Maternal Serum Prolactinand Prediction of Postpartumb-Cell Function and Riskof Prediabetes/DiabetesDiabetes Care 2016;39:1250–1258 | DOI: 10.2337/dc16-0043

OBJECTIVE

The insulin resistance of mid- to late pregnancy poses a physiologic stress test forthe pancreaticb-cells, which must respond bymarkedly increasing their secretionof insulin. This response is achieved through an expansion of b-cell mass inducedby the hormones prolactin and human placental lactogen (HPL). Conversely, thefuran fatty acid metabolite 3-carboxy-4-methyl-5-propyl-2-furanpropanoic acid(CMPF) has recently emerged as a negative regulator of b-cell function in preg-nancy. Given their respective roles in the b-cell response to the stress test ofgestation, we hypothesized that antepartum prolactin, HPL, and CMPFmay relateto a woman’s underlying glucoregulatory physiology and hence to her metabolicstatus after pregnancy.

RESEARCH DESIGN AND METHODS

Three hundred and sixty-sevenwomen underwentmeasurement of fasting serumprolactin, HPL, and CMPF in the late-2nd/early-3rd trimester, followed by an oralglucose tolerance test (OGTT) at 3 months postpartum that enabled assessmentof glucose tolerance, insulin sensitivity/resistance, and b-cell function (InsulinSecretion-Sensitivity Index-2 [ISSI-2]).

RESULTS

The postpartumOGTT identified 301 women with normal glucose tolerance (NGT)and 66 with prediabetes or diabetes. Serum prolactin in pregnancy was higher inwomen with postpartum NGT compared with those with postpartum prediabetes/diabetes (mean 98.2 vs. 80.2 ng/mL, P = 0.0003), whereas HPL and CMPF did notdiffer between the groups. On multiple linear regression analyses, antepartumprolactin was an independent determinant of postpartum ISSI-2 (b = 0.0016, t =2.96, P = 0.003). Furthermore, higher serum prolactin in pregnancy independentlypredicted a lower risk of postpartum prediabetes/diabetes (odds ratio 0.50, 95% CI0.35–0.72, P = 0.0002).

CONCLUSIONS

Serum prolactin in pregnancy predicts postpartum b-cell function and risk ofprediabetes/diabetes.

1Leadership Sinai Centre for Diabetes, MountSinai Hospital, Toronto, Ontario, Canada2Division of Endocrinology, University of Toronto,Toronto, Ontario, Canada3Lunenfeld-TanenbaumResearch Institute,MountSinai Hospital, Toronto, Ontario, Canada4Keenan Research Centre for Biomedical Science ofSt. Michael’s Hospital, Toronto, Ontario, Canada5Department of Nutritional Sciences, Universityof Toronto, Toronto, Ontario, Canada6Division of Obstetrics and Gynecology, MountSinai Hospital, Toronto, Ontario, Canada

Correspondingauthor:RaviRetnakaran, rretnakaran@mtsinai.on.ca.

Received 8 January 2016 and accepted 4 April2016.

This article contains Supplementary Data onlineat http://care.diabetesjournals.org/lookup/suppl/doi:10.2337/dc16-0043/-/DC1.

© 2016 by the American Diabetes Association.Readersmayuse this article as longas thework isproperly cited, the use is educational and not forprofit, and the work is not altered.

Ravi Retnakaran,1,2,3 Chang Ye,1

Caroline K. Kramer,1,2

Philip W. Connelly,2,4

Anthony J. Hanley,1,2,5 Mathew Sermer,6

and Bernard Zinman1,2,3

1250 Diabetes Care Volume 39, July 2016

PATH

OPHYS

IOLO

GY/COMPLICATIONS

The latter half of pregnancy is a state ofmarked insulin resistance that poses aphysiologic stress test for the pancreaticb-cells (1). For normal glucose homeo-stasis to be maintained in pregnancy,the b-cells must compensate for this in-sulin resistance by markedly increasingthe secretion of insulin (1). An insuffi-cient compensatory response will resultin maternal hyperglycemia, as occurs inthe setting of gestational diabetes mel-litus (GDM) (1,2). This insufficient re-sponse is indicative of an underlyingdefect in b-cell function in womenwho develop GDM that is also the path-ophysiologic basis for their high risk ofpostpartum progression to prediabetesand type 2 diabetes (T2D) in the yearsthereafter (3–6). Indeed, b-cell dys-function is the central defect in thepathophysiology of both GDM andT2D (2,7). As such, the adaptive re-sponse of the b-cells to the physiologicstress test of pregnancy may provideunique insight into a woman’s lifetimerisk of diabetes (1).Normal islet adaptation in pregnancy

is believed to be dependent upon amarked expansion of b-cell mass thatis stimulated by the circulating hor-mones prolactin and human placentallactogen (HPL) (8–11). Preclinical mod-els have demonstrated that prolactinand placentally derived HPL both bindto the prolactin receptor on the b-cellsand induce a series of downstream in-tracellular mediators that ultimatelystimulate b-cell growth and prolifera-tion (8–11). Indeed, the prolactin recep-tor has been shown to be essential forthe expansion of b-cell mass in preg-nancy (12,13). Whereas prolactin andHPL are key determinants of this normalphysiologic response, the furan fattyacid metabolite 3-carboxy-4-methyl-5-propyl-2-furanpropanoic acid (CMPF)was recently identified as a negativeregulator of insulin secretion and hencea mediator of b-cell dysfunction in GDM(14,15). Given their respective roles inthe b-cell response to the stress test ofpregnancy, we hypothesized that thesecirculating factors may be relevant to awoman’s underlying glucoregulatoryphysiology outside of gestation andhence to her risk of postpartum predia-betes and diabetes. Thus, our objectivein this study was to evaluate the longi-tudinal associations of serum prolactin,HPL, and CMPF in pregnancy with b-cell

function and glucose tolerance at 3months postpartum in a cohort ofwomenreflecting the full spectrum of gestationalglucose tolerance (from normal to GDM)and hence a broad range of future diabe-tes risk.

