.4 .15 yGTT: Fasting (mg/dL)* 105.4 G 12.9 102.4 G 14.2 .005 y1-h value (mg/dL)* 222.8 G 28.9 220.0 G 27.2 .48 y2-h value (mg/dL)* 197.4 G 33.6 194.7 G 32.1 .44 y3-h value (mg/dL)* 139.6 G 37.5 137.3 G 35.4 .59 yGestational age at diagnosis (wk)* 26.3 G 5.3 25.5 G 5.3 .02 yFasting R 105 mg/dL no. (%) 144 (54) 103 (44) .02Gestational age dietary education (wk)* 27.8 G 5.4 (n = 250) 27.2 G 5.2 (n = 224) .08 yWeight gain (lb)* 19.7 G 11.5 (n = 250) 22.3 G 12.9 (n = 224) .06 yGestational age medications began (wk)* 30.4 G 4.8 30.3 G 4.5 .81* Values are mean G SD.y P value from Wilcoxon 2 sample, 2 sided.P valuedistribution of the study population, infants were consideredlarge for gestational age (LGA), or small forgestational age (SGA) if their birth weight exceeded, orwas below, the 90th or 10th percentiles, respectively, withthe use of KPNC-derived algorithms. These algorithmswere based on 72,863 deliveries at KPNC from 1996through 1999 of women with live singleton births who hada GLT performed between 24 and 28 weeks. Macrosomiawas defined as birth weight 4,000 g or greater.Neonatal hyperbilirubinemia was defined as a totalbilirubin 12 mg/dL or greater within the first 7 days ofbirth. Neonatal polycythemia was defined as hematocritgreater than 60% and neonatal hypocalcemia wasdefined as calcium 7.0 mg/dL or less within 3 days ofbirth. Decision to check any laboratory value was madeby local providers on the basis of clinical indications.Several delivering facilities relied on capillary heel sticksfor BS evaluation performed in the nursery or neonatalintensive care unit (NICU) and values were not enteredinto the regional database. Definition of neonatal hypoglycemiawas therefore based on discharge coding. AnAdmission, Discharge, Transfer (ADT) database withall KPNC hospitalizations was searched with the use ofthe appropriate International Classification of Diseasesninthrevision-Clinical Modification (ICD-9-CM) codesfor the following outcomes: neonatal hypoglycemia,preeclampsia, congenital anomalies, birth injury, neonatalexchange transfusions, and phototherapy.The KPNC Neonatal Minimum Data Set (NMDS)contains comprehensive, high-quality neonatal and perinataldata on all infants admitted to all 5 NICUs and 2of the intermediate care nurseries involved in this study.NMDS database was used to obtain data on NICUadmissions including length of stay and need for oxygen/assistedventilation. NMDS defines oxygen/assistedventilation as an infant receiving supplemental oxygen(O 2 at 21% for O1 hour) or requiring ventilation duringhospitalization.Although all GDM patients were instructed to selfmonitorcapillary BS 4 times per day, glycemic controlcould only be assessed in patients managed by theDHMP where capillary BS values were routinely enteredin the PERMS database and therefore easily retrievable.During the period of this study patients used eitherwhole blood or plasma glucometers and a conversionfactor of !1.1 was used to convert whole to plasmavalues. Maternal hypoglycemia was defined as BSless than 60 mg/dL. Treatment failure with glyburidewas defined as beginning insulin therapy after startingglyburide.

