Lack of control by glucose of ultradian insulin
secretory oscillations in impaired glucose
tolerance and in non-insulin-dependent
diabetes mellitus.
N M O'Meara, … , E Van Cauter, K S Polonsky
J Clin Invest. 1993;92(1):262-271. https://doi.org/10.1172/JCI116560.
Normal subjects demonstrate the presence of ultradian oscillations (period 80-150 min) in insulin secretion rate (ISR) tightly coupled to glucose oscillations of similar period. These oscillations appear to be a function of the feedback loop linking glucose and insulin. The present study was undertaken to determine whether the control by glucose of the ultradian oscillations in insulin secretion is altered in impaired glucose tolerance IGT and in non-insulin-dependent diabetes mellitus (NIDDM). Patients with NIDDM (n = 7), IGT (n = 4), and matched nondiabetic controls (n = 5) were studied under three separate protocols that involved administration of glucose at either a constant rate of 6 mg/kg per min for 28 h or in one of two oscillatory patterns at the same overall mean rate. The amplitude of the
oscillations was 33% above and below the mean infusion rate, and their respective periods were 144 min (slow oscillatory infusion) or 96 min (rapid oscillatory infusion). Insulin, C-peptide, and glucose were sampled at 10-min intervals during the last 24 h of each study. ISRs were calculated by deconvolution of C-peptide levels. Analysis of the data showed that (a) the tight temporal coupling between glucose and ISR in the nondiabetic controls was impaired in the IGT and NIDDM groups as demonstrated by pulse analysis, cross-correlation analysis, and spectral analysis; (b) the absolute […]
Research Article
Find the latest version:
Lack of Control
by Glucose
of Ultradian
Insulin
Secretory
Oscillations
in ImpairedGlucose Tolerance
and in Non-insulin-dependent
Diabetes
Mellitus
Niall M.O'Meara, Jeppe Sturis,Eve Van Cauter,andKennethS. Polonsky
TheDepartmentofMedicine, UniversityofChicago, Pritzker School ofMedicine, Chicago, Illinois60637
Introduction
Normal subjects demonstratethepresenceofultradian
oscilla-tions (period 80-150 min) in insulin secretion rate (ISR)
tightlycoupledtoglucose oscillations of similar period.These
oscillationsappear tobeafunctionof the feedback loop linking glucoseandinsulin.Thepresent study was undertaken to deter-mine whether the control by glucoseofthe ultradian oscilla-tions in insulin secretion is altered in impaired glucose
toler-ance IGT and in non-insulin-dependent diabetes mellitus
(NIDDM). Patients withNIDDM(a=7), IGT(a= 4),and
matched nondiabetic controls (a=5) werestudied under three separate protocols that involved administration ofglucose at eitheraconstant rateof 6mg/kgperminfor 28h orinoneof
two oscillatory patterns at the same overall mean rate. The
amplitude of the oscillations was 33% above and below the
meaninfusionrate, andtheir respective periodswere144min (slow oscillatory infusion)or96 min (rapidoscillatory
infu-sion).Insulin,C-peptide,andglucoseweresampledat10-min
intervals during the last 24hof each study. ISRswere calcu-lated bydeconvolutionof C-peptide levels. Analysis of the data
showed that(a)thetight temporal couplingbetweenglucose
andISR in thenondiabeticcontrolswasimpairedin theIGT
and NIDDMgroups asdemonstratedbypulse analysis,
cross-correlationanalysis,andspectral analysis; (b)theabsolute am-plitude of the ISR pulsesprogressivelydeclined with the
transi-tion fromobesitytoIGTtoNIDDM;and(c)theabsolute
am-plitude of the ISR oscillations failedtoincreaseappropriately
with increasing absolute amplitude of glucose oscillations in the
IGT and NIDDMsubjects comparedwith the controlgroup. In
conclusion,thepresentstudydemonstrates thatimportant
dy-namicpropertiesof the feedbacklooplinkinginsulinsecretion
andglucosearedisruptednotonlyinestablishedNIDDM but
also in conditions where glucose tolerance is only minimally
impaired. Furtherstudiesareneededtodetermine howearlyin thecourseof beta-celldysfunctionthis lack of controlby
glu-coseof the ultradianoscillationsininsulin secretionoccursand todefinemoreprecisely ifthis phenomenon playsa
pathoge-netic role in theonsetof
hyperglycemia
ingenetically
suscepti-ble individuals. (J. Clin. Invest. 1993.
92:262-271.)
Keywords:entrainment*feedbackloop*glucose*insulin
secretion-oscillations *impaired glucosetolerance *
non-insulin-depen-dent diabetes mellitus
AddressreprintrequeststoKenneth S.Polonsky, M.D.,Universityof
Chicago, DepartmentofMedicine, MC1027, 5841 SouthMaryland Avenue,Chicago,IL60637.
Receivedfor publication 9September1992andinrevisedform3
February1993.
During thepast10-15 years, many studies havedemonstrated
thatinsulinsecretion isacomplex oscillatoryprocess charac-terized by rapid oscillations with a periodicity of 8-15 min,
whicharesuperimposedonslower (ultradian) oscillationswith
a periodof 1.5-2 h (1-9). Both rapid and slow oscillations
have beenobserved under fasting conditions and during
con-stantglucose infusion.Theultradianoscillations have in
addi-tionbeen demonstrated duringconstant enteral nutrition and
following meals. The rapidpulsespersistin the isolated
per-fusedpancreas (10) and in isolated islets ( 11 )exposed to a constantglucoseconcentration, suggestingthat they occur inde-pendently of oscillations inplasma glucose and that they origi-nateduetointraislet mechanisms.Theultradianoscillationsin insulinsecretionare,by contrast, tightly coupled to glucose (6, 8,9). Studiesin normal weight subjects during constant glu-coseinfusion and following meals have demonstrated a high
degree of concomitance between these secretory oscillations
andoscillationsin glucose of comparable periodicity.In
addi-tion,theperiodicity ofthe insulin secretory oscillationscan be entrainedtotheperiod ofanoscillatoryglucoseinfusion (9). These findingsareentirely compatible withthe view that the
ultradianoscillations in insulin secretionare aproductofthe
insulin/glucose feedback mechanism(12, 13).
Wehavestudiedthetemporal interactionsbetweenglucose
and insulinsecretion by determiningtheextent towhich ultra-dianoscillationsininsulinsecretioncanbeentrained by exoge-nousglucose. Theconceptof entrainmentiscentraltothefield of nonlinear dynamics and has been used inmany different scientificfields, includingthestudy of biologicalsystems( 14). Briefly,ifanonlinear, self-oscillatingsystemis perturbed exoge-nously withaperiodic stimulus, differenttypesofoscillatory behaviors may emerge, one of whichis entrainment. When
entrainmentoccurs,theoscillationof thesystemreacts tothe exogenousstimulus by
adjusting
itsperiod
tothat ofthestimu-lus. The ability to entrain depends on a number of
factors,
includingtheperiod ofthestimuluscomparedwith the natural
periodofthesystem,the
amplitude
ofthestimulus,thenonlin-earities ofthe system, and the
strength
ofthecoupling
between theexogenousstimulus and thesystem.In non-insulin-dependent diabetes mellitus
(NIDDM),'
the
oscillatorypatternof3-cell
secretion isdisrupted.
Theper-sistent, regular rapidoscillationspresentin normal
subjects
arereplaced byirregular cyclesof shorter duration both insubjects
with established NIDDMandintheirfirst-degreerelatives( 15,
16).Theultradianoscillations inNIDDMarealsomore
irregu-lar.Inaddition, theyareof smaller
amplitude
and the propor-tionoftheseoscillationsthatareconcomitant with oscillations1.Abbreviationsused in thispaper:BMI,bodymassindex; IGT,
im-paired glucose tolerance; ISR,insulin secretionrate;NIDDM,
non-in-sulin-dependentdiabetes mellitus.