RESEARCH DESIGN AND METHODS

This analysis was conducted in the set-ting of a prospective observational co-hort study of early events in the naturalhistory of T2D inwhich a cohort ofwomenrecruited at the time of antepartumscreening for GDM is undergoing longi-tudinal metabolic characterization inpregnancy and the postpartum period.The study protocol has previously beendescribed in detail (16). In brief, at our in-stitution, all pregnantwomen are screenedfor GDM by 50-g glucose challenge test(GCT) late in the 2nd trimester, followedby referral for diagnostic oral glucose tol-erance test (OGTT) if the GCT is abnor-mal (blood glucose $7.8 mmol/L at 1 hpostchallenge). For this study, healthywomen are recruited either before orafter the GCT, and all participants thencomplete a 3-h 100-g OGTT (regardlessof the GCT result). As previously de-scribed (6,16), the recruitment of womenafter an abnormal GCT serves to enrichthe study population for those withGDM. At 3 months postpartum, partici-pants return to the clinical investigationunit for reassessment of metabolic statusby 2-h 75-g OGTT. The protocol has beenapproved by theMount Sinai Hospital Re-search Ethics Board, and all women pro-vide written informed consent for theirparticipation. The current analysis was re-stricted to 367 women with singletonpregnancies in whom serum CMPF, HPL,and prolactin weremeasured at the ante-partum OGTT and who had completedthe 3-month postpartum visit.

Evaluation of Women in Pregnancyand at Three Months PostpartumOn the morning of the OGTT in preg-nancy, interviewer-administered ques-tionnaires were completed pertainingto medical, obstetrical, and family his-tory. As previously described (16), theantepartum 3-h 100-g OGTT enabled as-certainment of gestational glucose tol-erance status as follows: 1) GDM,defined as two or more glucose valuesabove the National Diabetes Data Group(NDDG) (17) diagnostic criteria on theOGTT(fasting blood glucose $5.8 mmol/L, 1-h

glucose$10.6mmol/L, 2-h blood glucose$9.2 mmol/L, or 3-h blood glucose$8.1mmol/L); 2) gestational impaired glucosetolerance (GIGT), defined as only one glu-cose value above NDDG thresholds; and3) normal glucose tolerance (NGT), de-fined as no glucose values above NDDGthresholds. Women diagnosed with GDMwere referred to thediabetes-in-pregnancyclinic for clinical care, where they re-ceived glucose-lowering treatment inpregnancy, consisting of dietary/lifestylecounseling with or without antepartuminsulin therapy.

At 3 months postpartum, participantsreturned for a 2-h 75-g OGTT, on whichcurrent glucose tolerance status was de-fined according to Canadian DiabetesAssociation guidelines (18). Prediabetesrefers to impaired glucose tolerance(IGT), impaired fasting glucose (IFG), orcombined IFG and IGT. At this visit, par-ticipants also underwent physical exam-ination, with measurement of weightand waist circumference.

Laboratory Measurements on OGTTand Physiologic IndicesAll OGTTs were performed in the morn-ing after overnight fast, with venousblood samples drawn for the measure-ment of glucose and specific insulin atfasting and at 30, 60, and 120 min (and180 min in pregnancy) after the inges-tion of the glucose load. Specific insulinwas measured with the Roche-Elecsys-1010 immunoassay analyzer and elec-trochemiluminescence immunoassaykit (Roche Diagnostics, Laval, Canada).

Glycemia was assessed with the areaunder the glucose curve (AUCglucose) onthe OGTT, calculated by trapezoidalrule. Insulin sensitivity was measuredwith the Matsuda index, an establishedmeasure of whole-body insulin sensi-tivity that has been validated againstthe euglycemic-hyperinsulinemic clamp(19). Insulin resistance (primarily hepatic)was evaluated with HOMA (HOMA-IR)(20). b-Cell function was assessed withthe Insulin Secretion-Sensitivity Index-2(ISSI-2), a validated measure of b-cellcompensation that is analogous to thedisposition index obtained from the in-travenous glucose tolerance test, againstwhich it has been directly validated(21,22). ISSI-2 is defined as the productof 1) insulin secretion measured by theratio of the AUCinsulin to the AUCglucoseand 2) insulin sensitivity measured by

care.diabetesjournals.org Retnakaran and Associates 1251

Matsuda index (21,22). A second mea-sure of b-cell function was providedby the insulinogenic index divided byHOMA-IR (insulinogenic index/HOMA-IR), an established measure defined asthe incremental change in insulin be-tween 0 and 30 min divided by the in-cremental change in glucose over thesame interval, divided by HOMA-IR (16).

Measurement of HPL, Prolactin, andCMPFCMPF, HPL, and prolactin were measuredfrom fasting serum at the OGTT in preg-nancy. CMPF was measured by ELISA kitno. BG-HUM10440 from Novatein Bio-sciences (Woburn, MA). Samples wereanalyzed in 11 runs and were run with-out dilution. A serum pool was used as acalibrator, and the assigned value of 49ng/mL was determined as the mean

value from all runs. The within-run coef-ficient of variation (CV) averaged 16.5%for serum duplicates. HPL was measuredby ELISA no. 20-HPLHU-E01 from Alpco(Salem, NH). Two quality-control poolswere included in each run with mean3.44 mg/L, CV 14.4%, and mean 15.0mg/L, CV 15.4%. The manufacturer’sstandards were calibrated against theNational Institute for Biological Stan-dards and Control international stan-dard for HPL International ReferencePreparation (73/545). All pipetting ofcontrols, samples, and standards wascompleted in 3 min with a Janus liquidhandler equipped with an eight-tip Va-rispan arm (Perkin-Elmer, Woodbridge,Ontario, Canada). Prolactin was mea-sured using the Meso Scale Discovery96-well multiarray human prolactin as-say (Meso Scale Diagnostics, Rockville,

MD). A control serum pool had a meanof 3.49 ng/mL and CV 8.8% (n = 11 runs).