122 Jacobson et alTable IV Comparison of posttreatment plasma glucose values for insulin and glyburide PERMS subgroupsInsulin Glyburide P valueFasting: No. women 122 137Fasting: Value (mg/dL)* 97.7 G 12.2 90.2 G 12.7 ! .001Fasting: No. values O100 mg/dL (%) 2182/6141 (35.5) 1044/6623 (15.8) ! .0011 h: No. women 89 1211 h: Value (mg/dL)* 137.8 G 23.6 131.4 G 23.3 ! .0011 h: No. values O155 mg/dL (%) 2182/12175 (17.9) 1772/16479 (10.8) ! .0012 h: No. women 33 162 h: Value (mg/dL)* 118.8 G 19.6 117.6 G 23.2 ! .052 h: No. values O130 mg/dL (%) 1183/4448 (26.6) 452/1873 (24.1) .04Maternal hypoglycemia: Values !60 mg/dL (%) 19/22,764 (0.08) 50/24,975 (0.20) ! .001Mean fasting and/or postprandial within goal: No. women (%) y 77 (63) 118 (86) ! .001Mean days tested z 66.7 G 34.3 62.6 G 32.2 .33Mean tests per day 2.8 2.9* Values are mean G SD.y Within goal defined as fasting !100 and either 1-h postprandial !155 or 2-h postprandial !130 mg/dL.z Calculated from date medication initiated to date of delivery.daily insulin dose was available for 249 women: mean34.4 G 28.1 U (median 28, range 2-242). Maximumdaily glyburide dose was available for 229 women: mean5.6 G 4.6 mg (median 5, range 1.25-20).Table II shows the maternal and neonatal outcomesof the insulin and glyburide groups. There were nosignificant differences for most outcomes; however,women in the glyburide group had a higher incidenceof preeclampsia (12%, 6%, P = .02) and neonates in theglyburide group were more likely to receive phototherapy(9%, 5%, P ! .05). Neonates in the insulin groupwere more likely to be admitted to the NICU (24%,15%, P = .008) though they had a significantly shorterNICU length of stay compared to the glyburide group(4.3 G 9.6 vs 8.0 G 10.1 days, P = .002). No significantdifferences were noted between the 2 groups for the mostcommon discharge diagnoses in the ADT database forneonates admitted to the NICU (data not shown).Results of logistic regression analysis are shown inTable III. Adjusted ORs remain significant for preeclampsia,phototherapy, and NICU admission aftercontrolling for variables shown to differ between the 2groups. There were no neonatal deaths, lethal anomalies,or exchange transfusions in either group. Both theinsulin and glyburide groups had 4 infants with 7identified anomalies. There was 1 intrauterine fetaldemise at 37 weeks in the glyburide group.One hundred twenty two (46%) women in the insulingroup and 137 (58%) in the glyburide group werefollowed by the DHMP and had self-reported capillaryglucose values entered into the PERMS database andwere therefore easily retrievable for analysis. Women inthe glyburide group had significantly lower posttreatmentmean fasting and postprandial BSs (Table IV).Average daily testing for women on insulin was 2.8 perday compared with 2.9 per day for women on glyburide.There were no significant differences in outcomes betweenthese 2 subgroups (data not shown), exceptneonates admitted to the NICU in the glyburide grouphad a longer length of stay (2.1 G 3.9 vs 4.7 G 5.5 days,P ! .001). Maternal hypoglycemia, though rare, wasmore common in the glyburide group (0.20% vs 0.08%,P ! .001). Logistic regression modeling in thesesubgroups found that compared with insulin, glyburidetreatment was significantly more likely to be associatedwith achieving mean glucose goals (adjusted OR 0.27,95% CI 0.13-0.52).Twenty-eight (12%) women failed glyburide and wereswitched to insulin: 8 for side effects primarily attributedto hypoglycemia, 14 for poor control, and for 6 thedocumentation was unclear. These women had thefollowing characteristics (means): BMI 31.6 kg/m 2 ,fasting on GTT 104.8 mg/dL, and 25.6 lb pregnancyweight gain. Only 3 women switched for poor controlwere on maximum dose of 20 mg/d. Insulin was startedat a mean gestational age of 31.9 G 4.2 weeks andreached a mean insulin dose of 46.8 G 24.2 U/d (range5-90). Mean birth weight was 3858 g and there were nobirth injuries in this group. An additional 11 (5%)women discontinued glyburide, most for side effectsattributed to hypoglycemia, and never started insulin.These women had the following characteristics (means):BMI 30.8 kg/m 2 , fasting on GTT 94.8 mg/d, 27.7 lbpregnancy weight gain. Mean dose of glyburide was 7.2mg/d. Mean birth weight was 3893 g and there were nobirth injuries in this group.Three neonates in the insulin group had birth injuries:1 clavicle fracture, 1 brachial plexus injury, and 1 boneinjury. Eight neonates in the glyburide group had birthinjuries: 4 clavicle fractures, 1 brachial plexus injury,