Abstract
J. Clin. Invest.
©The AmericanSocietyforClinicalInvestigation,Inc. 0021-9738/93/07/262/10 $2.00
in glucose is much smaller ( 17, 18). These latter observations raise the possibility that specific dynamic properties ofthe feed-back loop linking glucose and insulin are disrupted in NIDDM. Such an alteration, if it could be confirmed, might be indicative of an important early (3-cell abnormality in suscepti-ble individualsevenbefore theonsetofoverthyperglycemia. The present study was therefore undertaken to determine ifthe abilitytoentrain 3-cell secretionwasreduced insubjects with established NIDDM and inpatients with early impaired glu-cosetolerance(IGT) who hadnotyet developedovert hypergly-cemia.
Methods
Subjects
Studies wereperformed inpatients withproven NIDDM(n = 7, 6 males, 1 female;2Caucasian,5Black),impairedglucose tolerance (n
=4, 2males, 2 females, 1 Caucasian,3 Black), and inagroupof obese
subjects(n=5,4males,1 female,3Caucasian, 1 Black, 1Hispanic)
who hadnormal glucose tolerance and no family history of diabetes. Thecriteriausedtodeterminewhetherglucose tolerancewasnormal
orimpairedin therespectivegroupswerethoseoftheNational Dia-betes DataGroup (19).Themeanage,weight,andbodymassindex (BMI) of the three groupsaredemonstrated in TableI. Noneofthe diabeticpatientswasreceivinginsulintherapyatthetime ofthestudy. Oralhypoglycemicagentswerediscontinuedatleast 2 wkbefore the
commencementof theexperiments.
Allstudieswereperformedin theClinical Research Center(CRC)
of theUniversityofChicago.Theprotocolwasapproved bythe Institu-tionalReview Board,andallsubjectsgave written informedconsent.
Experimental protocol
Subjects were admittedto the CRC afteran overnightfast. Aftera
subjectwasadmitted,anintravenous catheterwasinserted into both
forearms,onefor glucose infusionandonefor bloodsampling. Begin-ningat0800hours, glucosewasinfusedintravenouslyforaperiod of28 h,duringwhich time thesubjectsremained in the recumbentposition. Eachstudy consisted ofaninitial 4-hperiod (0800-1200 hours)to
allowasteadystatetobeachievedfollowingthe transientglucoserise thatisassociated with the first few hoursofinitiatingaglucoseinfusion
(8).Thiswasfollowed byasubsequentperiod of24 h (1200-1200 hours)duringwhichtimesamplesweredrawnat10-min intervalsfor
glucose,insulin,andC-peptide.Lightsweredimmed between 2300 and 0700hours, toallow thesubjectstosleep. Potassium(40 meq)was givenorally every 12 h, andsubjectswereallowedfreeaccesstowater
butnotgivenanyfood forthedurationof theexperiment.
Eachsubjectwasstudied on at leasttwooccasions under the above protocol. On eachoccasion,glucosewasadministeredas a20% solu-tion viaacomputer-controlled pump(Flo-gard8000volumetric infu-sion pump; Travenol Laboratories, Deerfield,IL)followingdifferent patterns.Initially,eachsubjectreceivedaconstantglucoseinfusion ata rateof 6 mg/kg per min. The subsequent studies involved the adminis-trationofanoscillatoryglucose infusion with periods of 96 min and/or
144min, theseperiods having been shown to be 20% shorter and 20% longerrespectively than the endogenous period observed in normal
subjects duringconstantglucoseinfusion (9). The software controlling
theglucose infusion was custom written to allow the actual infusion
rate tobe recorded so that possible interruptions could be detected.
Eachoscillation was shaped as a sine wave and the amplitude was 33% aboveand below the meaninfusion rate. Subjects received identical volumesof glucose during each 28-h infusion period.
Analytical methods
Blood samples for insulin measurementswere allowed to clot at room
temperatureand theserumwasstoredat-20'C untilassayed. Samples forC-peptideweredrawninto tubesat
40C
containing500 Kallikreininhibitorunits/ml Trasylol and 1.2 mg/ml EDTA. The plasma was immediately separated and stored frozen until assayed. Serum insulin wasassayed by a double antibody technique (20). This assay has a lowerlimitofsensitivity of 20 pmol/liter, and the average intraassay coefficientofvariation is 8%. PlasmaC-peptide immunoreactivitywas
measured aspreviously described (21 ). The lower limit ofsensitivityof the assay is 0.02pmol/ml and theintraassaycoefficient of variation averages 6%. Plasma glucose was measured with a glucose analyzer (model23A; Yellow Springs Instrument Co., Yellow Springs, OH). Theintraassaycoefficient of variationofthis method is<3%.
Data
analysis
Therelative degree ofinsulin resistance was evaluated with the Homeo-stasis Model Assessment (HOMA) method (22),usingthe mean of 3-4 stressedfastingglucose and insulin levels obtained ondifferent occa-sionsfrom each subject.
SMOOTHING AND ESTIMATION OF INSULINSECRETION RATES
Theindividual glucose,insulin,andC-peptide profilesweresmoothed
usingathree-point movingaverageasinpreviousstudies ofoscillatory
insulin secretion ( 1, 6, 9). This procedure strongly dampens all
fluctua-tions shorter than 30 min, allowingabetter visualizationofslower
oscillationsat anexpenseofamodest reduction in theiramplitude.It also reduces measurement errorbyafactor of A. All further calcula-tionswereperformedonthe smoothedprofiles.Thesmoothed
C-pep-tide curve was used to derive insulin secretion rates by deconvolution (23). Thekinetic parameters used in the deconvolution program were derived aspreviouslyvalidated and described in detail (24).
PULSE ANALYSIS
Toidentify significant pulses, eachprofilewasanalyzedwith ULTRA,
acomputer program forpulsedetectionandquantification (25).The generalprincipleof thisalgorithmis the eliminationofallpeaksfor
which either the increment (difference between the peak and the
pre-ceding trough)orthe decrement(differencebetween thepeakand the nexttrough) does not exceed acertainthreshold related to measure-ment error.Thisproduces a so-called "clean"profile.Extensive simula-tion studies(25)haveindicated thatathreshold of twice theintraassay
coefficient ofvariation generally minimizes both false-positive and
false-negativeerrors.This thresholdwasthereforeusedtoidentify
sig-nificant pulsesofglucose. However, becausedeconvolution involves
anamplification ofmeasurementerror,amoreconservative threshold of three times theintraassay coefficient ofvariation ofC-peptidewas
usedtoquantify pulsesof insulin secretion. Inthis study, pulseanalysis wasperformedon theprofilessmoothed by thethree-point moving
averageso that the measurement errors werereduced by A. Thus
peaks of insulin secretionandglucosewereconsideredsignificantif
theirrespectiveincrements and decrements exceeded 10.39%(i.e.,3
X6%/ V3) and 3.46% (2 x 3%/
f/),
respectively. Peaksthatdid not meet these threshold criteriawereeliminated fromthe datasetusingan iterativeprocess. For eachsignificantpulse, the absoluteamplitudewasdefinedasthe difference between the level of thepeak and the level of thepreceding trough. Groupstatisticsonpulse amplitudeswerebased
onmedians,ratherthan means, becauseofthe non-Gaussiannatureof
pulse distribution.
Becauseglucose hasarelativelylonghalf-life,changes in
produc-tion and/orutilization maynotbereflectedassignificant peaksin the glucose concentration curve. We therefore identified all the
"shoulders"in the glucoseprofiles.Todothis, instantaneous deriva-tives of eachindividual clean glucoseprofilewereestimatedastheslope oftheglucosechanges during each sampling interval, and the shoulders were identified as slopes with an absolute value of zero (i.e.,
corre-spondingtotheeliminated peaks), preceded and followed by slopes of thesamesign.