Statistical AnalysesAll analyses were conducted using SAS9.4 (SAS Institute, Cary, NC). Continuousvariables were tested for normality ofdistribution, and natural log transforma-tions of skewed variables were used,where necessary, in subsequent analy-ses. The characteristics of women withNGT, prediabetes, and diabetes at 3months postpartum were compared us-ing the Kruskal-Wallis test for continu-ous variables and Fisher exact test forcategorical variables (Table 1). For Ta-bles 2 and 3, multiple linear regressionanalyses were performed to determinewhether serum levels of CMPF, HPL,and prolactin in pregnancy could predictthe following metabolic outcomes at 3

Table 1—Demographic, clinical, and metabolic characteristics of study population stratified into groups based on glucosetolerance status at 3 months postpartum

NGT (n = 301) Prediabetes (n = 60) Diabetes (n = 6) P

At OGTT in pregnancyWeeks’ gestation (weeks) 30 (28–32) 29 (28–30) 28 (27–28) 0.010Age (years) 34 (31–37) 35 (33–38) 36 (33–42) 0.075Ethnicity (%) 0.047White 72.1 58.3 66.7Asian 10.6 21.7 33.3Other 17.3 20.0 0

Family history of DM (%) 48.8 66.7 33.3 0.023Prepregnancy BMI (kg/m2) 23.8 (21.5–27.5) 24.2 (21.4–27.6) 25.4 (22.5–28.1) 0.773Gestational weight gain up to OGTT (kg) 10.3 (7.7–13.6) 10.5 (7.6–14.5) 5.0 (4.5–7.3) 0.045Insulin sensitivity/resistanceMatsuda index 4.7 (3.1–7.1) 3.4 (2.5–5.1) 7.3 (4.9–9.8) 0.001HOMA-IR 1.6 (1.0–2.7) 2.1 (1.4–3.1) 1.2 (0.6–1.9) 0.013

b-Cell functionISSI-2 736 (563–912) 557 (481–707) 418 (348–583) ,0.0001Insulinogenic index/HOMA-IR 10.3 (6.6–16.4) 6.8 (4.2–9.1) 3.3 (3.0–12.9) ,0.0001

Glucose tolerance on OGTT (%) ,0.0001Normal 58.8 33.3 0GIGT 19.6 13.3 16.7GDM 21.6 53.3 83.3

CMPF (ng/mL) 75.1 (57.7–101.5) 70.3 (55.5–99.6) 69.6 (53.2–98.0) 0.788HPL (mg/L) 2.0 (1.5–3.7) 2.0 (1.5–3.1) 1.5 (1.2–2.6) 0.312Prolactin (ng/mL) 93.4 (72.9–121.9) 82.7 (60.4–97.5) 79.2 (52.2–100.4) 0.004

At 3 months postpartumTime since delivery (months) 3 (3–4) 3 (3–3) 3.5 (3–4) 0.140BMI (kg/m2) 25.7 (22.9–29.1) 26.6 (24.4–30.9) 24.4 (22.5–27.4) 0.368Waist circumference (cm) 88 (82–96) 90 (83–96) 79 (75–82) 0.106Breast-feeding (%) 82.7 71.7 83.3 0.136Insulin sensitivity/resistanceMatsuda index 11.1 (7.2–15.8) 8.4 (5.4–10.7) 8.6 (7.0–11.6) ,0.0001HOMA-IR 0.8 (0.6–1.3) 0.9 (0.6–1.5) 1.0 (0.7–1.4) 0.402

b-Cell functionISSI-2 805 (620–1025) 601 (505–666) 303 (265–355) ,0.0001Insulinogenic index/HOMA-IR 10.7 (6.9–17.4) 7.5 (5.4–10.0) 3.8 (3.1–5.1) ,0.0001

OGTT resultsFasting glucose (mmol/L) 4.5 (4.3–4.8) 4.6 (4.4–5.0) 4.6 (4.5–5.5) 0.135AUCglucose 12.7 (11.1–14.2) 16.4 (15.1–17.5) 20.8 (19.2–21.4) ,0.0001

Continuous variables are presented as median (interquartile range). Categorical variables are presented as proportions. DM, diabetes.

1252 Prolactin and Postpartum Diabetes Risk Diabetes Care Volume 39, July 2016

months postpartum: Matsuda index,HOMA-IR, ISSI-2, insulinogenic index/HOMA-IR, fasting glucose, and AUCglucose.The models in Table 2 were adjusted forcovariates from pregnancy as follows:model I shows associations adjusted forweeks’ gestation at the antepartumOGTT, age, ethnicity, family history of di-abetes, prepregnancy BMI, and ges-tational weight gain up to the OGTT;model II includes the covariates frommodel I in addition to adjustment forCMPF, HPL, and prolactin; and model IIIincludes the covariates from model II inaddition to adjustment for GDM. Themodels in Table 3 were adjusted for co-variates from postpartum as follows:model I shows associations adjusted forage, ethnicity, family history of diabetes,postpartum BMI, breast-feeding, andtime since delivery; model II includes ad-ditional adjustment for CMPF, HPL, andprolactin in pregnancy; model III includesfurther additional adjustment for GDM.Logistic regression analysis of (depen-dent variable) prediabetes/diabetes at 3months postpartum was performed withthe following covariates from pregnancy:

weeks’ gestation at the OGTT, age, eth-nicity, family history of diabetes, prepreg-nancy BMI, gestational weight gain up tothe OGTT, GDM, and serum CMPF, HPL,and prolactin in pregnancy (Fig. 1A). Ananalogous logistic regression analysis wasperformed with covariates from postpar-tum as follows: time since delivery, age,ethnicity, family history of diabetes, cur-rent BMI, breast-feeding, GDM in the re-cent pregnancy, and serum CMPF, HPL,and prolactin in pregnancy (Fig. 1B). Thesame approach was applied to logistic re-gression analyses of dependent variableprediabetes (Fig. 2).