Jacobson et al 123and 3 bone injuries. Mean birth weights for injuredneonates in the insulin group was 4048 g and 3906 g inthe glyburide group. No infants in the ING group hadidentified birth injuries.Chart review did not reveal the reasons 80 eligiblewomen who had GDM diagnosed in 2001 through 2002were initially treated with insulin instead of glyburide(ING group). The glyburide and ING groups weresimilar (data not shown) with a few notable exceptions:women in the ING group had higher mean BMI (33.9 vs30.6 kg/m 2 , P = .003), were less likely to identifythemselves as Asian (24%, 37%, P = .02), and had ahigher mean fasting on GTT (108.7 vs 102.4 mg/dL,P = .001).Maternal and neonatal outcomes in the glyburide andING groups were similar (data not shown), includingbirth weight (3661 G 629 g vs 3608 G 564 g, P = .48),preeclampsia (12%, 5%, P = .06), and phototherapy(9%, 4%, P = .11). Admission rates to the NICU weresimilar (15%, 10%, P = .27), though length of staywas longer for the glyburide group (8.0 G 10.1 days vs2.8 G 2.8 days, P = .01).Comparison of all women who received diagnoses in1999 through 2000 (insulin group) with all women whoreceived diagnoses in 2001 through 2002 (glyburide andING groups) demonstrated no significant differences indemographic or descriptive variables (data not shown)with 1 exception. Women who had GDM diagnosed in1999 through 2000 were less likely to identify themselvesas Asian (24%, 34%, P = .008) and more likely toidentify themselves as white (43%, 28%, P ! .001),There was an 18% increase in the number of womenmeeting study diagnostic criteria over the 2 periods.CommentOur retrospective study suggests that glyburide can beeffectively introduced for the treatment of women withGDM and fasting plasma glucose 140 mg/dL or less onGTT, into a large and diverse managed-care populationachieving results similar to insulin therapy. We found nosignificant differences in birth weight or rates of LGA.In addition, women treated with glyburide achievedbetter glycemic control, a finding that remained significantafter adjusting for characteristics that differedbetween the groups, including BMI, ethnicity, degreeof glucose intolerance, and gestational age at diagnosisof GDM. Although it is possible that women onglyburide had better compliance than those on insulin,among women with retrievable BS values, daily testingrates were similar between the 2 groups suggestingsimilar levels of compliance.Our study had several findings not previously reported.The glyburide group had a higher incidence ofpreeclampsia compared with the insulin group even aftercontrolling for confounders such as BMI and ethnicity.Although metformin, a biguanide, has been associatedwith increased risk of preeclampsia, this has not beennoted with the sulfonylureas. 7 Animal studies suggestglyburide inhibits vascular smooth muscle ATP-sensitivepotassium channel activity and increases systemic vascularresistance, and human in vitro studies suggest thatglyburide antagonizes cicletanine-induced relaxation inarteries from women with preeclampsia and thereforemay effect the natural vasodilatory substances of pregnancy.8,9Another unexpected finding was the higher rate ofphototherapy in the glyburide group. Paradoxically,both groups had similar rates of hyperbilirubinemia.The higher NICU admission rate in the insulin groupmay be due to practitioners selectively admitting infantsborn to women on insulin for a short observationperiod; however, the low rate of NICU admission inthe ING group would argue against this. Of concern isthe longer NICU length of stay in the glyburide groupthat could be due to their higher rate of phototherapy. Itis reassuring that all groups had similar rates of assistedventilation suggesting similar rates of respiratory distresssyndrome between the groups. The higher rate ofbirth injuries in the glyburide group, although notstatistically significant, warrants further investigation.The retrospective nature of our study lends itself toinherent biases. Comparison of patients treated in 1999through 2000 (insulin group) with those treated in 2001through 2002 (glyburide and ING groups) suggestedthere was a clear demographic change in the populationreflected by decreased white and increased Asian ethnicitiesover time. Although all patients identified in2001 through 2002 were eligible for glyburide, only 75%were initially treated with glyburide. Comparison of theglyburide with the ING group suggests a selection biasfor administration of glyburide, women receiving glyburidehad lower BMIs, lower mean fasting on GTT,and were more likely to be Asian. Logistic regressionmodeling controlling for ethnicity and these clinicalvariables still found a higher rate of preeclampsia, needfor phototherapy, and lower NICU admission rate inthe glyburide group. However, adjusted analysis maynot completely account for differences in the groups thatoccur because of population differences due to ourretrospective study design.Our study includes significantly more patients thanother studies published since the publication by Langeret al. 10-15 Our patient population is particularly diversewith Hispanics and Asians, although only 4% wereblack. Twelve percent of the glyburide group wereswitched to insulin and another 5% discontinued glyburideand never started insulin. Overall, this is significantlymore than the 4% reported by Langer et al, butmuch closer to current reports. 9-12 However, unlike thecontrolled trial of Langer et al with a strict dosing