ANALYSIS OF THE TEMPORAL ASSOCIATION BETWEEN
OSCILLATIONS OF GLUCOSEAND OSCILLATIONS INISR
Pulse-by-pulseanalysis of concomitance. Temporal associations
analysis oftheconcomitance. Significantpulsesofglucose and insulin secretion were consideredconcomitantiftheirpeakvalues occurred within 10 min of each other. The concomitance ratio of glucose pulses with ISR pulses was calculated as the number of concomitant glucose and ISRpulses divided bythe total numberof glucosepulses.
Con-versely, theconcomitanceratio ofISR pulseswithglucose pulses was calculated as the numberofconcomitant glucose and ISR pulses di-videdby the total numberofISR pulses.
Cross-correlation analysis. For each glucose, insulin, and ISR pro-fileabest-fitcurve wascalculatedusingarobustalgorithmproposed by Cleveland(26)with awindowof 15 datapoints.Eachprofilewasthen
divided by itscorresponding best-fit curve, thereby yielding a de-trendedprofile for whichdiurnalvariationswereremoved. For each
study, coefficients ofoverallcrogs-correlationwerethen calculated be-tweenpairs ofdetrendedprofiles(glucose vs.insulin,glucose vs. ISR, andinsulinvs. ISR) atlags of0, ±10, ±20, etc., up to±300 min.
Cross-correlation profiles from different studies were pooled using Fisher'szvalues(27). This procedureallowed themaximalcoefficient of cross-correlation foreachglucoseinfusion protocolandstudygroup tobeidentified along withthelagatwhich this occurred,thus provid-ingmeasuresofthe overallconcomitanceaswell as thecorresponding
lag between the ultradianoscillationsin glucose,insulin,and ISR.
Spectral analysis. To investigate the temporal nature ofthe profiles
obtainedduring glucoseinfusion, spectral analysiswasperformedon eachglucose,insulin, and ISR series. The methods aredescribedby
Jenkinsand Watts(28).To removecircadiantrends,eachserieswas detrendedwiththefirst differencefilter,and then spectralestimations
wereobtained usingaTukeywindowwith a widthof72 datapoints.
Eachindividualspectrum wasnormalizedassumingthe totalvariance of theseriestobe 100%. Thisallowed average values across the three groups to bestatistically compared.
STATISTICALANALYSIS
Results areexpressedasmean±SEM. Statistical tests wereperformed usingtheStatistical Analysis System (version6.04for personal com-puters; SASInstitute Inc., Cary, NC). Repeatedmeasuresanalysisof
variance was used toidentify overall differences between the three
studygroups and thestudiesusingthe threedifferent glucose infusion
protocols.Tukey's studentizedrange testwasusedfor post-hoc
compar-isons. Differenceswereconsideredtobesignificantif P<0.05. Results
Basallevels ofplasma glucose, serum insulin, andISR. The
clinical characteristicsand basallaboratoryvaluesof the study
groups are recorded in Table I. All three groups were well
matched for age,weight,and BMI.Fastingglucose and
glyco-hemoglobinconcentrationsweresignificantly higherin the dia-betic patients, whilethese valueswerenormal in the patients
TableI. Ages, Weight,BMI, BasalInsulin, C-Peptide, Glucose, andHbAI inthe ThreeStudy Groups
Control IGT NIDDM
(n= ()(n=4) (n=7)
Age(yr) 40.4±4.6 44.5±7.0 46.0±4.7
Wt(kg) 122±5.9 110±11 106±10
BMI
(kg/M2)
38.4±1.3 38.8±4.4 34.9±2.5 Insulin(pmol/liter) 172±16.6 128±4 132±2.8C-peptide(nmol/liter) 1.06±0.13 0.84±0.08 0.87±0.13 Glucose(mmol/liter) 5.54±0.21 6.11±0.17 10.3±0.98*
Glycosylatedhemoglobin(%) 5.81±0.18 6.05±0.53
10.3±1.11
*P <0.00 1; t P<
0.01
vs.controlsubjects.with IGT. Although basal glucose levels were slightly higher in
thosewith IGT compared with the nondiabeticcontrol
sub-jects, the differences were notsignificant. Thedifferences in the
fastinginsulin andC-peptideconcentrations were not
signifi-cant at the 5% level.Use oftheHomeostasis Model Assessment
(HOMA) method revealed nosignificant differences in degree
of insulin resistance between the three study groups (control,
8.0+1.4;
IGT,
5.2+0.7;
NIDDM,7.9±1.4;P >0.43).Mean 24-h levels of plasma glucose, plasma C-peptide,
seruminsulin, andISR. Overall mean concentrations of
glu-cose,insulin, C-peptide,and the insulin secretory rates over the 24-h periodduring the constant glucoseinfusion studies are presented inTable II. In addition, the mean glucose levels and insulin secretory rates are shown forthe oscillatory studies. Insulin,C-peptidelevels, andinsulin secretion rates were
signif-icantly lower in the NIDDMpatientsthan in thenondiabetic
controls,and theglucose concentrationsintheNIDDM group
were significantly higher than in both the control and IGT
groups. The differences in glucose and insulin secretion rates between thepatientswith IGT and controls were not
signifi-cant. Theglucoselevels and insulinsecretionratesdidnot
de-pendonthe modeof glucose infusion. Fig. 1 depictsthemean
insulinsecretory rates andglucose levelsduringallstudies.The meanglucose levels were elevated andinsulin secretion rates
werereduced in thesubjects withNIDDMcompared to
con-trols, withvirtually no overlap between the two groups. Pa-tients with IGT tendedto havelowinsulin secretionrates in
relationtotheprevailing glucose concentration, although the
differences in secretion rate did not reach statistical
signifi-cance.
Number andamplitude ofoscillations. Representative pro-files ofglucose, insulin and ISRfromeachof the three study groupsduringthe constantandoscillatory infusionsare demon-strated inFigs.2-4. The number and absoluteamplitude ofthe
respectiveoscillations in eachexperimentalprotocol are shown in TableIIIandFig.5.The numberofglucose pulseswas
simi-TableII. Mean Values (±SEM)of Insulin, C-Peptide, Glucose, and InsulinSecretionduring Constant Glucose Infusion
andofGlucose and InsulinSecretionduring Oscillatory GlucoseInfusion
Control IGT NIDDM
Constantglucose (n=5) (n=4) (n=7)
Insulin(pmol/liter) 1174±145 591±101 397±99*
C-peptide(nmol/liter) 3.2±0.35 2.4±0.23 1.73±0.17*
Glucose(mmol/liter) 9.2±0.47 10.7±0.91 20.4±1.83t
Insulinsecretion
(pmol/m2permin) 426±48 310±26 229±23*
Oscillatoryglucose-144min (n=5) (n=3) (n=6)
Glucose(mmol/liter) 9.0±0.33 11.4±0.74 20.4±2.01 Insulin secretion
(pmol/m2permin) 416±31 345±19 242±26
Oscillatoryglucose-96min (n=5) (n=4) (n=5)
Glucose(mmol/liter) 9.4±0.52 10.3±0.54 19.1±1.91 Insulin secretion
(pmol/m2permin) 416±37 301±44 243±30
U
a ** *
* * * * *
5 10 15 20
MEAN GLUCOSE LEVEL
25
30i
(mmol/L)
Figure1. Relationship betweenmeaninsulin
secretory
rates normal-izedby bodysurfaceareaandmeanglucose
concentrationsduring
allglucoseinfusion studies. Eachmarking (control, +;IGT,.m,
NIDDM,*)
corre-spondstotheresultsof onestudy.
lar in the threegroupsexcept fora
significant
reductionduring
constant
glucose
infusion in thesubjects
with NTDDM(P
<
0.02).