RESULTS

Table 1 shows the demographic, clinical,and metabolic characteristics of thestudy population stratified into threegroups based on glucose tolerance sta-tus at 3 months postpartum: 1) womenwith NGT (n = 301), and 2) women withprediabetes (n = 60, consisting of 57with IGT, 2 with IFG, and 1 with IFG +IGT), and 3) women with diabetes (n =6). At the OGTT in pregnancy, among thewomen who went on to have postpartum

prediabetes and diabetes, there waspoorerb-cell function (ISSI-2 and insulino-genic index/HOMA-IR: both P , 0.0001)and a higher prevalence of GDM (P ,0.0001) than among those who had post-partum NGT. Of note, serum prolactin inpregnancy was significantly higher in thewomen with postpartum NGT comparedwith those with postpartum prediabetesand diabetes (median 93.4 vs. 82.7 vs.79.2 ng/mL, respectively; P = 0.004),whereas the antepartum concentrationsof CMPF and HPL did not differ betweenthe three groups (P = 0.79 and P = 0.31,respectively). As shown in SupplementaryFig. 1, mean serum prolactin in pregnancywas higher in women with postpartumNGT compared with the women compris-ing each strata of postpartum prediabetes(IFG, IGT, and IFG + IGT) and those withdiabetes. Accordingly, serum prolactin inpregnancy was significantly higher in thewomen with postpartum NGT comparedwith those with postpartum prediabetes/diabetes (mean 98.2 vs. 80.2 ng/mL, P =0.0003).

At 3 months postpartum, the NGT,prediabetes, and diabetes groups did

Table 2—Adjusted associations of CMPF, HPL, and prolactin in pregnancy with outcomes (dependent variables) at 3 monthspostpartum

CMPF HPL Prolactin

Coefficient t P Coefficient t P Coefficient t P

Log Matsuda indexModel I 20.0001 20.47 0.638 0.1771 1.18 0.239 0.0005 0.70 0.486Model II 20.0001 20.48 0.629 0.1771 1.13 0.259 0.0004 0.56 0.579Model III 20.0001 20.48 0.635 0.0151 1 0.320 0.0004 0.55 0.582

Log HOMA-IRModel I 0.0004 1.19 0.235 20.0285 21.91 0.058 20.0006 20.77 0.443Model II 0.0004 1.23 0.218 20.0281 21.86 0.064 20.0004 20.51 0.609Model III 0.0004 1.23 0.218 20.0284 21.87 0.062 20.0004 20.51 0.609

Log ISSI-2Model I 0.0001 0.47 0.642 0.0067 0.60 0.551 0.0014 2.55 0.011Model II 0.0002 0.71 0.475 0.0027 0.24 0.809 0.0014 2.57 0.011Model III 0.0002 0.75 0.454 20.0005 20.04 0.965 0.0014 2.61 0.010

Log IGI/HOMA-IRModel I 20.0001 20.27 0.784 20.0023 20.09 0.930 0.0027 2.04 0.042Model II 20.00003 20.07 0.947 20.0089 20.34 0.734 0.0027 2.05 0.041Model III 20.00003 20.06 0.949 20.0136 20.52 0.601 0.0027 2.07 0.039

Fasting glucoseModel I 0.0003 1.13 0.260 20.0042 20.33 0.744 20.0005 20.76 0.448Model II 0.0003 1.07 0.288 20.0049 20.38 0.706 20.0004 20.63 0.526Model III 0.0003 1.06 0.289 20.003 20.23 0.818 20.0004 20.63 0.530

AUCglucoseModel I 20.0016 21.05 0.294 0.0038 0.05 0.961 20.0082 22.17 0.031Model II 20.0019 21.27 0.204 0.0246 0.32 0.749 20.0087 22.28 0.023Model III 20.002 21.39 0.165 0.0633 0.88 0.381 20.0085 22.41 0.017

Model I: adjusted for age, ethnicity, family history of diabetes, prepregnancy BMI, gestational weight gain, and weeks’ gestation at OGTT. Model II:adjusted for covariates in model I plus CMPF, HPL, and prolactin in pregnancy. Model III: adjusted for covariates in model II plus GDM. Boldfacetype indicates P , 0.05. IGI, insulinogenic index.

care.diabetesjournals.org Retnakaran and Associates 1253

not differ with respect to time since deliv-ery, BMI, waist circumference, or breast-feeding. Compared with those with NGT,the women with prediabetes and diabeteshad lower whole-body insulin sensitivity(Matsuda index: P , 0.0001). As antici-pated, they continued to exhibit poorerb-cell function (ISSI-2 and insulinogenicindex/HOMA-IR: both P , 0.0001) andgreater glycemia (AUCglucose: P, 0.0001)than the women with NGT.

CMPF, HPL, and Prolactin inPregnancy as Predictors ofPostpartum Metabolic FunctionWe next performed a series of multiplelinear regression analyses to determinewhether serum levels of CMPF, HPL, andprolactin in pregnancy could predictpostpartum metabolic function at 3months after delivery. Table 2 showsmultiple linear regression models evalu-ating the independent associations ofeach of these analytes with features ofpostpartum metabolic function (insulinsensitivity/resistance, b-cell function,glycemia) after adjustment for the follow-ing covariates in turn: model I consists of

variables in pregnancy, including weeks’gestation at the antepartum OGTT, age,ethnicity, family history of diabetes, pre-pregnancy BMI, and gestational weightgain up to the OGTT; model II includesadditional adjustment for CMPF, HPL,and prolactin; and model III includes fur-ther additional adjustment for GDM.These analyses revealed that CMPF,HPL, and prolactin were not signifi-cant determinants of postpartum whole-body insulin sensitivity (Matsuda index)or insulin resistance (HOMA-IR). Impor-tantly, however, prolactin in pregnancyconsistently emerged as a significant in-dependent predictor of postpartumb-cell function in all models of ISSI-2 (allP # 0.011) and insulinogenic index/HOMA-IR (all P # 0.042). Furthermore,higher antepartum prolactin predictedlower AUCglucose on the postpartumOGTT in all models (all P # 0.031). CMPFand HPL were not associated with any ofthese outcomes. It thus emerges thatantepartum serum prolactin is an inde-pendent determinant of better b-cellfunction and lower glycemia at 3 monthspostpartum after complete adjustment

for covariates in pregnancy, includingGDM.