124 Jacobson et alprotocol, management may not have been consistentamong physicians in our retrospective study.Although we found no significant differences in meanbirth weights and rates of LGA and macrosomiabetween our groups, our overall birth weights and rateswere greater in our study than Langer et al reports.However, several studies since have reported higherrates of LGA and macrosomia. 10,11,13-15 There areseveral possible explanations. Our sample includedmore women with risk factors such as family historyof DM and history of GDM as well as greater maternalage. Our study included more women with greaterglucose intolerance as evidenced by mean GLT valuesand by our use of NDDG cut-offs, reflected in ourhigher mean fasting, 1-hour, and 2-hour values on GTT.Langer’s group had lower fasting and postprandial BSgoals and their mean daily glyburide and insulin dosesappear higher than doses in our patients. It is possiblethat lower birth weights in the study by Langer et alreflect inadequate dosing in our retrospective study.However, our actual mean posttreatment BSs appearsimilar to the actual fasting means by Langer et al, andamong our subgroup that used 2-hour postprandialtesting, the means were less than their goal of 120 mg/dL.We found no significant difference in the rate of neonatalhypoglycemia between the groups, and although ouroverall rate was higher than Langer et al reported, it issimilar to that reported by others. 5,11 In addition, Langeret al used a strict laboratory definition for neonatalhypoglycemia and we used a definition based ondischarge summary data.Glyburide appears to be a safe and effective alternativeto insulin for the treatment of GDM in women withfasting plasma glucose 140 mg/dL or less on GTT andwho fail dietary therapy. Although potentially avoidingthe need for self-injection, the rate of discontinuation ina nonresearch setting warrants investigation into alternativeadministration protocols to improve compliance.Larger randomized trials are needed to investigate lessfrequent complications such as preeclampsia, NICUadmission, need for phototherapy, and birth injury.AcknowledgmentsWe thank Dayakar Beeravolu, Susan M. Shaheen, andIsaac J. Ergas for their assistance in database abstraction.References1. Coustan DR. Gestational diabetes. Washington: American Collegeof Obstetricians and Gynecologists; 2001. ACOG Practice Bulletin,Number 30.2. Langer O, Conway DL, Berkus MD, Xenakis EM, Gonzales O. Acomparison of glyburide and insulin in women with gestationaldiabetes mellitus. N Engl J Med 2000;343:1134-8.3. Gabbe SG, Graves CR. Management of diabetes mellitus complicatingpregnancy. Obstet Gynecol 2003;102:857-68.4. Gabbe SG, Gregory RP, Power ML, Williams SB, Schulkin J.Management of diabetes mellitus by obstetrician-gynecologists.Obstet Gynecol 2004;103:1229-34.5. National Diabetes Data Group. Classification and diagnosis ofdiabetes mellitus and other categories of glucose intolerance.Diabetes 1979;28:1039-57.6. Shapiro SS, Wilk MD. An analysis of variance test for normality(complete samples). Biometrika 1965;52:591-611.7. Hellmuth E, Damm P, Molsted-Pedersen L. Oral hypoglycaemicagents in 118 diabetic pregnancies. Diabet Med 2000;17:507-11.8. Keyes L, Rodman DM, Curran-Everett D, Morris K, Moore LG.Effect of KCATP channel inhibition on total and regionalvascular resistance in guinea pig pregnancy. Am J Physiol1998;275:H680-8.9. Ebeigbe AB, Cabanie M. In vitro vascular effects of cicletanine inpregnancy-induced hypertension. Br J Pharmacol 1991;103:1992-6.10. Gilson G, Murphy N. Comparison of oral glyburide with insulinfor the management of gestational diabetes mellitus in AlaskaNative women. Am J Obstet Gynecol 2002;187:S152.11. Velazquez MD, Bolnick J, Cloakey D, Gonzalez JL, Curet LB.The use of glyburide in the management of gestational diabetes.Obstet Gynecol 2003;101(4 suppl):88S.12. Conway DL, Gonzales O, Skiver D. Use of glyburide for thetreatment of gestational diabetes: the San Antonio experience.J Matern Fetal Neonatal Med 2004;15:51-5.13. Fines VL, Moore TR, Castle SL. A comparison of glyburide andinsulin treatment in gestational diabetes mellitus on infant birthweight and adiposity. Am J Obstet Gynecol 2003;189:S161.14. Chmait R, Dinise T, Moore T. Prospective observational study toestablish predictors of glyburide success in women with gestationaldiabetes mellitus. J Perinat 2004;10:617-22.15. Kremer CJ, Duff P. Glyburide for the treatment of gestationaldiabetes. Am J Obstet Gynecol 2004;190:1438-9.