The numberofpulses
ofTSR didnotdiffer inanyof thethreegroupsinanyof the threeprotocols.Two-way
analy-sis of variance for repeatedmeasures showed that the
ampli-tudeof
glucose
oscillationsincreasedduring
theoscillatory
in-fusion
studies,
thechanges
being
most marked with a slow144-mmn
oscillatory period
(P<0.0001I).
Therewere nosignifi-cant differences in the
amplitude
of theglucose
oscillations between the threegroups. However, differences in theampli-tude of the TSR oscillations were
highly significant
both be-tweenthegroups(P<0.001I)
and betweentheglucose
infusionprotocols (P<
0.01I),
and thegroupdifferenceswererelatedto thedifferences in the infusionprotocols (P
<0.002).
Tn thecontrols, theincrease in
amplitude
of theglucose
oscillationsbrought
aboutby
theoscillatory glucose
infusionprotocols
wasthus
accompanied by
asignificant
increase in theamplitude
of the oscillations inTSR,
whereas in both the diabetic andTGTgroups,the
amplitude
ofTSRoscillations didnotincreasesignif-icantly
with the increase inamplitude
of theglucose
oscilla-tions(Fig. 5).
Temporal
association between oscillations inglucose
and ISR. Thetemporal association between oscillations inglucose
andTSRwasevaluated
using
threeseparateapproaches:
pulse-4~-4,
by-pulse
analysis
ofconcomitance,
cross-correlationanalysis,
and
spectral
analysis.
Eachapproach
totheanalysis
of the dataindicated a reduction in the normal
tight temporal coupling
between oscillations in glucose andTSR in the NIDDM and IGT
subjects although
thesignificance
of the differences varieddepending
onthetechnique.
Pulse-by-pulse analysis
of
concomitance. The concomi-tancebetween individual pulses ofglucose
and TSR for each groupandprotocol
areshown inTable ITT. When the concomi-tance ofpulses
in TSRwithglucose pulses
wasexamined withrepeated
measures two-wayanalysis
ofvariance,
therewas asignificant
difference between groups(NIDDM
significantly
different from TGT andcontrol; P<
0.002)
and betweenglu-coseinfusion
protocols
(P<0.0002).
Similarly,
thepercentageof
glucose pulses
thatwasconcomitant withanTSRpulse
wassignificantly
different between groups (NIDDMsignificantly
different from control; P <
0.03),
but not betweenglucose
infusion
protocols.
Inorder to determine if these differenceswereconfirmed with other
analytical approaches,
cross-corre-lation andspectral
analyses
wereapplied.
Cross-correlation
analysis. Fig.
6 illustrates the coefficient of cross-correlation betweenglucose
andTSRatdifferentlags
for the threegroupsand the three
glucose
infusionprotocols.
Repeated
measurestwo-wayanalysis
of variance showed that the maximalcoefficient of cross-correlation betweenglucose
and TSR was different between the three groups
(TGT
andNTDDM different fromcontrol;P<
0.0001I)
and between the threeglucose
infusionprotocols
(P <0.002).
Similarly,
the maximalcoefficients of cross-correlation betweenglucose
andinsulin weredifferent between thegroups
(TGT
andNTDDMdifferent fromcontrol; P<
0.001I)
and between the threeglu-coseinfusion
protocols
(P<0.002).
Theresults forcorrelationbetween TSR and insulin were
only
significant
between thegroups
(NTDDM
differentfromcontrol;P<0.04).
Tn ordertoexamine which of the three
glucose
infusionprotocols
mostclearly brought
outgroupdifferences,
one-wayanalysis
ofvari-ance was
performed
ondatafrom eachglucose
infusionstudy
separately.
The 144-mmnprotocol
consistently
revealed the mostsignificant
differences. In addition to reduced overall10 9
8-7 6
1600
14001-8200
200 W-\ \64600
00 ~4001 200~
1500
-1750-1 250
1000 1250
2500 2501J
0 240 480 720 960 1200 1440
TIME (min) 0 240 480TIME720(min)960 1200 1440
Figure2.Profiles of theglucoseinfusion rate, theglucoseandinsulinconcentrations,and theinsulinsecretoryratesinonerepresentativecontrol
subject.All three studiesperformedinthissubjectareshown:constantglucose infusion, left;slowoscillatory infusion, middle; rapid oscillatory infusion, right. Significant peaksareshownas arrowsin theglucoseand ISRprofiles,andshoulder(s)areshown in theglucose profileobtained
duringconstantglucoseinfusion.Afewinterruptionsoccurredthatwereallof short duration(<2min),andtheydidnothave any visible effectupontheglucoseprofiles.
700
-E-
600-NN
E 500
-
NI-0
E 40
(i) 300
-j~
200-+
+ .- a*
ui0200-zE
0-0-0
0 8
110
25001-a10001
500~
11
9
7]fJJVk\AJ
0 240 480 720 960 1200 1440
TIME (min)
400
- 200-0
300-°, o 200-1
2 \ 100]
zE _ E
o-151 W
-14]
8 ° 13]
, E 12
J
E
10
16-14 -12
-101
400 200
0
141
12
-10
-8-i
~~~~~~~~~~~~12001-.E10001 1000
Z
E ,51000
E
0.1
600 600J600
1250 1500
1500-I1000-10
01000
500 ... 500 ..,,.,,,, 500
0 240 480 720 960 1200 1440
TIME (min) 0 240 480TIME720(min)960 1200 1440
0 240 480 720 960 1200 1440 TIME (min)
Figure 3.Profiles ofthe glucoseinfusionrate, theglucose andinsulin concentrations,and theinsulinsecretoryratesin one representative IGT
patient.Allthreestudiesperformedinthispatientareshown:constantglucoseinfusion, left;slowoscillatory infusion,middle; rapid oscillatory
infusion,right.Afewinterruptionsoccurred that were allofshort duration (< 2 min), and they did not have any visible effect upon the glucose
profiles.
cross-correlation, theNIDDMgroup also demonstrated a
con-siderably increased lagat which the maximal coefficient was observed (see Fig. 6). Inotherwords, whiletheoverall
coeffi-cient ofcross-correlation between glucose and ISR was greatest
at alag ofzero min in the control and IGT groups for all
glucose infusion protocols, there was a lag in the NIDDM group which ranged from 15 to 35 min (peaks ofglucose occur-ring before peaksofISR) depending on the glucose infusion
protocol.Similarresults wereobtainedfor the cross-correlation betweenglucoseandinsulin.
Spectral analysis. Spectral analysisof theoscillatory
pro-files (Fig. 7)confirmedthe existenceofpeaks in the plasma glucosespectra in the three study groups at 96 and 144min, correspondingtotherespectiveperiods of the exogenous
infu-sion. Similar spectral peaksin insulin (not shown) and ISR
wereobserved in thenondiabeticcontrolsubjects.However the
Z 400 400
200. 200 \
0
0E
20 -W
191
8 -
18-, E
17-_E 16]
15
231 21 ]
19] 17]
15
A
\AA/V\/\
power of the insulin and ISR spectralpeaks was reduced in
boththe IGT andNIDDMgroups, this reductionbeingmost
marked in theNIDDMgroup. Duringthe96-minoscillatory
infusion the normalized powerofthe ISR spectrumfell from
29.1 in the controls to 13.6 in the subjects with IGT (P
<0.001 ) and 5.5 (P<0.001)in the subjectswith NIDDM.