We then constructed models to ad-dress whether prolactin in pregnancypredicts these outcomes after adjust-ment for postpartum covariates. Table3 shows multiple linear regression mod-els evaluating the antepartum analytesCMPF, HPL, and prolactin using an ap-proach tomodel construction analogousto that in Table 2 butwith adjustment forpostpartum factors as follows: model Ishows associations adjusted for age,ethnicity, family history of diabetes,postpartum BMI, breast-feeding, andtime since delivery; model II includesadditional adjustment for CMPF, HPL,and prolactin in pregnancy; and modelIII includes further additional adjust-ment for GDM. The findings from theseanalyses (Table 3) were unchangedfrom those in Table 2, with prolactin inpregnancy consistently emerging in allmodels as a significant independentpredictor of b-cell function (ISSI-2, allP = 0.003, and insulinogenic index/HOMA-IR, all P # 0.018) and lowerAUCglucose (all P # 0.006).

Table 3—Adjusted associations of CMPF, HPL, and prolactin in pregnancy with outcomes (dependent variables) at 3 monthspostpartum

CMPF HPL Prolactin

Coefficient t P Coefficient t P Coefficient t P

Matsuda indexModel I 0.00003 0.11 0.913 0.0141 1.01 0.313 0.0006 0.92 0.361Model II 0.00003 0.11 0.910 0.0124 0.87 0.383 0.0006 0.83 0.406Model III 0.00003 0.09 0.927 0.0092 0.65 0.515 0.0005 0.79 0.431

Log HOMA-IRModel I 0.0002 0.70 0.482 20.0248 21.77 0.078 20.0006 20.8 0.422Model II 0.0002 0.76 0.448 20.0237 21.67 0.096 20.0004 20.58 0.559Model III 0.0002 0.76 0.446 20.0232 21.63 0.104 20.0004 20.58 0.564

Log ISSI-2Model I 0.0002 0.75 0.454 0.0105 0.92 0.358 0.0016 2.96 0.003Model II 0.0002 0.99 0.321 0.0053 0.46 0.644 0.0017 2.97 0.003Model III 0.0002 0.98 0.326 0.0016 0.15 0.882 0.0016 2.96 0.003

Log IGI/HOMA-IRModel I 20.0002 20.04 0.972 0.0071 0.27 0.786 0.0031 2.41 0.017Model II 0.0001 0.17 0.864 20.0018 20.07 0.946 0.0032 2.4 0.017Model III 0.0001 0.15 0.884 20.0079 20.31 0.760 0.0031 2.39 0.018

Fasting glucoseModel I 0.0002 0.66 0.508 20.0028 20.22 0.825 20.0006 20.96 0.336Model II 0.0001 0.58 0.564 20.0027 20.22 0.829 20.0005 20.89 0.375Model III 0.0001 0.60 0.547 20.0005 20.04 0.971 20.0005 20.85 0.395

AUCglucoseModel I 20.002 21.26 0.208 20.0269 20.34 0.731 20.0106 22.79 0.006Model II 20.0023 21.51 0.132 0.0039 0.05 0.960 20.0111 22.89 0.004Model III 20.0022 21.55 0.121 0.0488 0.67 0.503 20.0106 22.98 0.003

Model I: adjusted for age, ethnicity, family history of diabetes, BMI at 3 months postpartum, breast-feeding, and time since delivery. Model II:adjusted for covariates in model I plus CMPF, HPL, and prolactin in pregnancy. Model III: adjusted for covariates in model II plus GDM. Boldface typeindicates P , 0.05. IGI, insulinogenic index.

1254 Prolactin and Postpartum Diabetes Risk Diabetes Care Volume 39, July 2016

CMPF, HPL, and Prolactin inPregnancy as Predictors ofPostpartum Prediabetes/DiabetesLogistic regression analyses were thenperformed to determine whether pro-lactin in pregnancy is a predictor ofprediabetes/diabetes at 3 months post-partum (Fig. 1). In a model consistingonly of predictors from pregnancy (Fig.1A), higher antepartum serum prolactinwas a significant independent determi-nant of a lower risk of postpartum pre-diabetes/diabetes (odds ratio [OR] 0.51,95% CI 0.36 to 0.74, P = 0.0003). The onlyother significant predictors were GDM(OR 4.94, 95% CI 2.64 to 9.24, P, 0.0001)and Asian ethnicity (OR 3.31, 95% CI 1.33to 8.20, P = 0.01). The C statistic for thismodel was higher than that of the modelwithout CMPF, HPL, and prolactin in preg-nancy (0.7998 vs. 0.7606, P = 0.024), in-dicating that these antepartum circulatingfactors significantly improved the predic-tion of postpartumprediabetes/diabetes.In addition, the significant predictorswere unchanged with further adjustmentfor GIGT (data not shown).

These findings were unchanged in amodel adjusting for postpartum factors(Fig. 1B), with serum prolactin in preg-nancy again emerging as a significant in-dependent predictor of a lower risk ofprediabetes/diabetes at 3 months post-partum (OR 0.50, 95% CI 0.35 to 0.72, P =0.0002). Again, the C statistic for thismodel was significantly higher thanthat of the model without the antepar-tum serum factors (0.7946 vs. 0.7420,P = 0.0066). In addition, the significantpredictors were again unchanged withfurther adjustment for GIGT (data notshown).

Given the limited number of womenwith postpartum diabetes, the logisticregression models were repeated withdependent variable postpartum predia-betes (Fig. 2). In a model consisting onlyof predictors from pregnancy (Fig. 2A),higher antepartum serum prolactin re-mained a significant independent deter-minant of a lower risk of postpartumprediabetes (OR 0.50, 95% CI 0.35 to0.73, P = 0.0003). The only other signifi-cant predictors were GDM (OR 4.56, 95%

CI 2.40 to 8.69, P , 0.0001) and Asianethnicity (OR 3.09, 95% CI 1.20 to 7.97,P = 0.019). Again, thesefindingswere sim-ilar in a model with adjustment for post-partum factors (Fig. 2B), with serumprolactin in pregnancy remaining a signif-icant independent predictor of a lower riskof prediabetes at 3 months postpartum(OR 0.51, 95% CI 0.35 to 0.73, P = 0.0003).

Serum prolactin in pregnancy ,115ng/mL had 84.9% sensitivity for predict-ing postpartum prediabetes/diabetes butwith 30.8% specificity (SupplementaryTable 1). As would be anticipated andconsistent with the central role of b-celldysfunction in the pathophysiology ofdiabetes, ISSI-2 in pregnancy ,780 hadbetter operating characteristics as a predic-tor of postpartum prediabetes/diabetesand slightly higher area under the receiveroperating characteristic curve (0.70 vs.0.63) (Supplementary Table 1).

CONCLUSIONS

In this study, we demonstrate that serumprolactin in pregnancy is higher inwomenwho subsequently maintain NGT at 3

Figure 1—Logistic regression analyses of (dependent variable) prediabetes/diabetes at 3 months postpartum, using predictors from pregnancy only(A) and additional predictors from postpartum (B). The reference group for ethnicity variables is white ethnicity.

care.diabetesjournals.org Retnakaran and Associates 1255

months postpartum compared with thosewho have postpartum prediabetes/diabetes. On fully adjusted analyses,antepartum prolactin emerges as an in-dependent determinant of postpartumb-cell function. Most importantly,higher serum prolactin in pregnancy in-dependently predicts a lower risk ofprediabetes/diabetes at 3 months afterdelivery. Thus, serum prolactin in preg-nancy can provide insight into postpar-tum diabetes risk, likely through itsimplications for b-cell function.Although best known for its lactogenic

activity, prolactin is also recognized as aprincipal determinant of islet adaptationto pregnancy (8–11). Indeed, a substan-tial body of evidence from preclinicalstudies has shown that both prolactinand HPL bind to the prolactin receptoron b-cells and induce the expansion ofb-cell mass through a series of down-stream intracellular mediators, includingtryptophan hydroxylase 1 (the rate-limitingenzyme in serotonin synthesis), cell-cycle

regulators, survivin, forkhead proteins,and menin (9,23–26). Of note, the peakof b-cell proliferation coincides with in-creased circulating levels of prolactinand HPL, supporting the importance ofthese hormones in this context (9). Fur-thermore, the prolactin receptor towhichthey bind has been shown to be essentialfor the expansion of b-cell mass in preg-nancy (12,13). Beneficial effects of prolac-tin on b-cell physiology may also extendto the nonpregnant state. Specifically, incell culture systems, prolactin has beenshown to increase b-cell proliferation,inhibit key caspases of the extrinsic andintrinsic pathways that lead to islet apo-ptosis, and promote b-cell survival (27–29). Moreover, prolactin supplementa-tion of pretransplant culture media canimprove b-cell survival in human isletpreparations (30). Taken together, thesedata suggest that endogenous prolactinpotentially could hold implications forb-cell function and overall glucose ho-meostasis outside of pregnancy.

To date, however, there has been onlylimited study of this question in humans.In a cross-sectional study of 2,377 individ-uals in China, those with higher circulat-ing prolactin concentrations had betterb-cell function (measured by HOMA)and a lower prevalence of prediabetesand diabetes (31). Similarly, in a study of3,993 German adults, there was an in-verse association between prolactin con-centration and the prevalence of T2D inboth men and women (32). In the settingof this background, the current study ex-tends the literature in three key ways.First, we have evaluated prolactin duringlate pregnancy, representing a strategicpoint in time when it is playing a centralrole in the b-cell response to the stresstest of gestation. Second, we demon-strate that, whenmeasured in the settingof this physiologic challenge, serum pro-lactin is independently associated withsubsequent b-cell function at 3 monthspostpartum (as assessed by both ISSI-2and insulinogenic index/HOMA-IR) after

Figure 2—Logistic regression analyses of (dependent variable) prediabetes at 3 months postpartum, using predictors from pregnancy only (A) andadditional predictors from postpartum (B). The reference group for ethnicity variables is white ethnicity.

1256 Prolactin and Postpartum Diabetes Risk Diabetes Care Volume 39, July 2016

adjustment for diabetes risk factors bothduring and after pregnancy. Third andmost importantly, the current study dem-onstrates a longitudinal independent rela-tionship between higher serum prolactinin pregnancy and subsequent lower risk ofpostpartum prediabetes/diabetes.While these data suggest that bene-

ficial effects on b-cell function are likelyresponsible for the lower diabetes risk,the underlying mechanism remains un-clear. One possibility is that the antepar-tumprolactinmeasurement is providing asurrogate indicator of the degree ofb-cellmass expansion in pregnancy, which inturn may hold ongoing implications forinsulin secretory capacity at 3 monthspostpartum. Althoughb-cellmass returnsto pregravid levels shortly after delivery inrodent models (33), the precise timecourse of the analogous regression in hu-mans remains to be established. Anotherpossibility is that the prolactin response indriving islet adaptation to the stress testof pregnancy may relate to the beneficialb-cell effects that might be anticipatedfrom lactation and its associated prolac-tinemia after delivery. In this regard,however, it should be noted that the in-dependent associations of antepartumprolactin with postpartum b-cell functionand glucose tolerance were all adjustedfor breast-feeding status. Finally, a thirdpossibility is that measurement of pro-lactin in the setting of the stress test ofpregnancy (and its central role therein)may be providing unique insight into un-recognized elements of the underlyingglucoregulatory physiology of themotherthat contribute to her postpartum b-cellfunction and glucose tolerance. Our find-ings suggest that further study is neededto elucidate the glucose homeostatic im-plications of the serum prolactin concen-tration in pregnant women.From a clinical perspective, it is note-

worthy that the inverse relationship be-tween serum prolactin in pregnancy andpostpartum prediabetes/diabetes per-sisted after adjustment for classical di-abetes risk factors (age, ethnicity, familyhistory, weight, and breast-feeding sta-tus) and even GDM, which is a particu-larly powerful predictor of future risk ofT2D (34). Indeed, coupled with the in-creased C statistic from the enhancedmodels, these data suggest that prolac-tin offers insight into postpartum diabe-tes risk above and beyond GDM status.This finding raises the possibility that