Correspondingly,the ISR spectra basedonthe144-min oscilla-tory infusionsyielded respectivepowers of26.2, 5.9,and 3.0 in eachofthethree groups(P<0.001). Similarresultswere ob-tained for the insulin spectra.Noclear group differenceswere detected between the spectra from the constantglucose infu-sion protocol. Examinationof individualspectra also showed
that in a large
proportion
of the IGT (three of four) andNIDDM (five of seven) patients,a discrepancy betweenthe dominantperiodof ISR andinsulincould be detected.In con-trast, noneofthe five nondiabetic controlsubjectsexhibited
400
/VV\/V\ 200
0
28-26 -24 -22
20-18
1200
1o00o 1 1100
E 1000
800 , 800 700
1000 1000
1500 ..5 0
0 240 480 720 960 120014 0 240 480 720 960 1200 1440 TIME (min) TIME(min)
0 240480 720 9601200l1440
TIME (min)
Figure 4. Profiles of the glucose infusion rate, theglucoseand insulinconcentrations, and the insulinsecretoryratesinonerepresentative
NIDDMpatient. All threestudiesperformedinthispatientareshown:constantglucoseinfusion, left;slowoscillatory infusion, middle; rapid
oscillatory infusion, right.Afewinterruptionsoccurred thatwereallof short duration(<2min),andtheydidnothave anyvisibleeffectupon
Table III. Pulse Analysis ofthe Glucose and ISR Oscillations
Glucose ISR
% ofglucose % of ISR No.of Absolute pulses with No. of No. of Absolute pulses with Study Group pulses increment ISRpulses shoulders pulses increment glucosepulses
Constant glucose infusion Control(n=5) 9.8±0.6 1.10±0.12 61.2±7.9 3.8±0.4 14.6±1.0 334±52 40.4±3.8 IGT (n=4) 8.3±0.9 1.18±0.30 61.3±9.5 3.5±1.5 13.5±1.0 260±68 39.1±9.9 NIDDM(n =7) 6.7±0.6 1.14±0.11 54.4±9.5 6.0±0.7 12.9±1.2 158±17 28.1±6.1
Oscillatory glucoseinfusion Control(n= 5) 10.2±0.2 3.43±0.26 70.5±4.5 - 11.0±0.5 566±91 66.4±2.8
(144min) IGT (n= 3) 10.0±0.0 3.04±0.23 63.3±16.7 14.3±0.9 248±30 46.0±13.2
NIDDM(n =6) 9.7±0.2 2.87±0.33 40.9±7.2 13.8±0.7 145±13 28.5±5.0
Oscillatoryglucoseinfusion Control(n= 5) 14.2±0.2 2.47±0.13 80.4±4.0 - 14.2±0.4 476±47 80.3±2.7
(96min) IGT (n=4) 14.3±0.3 2.07±0.20 61.3±9.5 13.0±1.2 292±76 66.1±4.5
NIDDM(n=5) 13.4±0.2 1.57±0.05 45.8±10.2 14.2±1.0 159±17 43.6±10.2
Results ofstatisticalanalyses are in the text.
anydiscrepancy betweeninsulinand ISR.Intheseindividuals,
the largest spectral peak was always observed at the
corre-sponding periodof the exogenousglucoseinfusion. Examples
of individual spectra illustrating these results are shown in Fig. 8.
changes to assume the period of exogenous pulsesofglucose duringexogenousoscillatory glucoseinfusions.Thus these
ul-tradian oscillations in insulin secretion were completely
en-trainable by exogenous glucose, implying that the inherent
Discussion
We havepreviously demonstrated thatin normal volunteers (9), ultradian oscillations of insulin secretionarepresent dur-ing constant intravenous glucose infusion and their period
CONTROL
E 0E
-~ ~ ~ ~ ~ ~ 5
IGT
N
C
.E
._
soo 500
200
100
CL
Figure 5. Relationships between * absoluteamplitudesin the three
different groups and the three
dif-ferent protocols.C,R,and S de-noteconstant,rapid oscillatory,
andslowoscillatory glucose
infu-sion, respectively. Inall groups, N administration ofoscillatory
glu-E cose infusion leads to an amplifi-* cation of the glucose oscillations *, (lefithandbars),butonly in the o controlgroup isthis associated with asignificant increase in the
0: absoluteamplitude of the ISR
oscillations(righthand bars).
CONSTANT INFUSION GLUCOSE LEADING ISR
-CONTROL
---IGT
---NIDDM
RAPID OSCILLATORY INFUSION GLUCOSE LEADING ISR
CONTROL
---IGT ---NIDDM
)-.... .. .
-300 -200 -100 0 100 200 300 LAG (min)
SLOW OSCILLATORY INFUSION GLUCOSE LEADING ISR
.31.000
cr
0
Ir
C.) 0.00
U- 0.50 S2
7-o
-CONTROL ---IGT
----NIDDM
I-~~ ~ ~ ~ ~~ ~
-300-200 -100 0 1600 200 300 LAG (min)
Figure 6. Coefficient of
cross-correlation be-tweenglucoseand ISR inthe threestudy
groupsduringconstant
(top), rapid oscillatory (center),and slow
oscillatory (bottom) glucoseinfusion.
-E
E
0 n
nx E
E
E
E
0
0.
NIDDM
4- A
1- ~~~~~~~~~~~~-26
0
.p
CONSTANT INFUSION 96-MIN INFUSION 144-MIN INFUSION -J 0 F-z 0 0)
----GLUCOSE °70 GLUCOSE ISR E 60- ISR
IE4 jX40 030 20. 20
---144192 240 288 0 48 96144192 240 28 ID(min) PERIOD(min)
Figure 7. Mean glucose(-- -)and ISR ( )spectraof the control subjects (top),IGTpatients (middle) and NIDDM patients (bot-tom). Error bars have been omitted for clarity. Left, middle, and right columnsshow the results from theconstant,rapid, and slow
oscilla-toryinfusionstudies, respectively.
properties of the insulin/glucose feedback loop playan
impor-tantrole inregulating and generating the oscillations. Thepresentstudywasdesignedtodetermine if specific dy-namic abnormalitiesare presentinthe feedback loop linking insulinsecretion and glucose in NIDDM duringconstantand oscillatory intravenous glucose infusion. Studies were per-formed in patients with overtNIDDM and matchedgroups
with impaired and normal glucose tolerance who received intra-venousglucose infusions eitherataconstantoroscillatoryrate
withaperiodicity of 96or 144min.In all threegroups, oscilla-tionsinglucose and ISRwereapparentduringconstantglucose infusion. In the groupwith normalglucose tolerance atight temporal couplingwasevident between the oscillations ofglu-coseand ISR. Incontrast,inpatients with IGT andNIDDM, therewasevidence ofuncoupling of this relationship between glucose and ISR andmany glucose independent oscillations wereapparent.Alterations in the ultradian oscillations in ISR have notpreviously been characterizedinsubjects with IGT. The abnormalrelationship between oscillations ofglucose and ISR in thisparticulargroup,which issimilartothat observed in theNIDDM group, suggests thatan uncouplingof the
tem-poral association between ultradian oscillationsinglucoseand
ISR isanearly manifestation of ,-cell dysfunctioninNIDDM. In thesestudies, evidence for lack of control by glucose of the ultradianoscillations ofISRwasdemonstratedusingthree different analytical approaches: (a) by analysis of concomi-tance of theindividual pulses ofglucose and ISR using the
programULTRA; (b) by cross-correlation analysis; and (c) by spectral analysis. Analysis of theconcomitance of the
individ-ualpulses demonstratedthateventhoughthe total numberof
70- 60-E
°
50-X
40-0 3c 30.
0~ 20-M 10
X)
C] z ID 70-.N E 60-° 50-X 40-LhJ 30 20- 10-LUJ en 0 I' I I I I II I I I I48 96 144 19 PERIOD (m
Il
I I
48 96 144 19
PERIOD (mi
Figure 8. Individual
powerspectraof the datafromone
nondia-beticcontrolsubject (top),oneIGT patient
(center), andone 2 240 288 NIDDM patient
(bot-in) tom). Theuppertwo
panels correspondtothe
data showncenterof
---GLUCOSE Figs. 2 and3, while the
--INSULN lower
panelisfrom a
ISR
diabetic subject
differ-entfrom theone
dis-played inFig. 4.Inthe controlsubject, glucose (-- ),insulin(---),
andISR ( ) all
ex-hibitadominant peak
2 240 288 at 144min(theperiod
in) of theglucoseinfusion).