prolactin measurement in pregnancypotentially could contribute to postpar-tum risk stratification in women withGDM by identifying those in whom itmay be particularly important to ensurecompletion of the OGTT that is recom-mended in the first 6 months after de-livery (35), owing to their enhanced riskof prediabetes/diabetes. The potentialvalue of this additional risk stratificationis underscored by the low rates of com-pletion of this OGTT in current clinicalpractice, representing a missed oppor-tunity for early diagnosis and interven-tion in this high-risk patient population(36). The mechanism by which prolactinprovides insight into postpartum diabe-tes risk beyond GDM status is also ofinterest. In this regard, one possibilityis that the antiapoptotic effects of pro-lactin (27–29) may play a role in modi-fying future diabetes risk in the settingof GDM, wherein hyperglycemia mayotherwise contribute to b-cell death(37–39) and hence the risk of postpar-tum glucose intolerance. Further mech-anistic study is needed in this regard.

A limitation of this study is that prolac-tin was not also measured at 3 monthspostpartum, such that the impact of cir-culating levels of prolactin isoforms at thetime of the assessment of glucose toler-ance cannot be determined. Similarly,measurement of prolactin prior to preg-nancy would also have been of interest todetermine whether the higher antepar-tum prolactin that was associated withlower risk of postpartum prediabetes/di-abetes existed before pregnancy or if itwas a reflection of greater antepartumelevation. However, pregravid assess-ment was not possible, since participantswere recruited during pregnancy. An-other limitation is that the modest num-ber of women with diabetes at 3 monthspostpartum (n = 6) necessitated that, de-spite being distinct metabolic states, pre-diabetes and diabetes be combined into asingle group in some of the analyses. Fi-nally, although this study shows a robustrelationship between higher serum pro-lactin in pregnancy and lower risk ofprediabetes/diabetes at 3 months afterdelivery, it does not provide insight intodiabetes risk beyond early postpartum.Further study is thus needed to determinewhether antepartum prolactin relates tothe long-term future risk of T2D inwomen.

In summary, serumprolactin inpregnancyis higher in women who subsequently

maintain NGT at 3 months postpartumcompared with those who have postpar-tum prediabetes/diabetes. Indeed, higherprolactin in pregnancy is an independentpredictor of both better b-cell functionand a lower likelihood of prediabetes/diabetes after complete adjustment forT2D risk factors. It thus emerges thatserum prolactin in pregnancy is a pre-viously unrecognized factor that canprovide novel insight into postpartumdiabetes risk in young women.

Funding. This study was supported by an op-erating grant from the Canadian Institutes ofHealth Research (CIHR MOP-84206). R.R. issupported by a Heart and Stroke Foundationof Ontario Mid-Career Investigator Award, andhis research program is supported by an OntarioMinistry of Research and Innovation Early Re-searcher Award. C.K.K. holds a Banting and BestDiabetes Centre New Investigator Award. A.J.H.holds a Tier-II Canada Research Chair in DiabetesEpidemiology. B.Z. holds the Sam and JudyPencer Family Chair in Diabetes Research atMount Sinai Hospital and University of Toronto.Duality of Interest. R.R. is part holder of U.S.Provisional Patent Application No. 61/703,867“CMPF as a biomarker for diabetes and associ-ated methods.” No other potential conflicts ofinterest relevant to this article were reported.Author Contributions. R.R., P.W.C., A.J.H.,M.S., and B.Z. designed and implemented thestudy. R.R., C.Y., and C.K.K. contributed to theanalysis plan and interpretation of data. C.Y.performed the statistical analyses. R.R. wrotethe first draft. All authors critically revised themanuscript for important intellectual contentand approved the final manuscript. R.R. is theguarantor of this work and, as such, had fullaccess to all the data in the study and takesresponsibility for the integrityof thedataand theaccuracy of the data analysis.

References1. Retnakaran R. Glucose tolerance status inpregnancy: a window to the future risk of di-abetes and cardiovascular disease in youngwomen. Curr Diabetes Rev 2009;5:239–2442. Buchanan TA, Xiang AH. Gestational diabetesmellitus. J Clin Invest 2005;115:485–4913. Retnakaran R, Qi Y, Sermer M, Connelly PW,Hanley AJ, Zinman B. Beta-cell function declineswithin the first year postpartum in women withrecent glucose intolerance in pregnancy. Diabe-tes Care 2010;33:1798–18044. Xiang AH, Kjos SL, Takayanagi M, Trigo E,Buchanan TA. Detailed physiological character-ization of the development of type 2 diabetes inHispanic women with prior gestational diabetesmellitus. Diabetes 2010;59:2625–26305. Xiang AH, Takayanagi M, Black MH, et al.Longitudinal changes in insulin sensitivity andbeta cell function between women with andwithout a history of gestational diabetes melli-tus. Diabetologia 2013;56:2753–27606. Kramer CK, Swaminathan B, Hanley AJ, et al.Each degree of glucose intolerance in pregnancy

care.diabetesjournals.org Retnakaran and Associates 1257

predicts distinct trajectories of b-cell function, in-sulin sensitivity, and glycemia in the first 3 yearspostpartum. Diabetes Care 2014;37:3262–32697. Kahn SE, Zraika S, Utzschneider KM, Hull RL.The beta cell lesion in type 2 diabetes: there hasto be a primary functional abnormality. Diabe-tologia 2009;52:1003–10128. Angueira AR, Ludvik AE, Reddy TE, WicksteedB, Lowe WL Jr, Layden BT. New insights intogestational glucose metabolism: lessons learnedfrom 21st century approaches. Diabetes 2015;64:327–3349. Newbern D, Freemark M. Placental hor-mones and the control of maternal metabolismand fetal growth. Curr Opin Endocrinol DiabetesObes 2011;18:409–41610. Ernst S, Demirci C, Valle S, Velazquez-GarciaS, Garcia-Oca~na A. Mechanisms in the adapta-tion of maternal b-cells during pregnancy. Di-abetes Manag (Lond) 2011;1:239–24811. Ramos-Roman MA. Prolactin and lactationas modifiers of diabetes risk in gestational di-abetes. Horm Metab Res 2011;43:593–60012. Huang C, Snider F, Cross JC. Prolactin recep-tor is required for normal glucose homeostasisand modulation of beta-cell mass during preg-nancy. Endocrinology 2009;150:1618–162613. Sorenson RL, Brelje TC. Prolactin receptorsare critical to the adaptation of islets to preg-nancy. Endocrinology 2009;150:1566–156914. Prentice KJ, Luu L, Allister EM, et al. Thefuran fatty acid metabolite CMPF is elevatedin diabetes and induces b cell dysfunction. CellMetab 2014;19:653–66615. Nolan CJ. Lipotoxicity, b cell dysfunction,and gestational diabetes. Cell Metab 2014;19:553–55416. Retnakaran R, Qi Y, SermerM, Connelly PW,Hanley AJ, Zinman B. Glucose intolerance inpregnancy and future risk of pre-diabetes or di-abetes. Diabetes Care 2008;31:2026–203117. National Diabetes Data Group. Classifica-tion and diagnosis of diabetes mellitus andother categories of glucose intolerance. Diabe-tes 1979;28:1039–105718. Goldenberg R, Punthakee Z; Canadian Dia-betes Association Clinical Practice Guidelines

Expert Committee. Definition, classification anddiagnosis of diabetes, prediabetes and metabolicsyndrome. Can J Diabetes 2013;37(Suppl. 1):S8–S1119. Matsuda M, DeFronzo RA. Insulin sensitiv-ity indices obtained from oral glucose tolerancetesting: comparison with the euglycemic insulinclamp. Diabetes Care 1999;22:1462–147020. Matthews DR, Hosker JP, Rudenski AS,Naylor BA, Treacher DF, Turner RC. Homeostasismodel assessment: insulin resistance and beta-cell function from fasting plasma glucose andinsulin concentrations in man. Diabetologia1985;28:412–41921. Retnakaran R, Shen S, Hanley AJ, Vuksan V,Hamilton JK, Zinman B. Hyperbolic relationshipbetween insulin secretion and sensitivity on oralglucose tolerance test. Obesity (Silver Spring)2008;16:1901–190722. Retnakaran R, Qi Y, Goran MI, Hamilton JK.Evaluation of proposed oral disposition indexmeasures in relation to the actual dispositionindex. Diabet Med 2009;26:1198–120323. Xu Y,WangX,Gao L, et al. Prolactin-stimulatedsurvivin induction is required for beta cell massexpansion during pregnancy in mice. Diabetologia2015;58:2064–207324. Kim H, Toyofuku Y, Lynn FC, et al. Serotoninregulates pancreatic beta cell mass during preg-nancy. Nat Med 2010;16:804–80825. Zhang H, Zhang J, Pope CF, et al. Gestationaldiabetes mellitus resulting from impaired beta-cell compensation in the absence of FoxM1, anovel downstream effector of placental lacto-gen. Diabetes 2010;59:143–15226. Karnik SK, Chen H,McLean GW, et al. Menincontrols growth of pancreatic beta-cells in preg-nant mice and promotes gestational diabetesmellitus. Science 2007;318:806–80927. Labriola L, Montor WR, Krogh K, et al. Ben-eficial effects of prolactin and laminin on humanpancreatic islet-cell cultures. Mol Cell Endocri-nol 2007;263:120–13328. Terra LF, Garay-MalpartidaMH,WailemannRA, Sogayar MC, Labriola L. Recombinant hu-man prolactin promotes human beta cell sur-vival via inhibition of extrinsic and intrinsic

apoptosis pathways. Diabetologia 2011;54:1388–139729. Fujinaka Y, Takane K, Yamashita H, VasavadaRC. Lactogens promote beta cell survival throughJAK2/STAT5 activation and Bcl-XL upregulation.J Biol Chem 2007;282:30707–3071730. Yamamoto T, Mita A, Ricordi C, et al. Pro-lactin supplementation to culture medium im-proves beta-cell survival. Transplantation 2010;89:1328–133531. Wang T, Lu J, Xu Y, et al. Circulating pro-lactin associates with diabetes and impairedglucose regulation: a population-based study.Diabetes Care 2013;36:1974–198032. Balbach L, Wallaschofski H, Volzke H, NauckM, Dorr M, Haring R. Serum prolactin concen-trations as risk factor of metabolic syndrome ortype 2 diabetes? BMC Endocr Disord 2013;13:1233. Rieck S, Kaestner KH. Expansion of beta-cellmass in response to pregnancy. Trends Endocri-nol Metab 2010;21:151–15834. Bellamy L, Casas JP, Hingorani AD, WilliamsD. Type 2 diabetes mellitus after gestational di-abetes: a systematic review and meta-analysis.Lancet 2009;373:1773–177935. American Diabetes Association. 12. Man-agement of Diabetes in Pregnancy. DiabetesCare 2016;39(Suppl. 1):S94–S9836. Shah BR, Lipscombe LL, Feig DS, Lowe JM.Missed opportunities for type 2 diabetes testingfollowing gestational diabetes: a population-based cohort study. BJOG 2011;118:1484–149037. Butler AE, Janson J, Bonner-Weir S, Ritzel R,Rizza RA, Butler PC. Beta-cell deficit and in-creased beta-cell apoptosis in humans withtype 2 diabetes. Diabetes 2003;52:102–11038. Federici M, Hribal M, Perego L, et al. Highglucose causes apoptosis in cultured humanpancreatic islets of Langerhans: a potentialrole for regulation of specific Bcl family genestoward an apoptotic cell death program. Diabe-tes 2001;50:1290–130139. Folli F, Okada T, Perego C, et al. Altered in-sulin receptor signalling and b-cell cycle dynam-ics in type 2 diabetes mellitus. PLoS One 2011;6:e28050

1258 Prolactin and Postpartum Diabetes Risk Diabetes Care Volume 39, July 2016