Inthe IGT andNIDDM
patients, although
---GLUCOSE plasma glucose isclearly ---INSUUN entrainedat144min
ISR
and 96min,
respec-tively, this is neither the
casewith insulinnor
with ISR. Furthermore, thedominant period of insulin is different from the dominantperiod of ISRinboth theIGT
i2 240 288
in) and NIDDMpatients.
glucose and ISR pulses in the 24-h profileswassimilar between thegroups, asignificant reduction in the proportion ofpulses inISR that wasconcomitant withpulses in glucose and vice versa wasobserved in NIDDM. Cross-correlation analysis re-vealed a reduction in the overall concomitance between the ultradian oscillations ofglucose, insulin, andISR, whichwas
significant for both the IGT and NIDDMgroups. The differ-ences weremostsignificant when the144-minoscillatory glu-coseinfusionprotocolwasused.In additiontoreduced coeffi-cients ofcross-correlation, this analytical approach also re-vealed that in the NIDDMgroupthelagatwhich the maximal correlationbetween oscillations ofglucoseand insulin and
be-tween oscillations ofglucoseandISRoccurredwasincreasedin theNIDDMgroup.Spectral analysisof the 24-hprofiles from theoscillatory glucoseinfusion studies confirmed the failure of
glucosetocontrol theperiodicityoftheISR oscillations.Inthe
control subjects, single coinciding peaks representingthe
pe-riod of infusionwereclearlyapparent inboth theglucoseand
ISR spectra. By contrast, in the subjects with NIDDM and
IGT, althoughaclearpeak representing the periodof infusion
was evident in theglucose spectra, the ISR profiles revealed eitherabroadspectrum withoutaclearlydominantperiodora dominantperioddifferent from theperiodofinfusion andwith
significantlyreducedpowerincomparisontothe controls.We have taken this dissociationbetween the dominantpeaksinthe spectra ofglucoseandISRtoreflectanunderlyingdefectinthe
insulin/glucose feedback mechanism. The results of the spec-tralanalysisgohand in hand with the results shown in Fig. 5, which illustrate thatamplification oftheglucose oscillations results inasimilar amplificationof the ISR oscillations in the
controlgroup,butnotin the IGT and NIDDMgroups.
Thedissociation between the glucose and ISR oscillations
duringtheoscillatoryinfusions in the IGTgroupin this study canthus be addedtothe list of abnormalities ind-cellfunction that have been observed inpatients studied in theearlystages ofNIDDM.Attenuated insulinresponsestooralglucosehave
also been described in subjects with IGT who later develop
NIDDM(29-31 ) and innormoglycemicco-twins of patients
withNIDDM(32),agroupwithaveryhighrisk ofdeveloping
NIDDM. O'Rahilly and co-workers (16) reported alterations in therapid oscillations in insulin in first-degreerelatives of
patients with NIDDM, and Gerich and co-workers (33) ob-serveddelayed attenuatedresponses tooralglucoseinpatients
with mild IGT. Therelationship between theseabnormalities
and thealteredultradian oscillationsin thepresentstudyisnot knownnorhas themostsensitivetestofearly d-cell
dysfunc-tion beendefined.Nevertheless the data from thecurrentstudy
taken inconjunction withthestudies quotedaboveillustrate that with the appropriate experimental approaches, intrinsic defects in the (3 cellmaybedemonstrable inpatientswith
im-paired glucose tolerance but normalglycosylatedhemoglobin
values.
Theprecisemechanismsresponsible for the loss of
entrain-mentin the IGT and NIDDMgroups are notknown. However
amodel of the interactions betweenglucose,insulinsecretion,
andaction recentlypublishedbyour group(12)mayprovide
aninitial frameworkwithwhichtoaddress thesequestions. In
this model,themaintenance oftheregularpattern of oscilla-torysecretionobserved in normalsubjectswasdependenton a numberoffactors, themostimportant ofthesebeingthe sim-plesigmoidal relationshipbetweenglucose and insulin
secre-tion,theexistence ofatime delay in the order of 25-50min betweentheincrease ininsulinsecretion and thesuppressionof
hepaticglucoseproduction,and thefact that insulinmust dif-fuse intoaremoteperipheralcompartmentbefore stimulating peripheralglucose utilization. Alterations in thedose-response
relationship between glucose and insulin secretion (34-38),
hepatic resistance to the action of insulin (39-43), or
pro-longedtime lags in the peripherydueto diminishedglucose
disposal (42)could allcontributetothealterations in the dy-namic interactions betweeninsulin secretionandglucose
ob-served in the IGTand NIDDMgroupsin this study. To
under-standfullythe lackof entrainment observed in thesegroups,
however, itmay be necessary to modifythe model toinclude
factors other than the insulin/glucose feedback mechanism. Thepresent study alsohighlightsthe systematic change in
(-cell
secretoryresponses asobesity progresses to IGT and ulti-mately NIDDM.In TableTI,
themean 24-h insulin andC-pep-tideconcentrationsas well as themean insulinsecretoryrates in the IGT group are intermediate between those observed in
theobeseandNIDDM groups. The data in Table III
demon-strate aprogressive decline in the absolute amplitude of the glucose and ISRpulses with thetransition from obesity to IGT toNIDDM.Thesimilarityin basal insulin and C-peptide levels betweentheNIDDMandIGT groups in Table I (both ofwhich were lower than the correspondingconcentrations in the obese
[nondiabetic] subjects)
raises thepossibility
that(3-cell
secre-toryfunction decreases beforethe
development
ofseverehyper-glycemia in NIDDM. This is further illustrated in Fig. 1,
de-picting the relationship between mean glucose concentration
andmean ratesof insulin secretionthroughout the 24-h study where theinsulin secretoryresponsesinmanyof the IGT and NIDDMsubjectswere notdifferent despite the marked differ-encesinglucose levels. In thisregard, it is also ofinterest that all three methods usedtoanalyze the interaction between glucose
and insulin secretion demonstrated abnormalities in the
pa-tients with IGT thatwerequalitatively similar but lesssevere than the defects identified in the subjects with NIDDM. It is thereforepossible that theuncoupling of the ISR and glucose oscillations isakey factor in thedevelopmentofglucose intoler-anceinsusceptibleindividuals.
The lower association observed by both cross-correlation andspectralanalysisbetweenthepatternsof insulin and ISR oscillations in the IGT and NIDDMgroupsthan in the control groupisnotreadily explained in view of theequimolar secre-tion of insulin andC-peptidefrom the(-cell,which has been well documented(44-46).In contrast tonondiabetic subjects, NIDDM patients have elevated levels ofproinsulin and of proinsulin conversion intermediates (split proinsulins),
sug-gestingthat the(-cellinthis instance releases immature secre-torygranules into the circulation(47-52). Since conventional insulin radioimmunoassays cross-react with proinsulin, it is also possible that the dissociation observed between insulin
andISR in theNIDDMand IGTgroups canbeattributed in
part tooscillating proportions ofproinsulin secretion in
rela-tiontothesecretionofinsulin. Proinsulin concentrationswere
not,however, measured in theseexperiments.
Although(-cellentrainmentwaspreserved inobesity, the presentstudy raises thepossibilitythat obesesubjectsmayhave subtle abnormalities in (3-cell function. Previous studies in obeseindividualsquantitatinginsulin secretionratesand evalu-ating pulsatility following meals have demonstrated (3-cell se-cretory patterns to be similar to those observed in normal weight subjects(7,53). Inthe present study,detailed analysis
of the proportion ofISR pulses thatwereconcomitant with
pulses in glucoseyieldedconcomitance ratios of 66%and80% inthe96 min and 144 minoscillatory infusions, respectively. This contrasts with concomitance ratios exceeding 90% that werepreviously reportedforyounger, normalweight subjects
(9)studied under identicalexperimentalprotocols. The obese
subjectsin this studywerehoweverolder and presumablymore insulin resistant than the normalweight subjects studied
previ-ously. Additionalstudiesarecurrently underway todetermine
if thesedifferences could bedue to the different agesof the
subjects, and/or differences in weight and degree of insulin resistance. In light ofthe importance ofthe insulin/glucose feedback mechanism incontrollingtheoscillations in nondia-beticsubjects,thenotion that insulin resistancemay,together with intrinsic defectsin the( cell, contribute to alterations in ultradian oscillations isanintriguingpossibility thatshould be
systematicallyaddressed infuture studies.
Inconclusion,thepresent study demonstrates that the
insu-lin/glucose feedback mechanism, which plays an important
role intheregulation and generation oftheultradian oscilla-tions innondiabetic subjects, is disrupted inNIDDMand even
in conditionswhere glucosetolerance isonly minimally
im-paired. Theresultantuncoupling ofthe normaltight
dis-crepancybetweenthedominantperiodofthe spectra ofinsulin secretion and glucose. Similar changes are seen using cross-correlation analysis and pulse analysis. The dissociation be-tweentheoscillations becomesmoreseverewith the
progres-sion fromimpaired glucose tolerance to NIDDM. Whether the
abnormalsecretoryoscillationsplay apathogeneticroleinthe progression to overtdiabetesor whetherthey are merely reflec-tive of underlying defects in the : cell and peripheral tissues
remainstobedetermined.
Acknowledgments
Theauthors wishtothank the nursing staff of the ClinicalResearch
Center for expertcare ofthe subjects whoparticipated in the study. The authors also greatly appreciate the advice given by Theodore Karrison regarding statistical analyses and thetechnical contributions of Paul Rue and MarshaJackson.
Thiswork wassupported byNational Institutes of Health grants DK-31842, DK-13941, DK-20595, DK-26678, (Diabetes Research and Training Center, Clinical Nutrition Research Unit), and
RR-00055 (Clinical Research Center). Dr. O'Meara (390343) and Dr. Sturis (392574) wererecipients of Post Doctoral Fellowships from the Juvenile DiabetesFoundationInternational.
References
1. Lang, D. A., D.R. Matthews, J. Peto, and R.C. Turner. 1979. Cyclic
oscillations ofbasalplasma glucose and insulinconcentrationsin humanbeings.
N.Engl.J.Med. 301:1023-1027.
2. Lang, D. A., D.R. Matthews, M. Burnett, G. M. Ward, and R. C. Turner. 1982. Pulsatile,synchronous basalinsulin and glucagon secretion in man. Dia-betes. 31:22-26.
3. Hansen, B. C., K. C. Jen, S. B. Pek, and R. A.Wolfe.1982. Rapid oscilla-tions in plasma insulin, glucagon, and glucose in obese and normalweight hu-mans. J.Clin.Endocrinol.&Metab.54:785-792.
4. Matthews, D. R., D. A. Lang, M. A. Burnett, and R. C. Turner. 1983. Controlof pulsatile insulinsecretion inman.Diabetologia.24:231-237.
5.Simon, C.,M.Follenius, andG. Brandenberger.1987.Postprandial
oscilla-tions of plasma glucose,insulin and C-peptideinman.Diabetologia.30:769-773. 6. Simon, C., G.Brandenberger, andM.Follenius. 1987.Ultradian
oscilla-tionsofplasma glucose,insulin,andC-peptidein manduring continuousenteral nutrition.J. Clin.Endocrinol. & Metab. 64:669-674.
7.Polonsky, K. S., B. D.Given,and E. VanCauter.1988.Twenty-four-hour
profiles and pulsatile patternsof insulin secretion in normaland obesesubjects.J.
Clin.Invest.81:442-448.
8.Shapiro,E.T., H.Tillil,K.S. Polonsky,V.S. Fang,A.H.Rubenstein,and E. VanCauter. 1988. Oscillations in insulin secretionduringconstantglucose infusion innormal man:relationshiptochangesinplasmaglucose.J.Clin. Endo-crinol. & Metab. 67:307-314.
9.Sturis,J., E. VanCauter,J. D.Blackman,and K.S.Polonsky.1991. En-trainmentofpulsatile insulin secretion byoscillatoryglucose infusion. J. Clin.
Invest.87:439-445.
10. Stagner, J.I.,E.Samols, and G.C. Weir.1980.Sustained oscillations of
insulin, glucagon, andsomatostatin from theisolated caninepancreasduring
exposure to a constantglucoseconcentration.J. Clin.Invest.65:939-942.
11. Chou, H. F.,and E. Ipp. 1990.Pulsatileinsulin secretioninisolatedrat
islets. Diabetes. 39:112-117.
12.Sturis, J., K. S. Polonsky, E.Mosekilde,and E. VanCauter. 1991. Com-puter modelfor mechanisms underlying ultradian oscillations of insulin and glucose. Am. J.Physiol. 260:E801-E809.
13.Sturis, J., K. S. Polonsky, J. D.Blackman,C.Knudsen,E.Mosekilde,and E. Van Cauter. 1992.Aspectsofoscillatory insulin secretion. InComplexity,
Chaos,andBiological Evolution.NATO ASI series B. E.Mosekildeand L. Mose-kilde, editors. PlenumPress, New York. 75-93.
14.Glass, L., and M.C. Mackey. 1988. From ClockstoChaos: theRhythms
of Life. PrincetonUniversityPress, Princeton. 248 pp.
15. Lang, D. A., D. R.Matthews,M.Burnett,and R. C. Turner. 1981.Brief,
irregularoscillations of basal plasmainsulinandglucose concentrationsin
dia-beticman.Diabetes.30:435-439.
16.O'Rahilly, S., R. C. Turner, and D. R. Matthews. 1988.Impairedpulsatile
secretion of insulin inrelatives of patients with non-insulin-dependentdiabetes. N.Engl. J.Med.318:1225-1230.
17.Polonsky, K.S., B. D.Given, L.J.Hirsch, H.Tillil, E. T.Shapiro,C.
Beebe, B. H. Frank, J. A. Galloway, and E. Van Cauter. 1988. Abnormal patterns of insulin secretion innon-insulin-dependentdiabetesmellitus. N. Engl. J. Med. 318:1231-1239.
18.Simon,C., G. Brandenberger, M.Follenius,and J. L.Schlienger. 1991. Alteration in the temporalorganisation of insulinsecretion in Type 2
(non-insu-lin-dependent) diabeticpatientsundercontinuousenteral nutrition.
Diabetolo-gia.34:435-440.
19.NationalDiabetesData Group. 1979.Classification and diagnosis ofdia-betesmellitusandothercategories ofglucoseintolerance.Diabetes.
28:1039-1057.
20.Morgan,C.R., and A. Iazarow. 1963. Immunoassay of insulin: two
antibodysystem:plasma insulinlevelsofnormal,subdiabeticand diabeticrats.
Diabetes. 12:115-126.
21. Faber,0.K.,C.Binder,J. Markussen, L.G. Heding, V. K. Naithani, H. Kuzuya, P.Blix,D.L.Horwitz,and A. H.Rubenstein.1978. Characterization of sevenC-peptide antisera. Diabetes.27(Suppl.1):170-177.
22. Matthews, D. R., J. P. Hosker, A. S. Rudenski, B. A. Naylor, D. F.
Treacher, andR.C.Turner.1985.Homeostasismodel assessment:insulin resis-tanceand,-cell function fromfastingplasmaglucoseandinsulinconcentrations
inman.Diabetologia. 28:412-419.
23. Polonsky,K.S., J.Licinio-Paixao, B.D.Given,W. Pugh, P. Rue, J.
Galloway,T.Karrison,and B. Frank.1986.UseofbiosynthetichumanC-peptide in the measurementofinsulinsecretion rates in normal volunteers and type I
diabetic patients.J.Clin.Invest.77:98-105.
24. VanCauter,E., F. Mestrez, J. Sturis, and K. S. Polonsky. 1992.
Estima-tion of insulin secreEstima-tionratesfrom C-peptidelevels:comparison of individual andstandardkineticparametersfor C-peptideclearance.Diabetes.41:368-377. 25. Van Cauter, E. 1988.Estimatingfalse-positiveandfalse-negativeerrors in
analysesof hormonal pulsatility.Am. J.Physiol.254:E786-E794.
26. Cleveland,W.1979.Robustlocallyweightedregressionandsmoothing
scatterplots.J. Am.Stat.Assoc.74:829-836.
27.Fisher,R. A. 1958.StatisticalMethodsfor Research Workers. Oliver &
Boyd,Edinburgh.356 pp.
28.Jenkins, G. M.,and D.G.Watts.1968.SpectralAnalysis andIts Applica-tions. Holden Day,SanFrancisco. 525pp.
29.Kosaka,K., R.Hagura,and T.Kuzuya. 1977. Insulinresponsesin equivo-cal anddefinitediabetes,withspecialreferencetosubjects who had mild glucose intolerance but laterdeveloped definite diabetes.Diabetes. 26:944-952.
30.Kadowaki, T.,Y.Miyake,R.Hagura,Y.Akanuma,H.Kajinuma,N. Kuzuya, F.Takaku,and K. Kosaka. 1984. Riskfactors for worseningtodiabetes
insubjects with impairedglucosetolerance.Diabetologia.26:44-49.
31.Efendic, S.,R.Luft,and A.Wajngot. 1984.Aspectsofthepathogenesis of
type 2diabetes.Endocr.Rev.5:395-410.
32. Barnett, A. H., A. J.Spiliopoulos,D. A.Pyke, W. A. Stubbs, J.Burrin,and K.G. Alberti. 1981.Metabolic studies inunaffected co-twinsof non-insulin-de-pendent diabetics.Br.Med.J.282:1656-1658.
33.Mitrakou,A., D.Kelley,M.Mokan,T.Veneman,T.Pangburn,J.Reilly,
and J.Gerich. 1992. Role ofsuppressionof glucoseproductionanddiminished early insulin release inimpairedglucose tolerance.N.Engl.J.Med. 326:22-29. 34. Ward,W.K.,D.C.Bolgiano,B.McKnight,J. B.Halter,and D. Porte, Jr.
1984. DiminishedB cellsecretory capacityinpatientswith noninsulin-dependent diabetes mellitus.J.Clin.Invest.74:1318-1328.
35.Pfeifer,M. A., J. B.Halter,and D. Porte, Jr. 1981. Insulin secretion in diabetes mellitus.Am.J.Med. 70:579-588.
36.Garvey,W.T.,J. M. Olefsky,J.Griffin,R. F.Harnman,and0.G. Kolterman.1985. Theeffect ofinsulintreatmentoninsulin secretionandinsulin
actionin type IIdiabetesmellitus.Diabetes. 34:222-234.
37.Ferner,R.E.,L.Ashworth,B.Tronier,andK.G.M.M.Alberti. 1986.
Effects ofshort-termhyperglycemiaoninsulin secretion innormalhumans.Am.
J.Physiol. 250:E655-E661.
38.Nesher, R., L.DellaCasa,Y.Litvin,J.Sinai,G. DelRio,B.Pevsner,Y. Wax, and E.Cerasi. 1987.Insulindeficiencyandinsulin resistancein type 2
(non-insulin-dependent)diabetes:quantitativecontributions ofpancreaticand peripheralresponsestoglucosehomeostasis.Eur.J.Clin.Invest. 17:266-274.
39.Firth,R., P.Bell,and R.Rizza. 1987.Insulin actioninnon-insulin
depen-dentdiabetesmellitus:therelationshipbetweenhepaticandextrahepaticinsulin
resistanceandobesity. Metab. Clin. Exp. 36:1091-1095.
40.Glauber, H.,P. Wallace,andG. Brechtel. 1987.Effects offastingon
plasmaglucoseandprolongedtracer measurementofhepatic glucoseoutput in NIDDM.Diabetes.36:1187-1194.
41.Bogardus, C.,S.Lillioja,B.V.Howard,G.Reaven,andD.Mott.1984.
Relationships betweeninsulinsecretion,insulinaction,andfasting plasma
glu-coseconcentrationinnondiabeticandnoninsulin-dependentdiabeticsubjects.J.
Clin.Invest.74:1238-1246.
42.Kolterman, 0.G.,R.S.Gray,J.Griffin,P.Burstein,J.Insel,J.A.Scarlett,
and J.M.Olefsky. 1981.Receptorand postreceptor defects contributetothe insulinresistance in noninsulin-dependent diabetesmellitus. J. Clin. Invest. 68:957-969.
43.Powrie,J.K.,G. D.Smith,T. R.Hennessy,F.Shojaee-Moradie,J. M.
glucoseproductioninnon-insulin dependent diabetesmellitus measured with
[6,6-2H2]glucoseenrichedglucoseinfusionduringthehyperinsulinaemic
eugly-caemic clamps.Eur.J. Clin. Invest. 22:244-253.
44.Rubenstein,A.H.,J. L.Clark,F.Melani, and D. F. Steiner. 1969. Secre-tion of proinsulin, C-peptide by pancreatic B cells and its circulaSecre-tion in blood. Nature(Lond.).224:697-699.
45. Steiner, D. F. 1978. On the role ofproinsulin C-peptide. Diabetes 27(Suppl. 1):145-148.
46.Horwitz, D.L., J. I.Starr,M. E.Mako, W. G. Blackward, and A. H. Rubenstein. 1975.Proinsulin, insulinandC-peptide concentrationsin human portal andperipheral blood.J.Clin.Invest.55:1278-1283.
47.Duckworth, W. C., and A. E.Kitabachi. 1972.Directmeasurementsof
plasmaproinsulin in normal and diabetic subjects. Am. J.Med.53:418-427. 48. Mako, M. E., J. I.Starr,A.H.Rubenstein. 1977.Circulating proinsulin in
patientswithmaturityonsetdiabetes.Am.J.Med. 63:865-869.
49. Ward, W. K., E. C. LaCava, T. L. Paquette, J. C. Beard, B. J. Wallum, and D.Porte,Jr. 1987.Disproportionateelevation of immunoreactive proinsulinin
type 2(non-insulin-dependent)diabetesmellitusandin experimental insulin
resistance. Diabetologia.30:698-702.
50.Yoshioka,N., T. Kuzuya, A. Matsuda, M. Taniguch, and Y. Iwamoto. 1988. Serum proinsulin levels at fasting and after oral glucose load in patients with type 2(non-insulin-dependent) diabetes mellitus. Diabetologia. 31:355-360.
51. Temple, R. C., C. A. Carrington, S. D. Luzio, D. R. Owens, A. E.
Schneider,W.J.Sobey,andC. N. Hales. 1989. Insulindeficiencyin
non-insulin-dependentdiabetes. Lancet 1 (8633):293-295.
52. Saad, M. F., S. E. Kahn, R. G. Nelson, D. J. Pettitt, W. C. Knowler, M. W.
Schwartz, S. Kowalyk,P. H.Bennett, and D. Porte, Jr.1990.Disproportionately
elevatedproinsulin in Pima Indians with noninsulin-dependentdiabetes melli-tus. J.Clin. Endocrinol. &Metab.70:1247-1253.