The relationship between large-for-gestational-age infants and glycemic control in women with gestational diabetes
Oded Langer, MD, and Roger Mazze, PhD
Bronx, New York
In this prospective study 246 women with gestational diabetes were followed up to determine the characteristics of metabolic control associated with large-for-gestational-age infants. Memory-based reflectance meters were used for self-monitoring blood glucose. Ambulatory glucose profiles were produced to characterize glycemic control levels throughout pregnancy. With these novel approaches to the collection and representaiion of glucose data, the severity of glucose intolerance (hyperglycemia)_ was found to be associated with both maternal and neonatal morbidity in terms of infant size and cesarean section rate. By use of hierarchical cluster analysis to identify three groups on the basis of control levels (low <87 mg/di, mid 87 to 105 mg/di, high >105 mg/di) we were able to show a positive outcome in the low group with reduced rates of large-for-gestational-age (2%) and macrosomatic (0%) infants. Furthermore, we showed that as mean blood glucose levels and instability in glycemic control ihcreased from group to group, incidence of large-for-gestational-age and macrosomatic infants increased. Whereas obesity increased the relative risk of adverse neonatal outcome, type of treatment (insulin versus diet) did not appear to be significant. Appropriately monitored toward stability within a narrow range to achieve tight metabolic control, ambulatory glycemia in pregnancy is associated with a decreased risk of maternal and fetal complications. (AM J 0BSTET GYNECOL 1988;159:1478-83.)
Key words: Gestational diabetes, ambulatory glucose profile, macrosomatia, level of control
Perhaps the most perplexing problem in the treatment of gestational diabetes is the significant incidence of large-for-gestational-age (LGA) and macrosomatic infants despite recent advances in the detection and treatment of gestational diabetes. 1·
3 It is well recognized that a dichotomy exists between metabolic control and neonatal complications. Despite numerous studies, the exact nature of the relationship between glycemic control in gestational diabetes and neonatal size remains unclear. Whereas there is evidence of the likely association between glycemic control before and during pregnancy of the woman with pregestational diabetes, no similar evidence to link the level of glycemic control at diagnosis or thereafter to neonatal outcome exists in gestational diabetes. Why? Some investigators have suggested the lack of an association between glucose levels and neonatal complications may be attributed to dif-
From the Departments of Obstetrics and Gynecology and Epidemiology and Social Medicine and the Diabetes Research and Training Center, Albert Einstein College of Medicine.
Supported in part by a National Institutes of Health Diabetes Research and Training Center grant and the Diagnostic Division of Miles Laboratories.
Presented at the Thirty-fifth Annual Meeting for the Society for Gynecologic Investigation, Baltimore, Maryland, March 17-20, 1988.
Reprint requests: Oded Langer, MD, Department of Obstetrics and Gynecology, Weiler Hospital, Albert Einstein College of Medicine, 1825 Eastchester Road, Bronx, NY 10461.
1478
ficulty in the assessment of glycemic control without the subjection of the patient to long-term hospitalization. Indeed, there is substantial evidence that alternative sources of data collected to reflect glycemic control (e.g., patient self-report, in-office randorri blood glucose and glycosylated hemoglobin tests) are subject to such a wide margin of error that it would not be possible to draw meaningful clinical conclusions. With the advent of the memory-based reflectance meter, however, a method for the collection and representation of ambulatory glycemic patterns has been developed; this has made it possible to examine the nature of the relationship between glycemic control and neonatal outcome.•· 5 This study sought to characterize metabolic control by means of a novel feature. All participants were equipped with memory-based reflectance meters from the time of diagnosis to delivery. The meters recorded all blood glucose values in correspondence with time and date. With the collection of accurate and verified self-monitored blood glucose data, we sought to determine the association between overall metabolic control and neonatal outcome among women with gestational diabetes treated with insulin or diet.
Method and materials
Women whose conditions were diagnosed according to the National Diabetes Data Group criteria for gestational diabetes were admitted to the study.6 They were
Volume 159 Number 6
359
1390
259
289
159
189
59
'--
LGA infants and glycemic control 1479
...
Fig. 1. Typical ambulatory glucose profile in which median is represented by middle curve, inter
quartile range is represented by distance between second and fourth curves, and number and
magnitude of reversals are measured from median curve.
Table I. Maternal and neonatal characteristics
Characteristics
Age Body mass index Obese(%) Previous LGA (%) Previous stillbirth (%) Previous gestational diabetes
(%) Family history of diabetes (%) Weight gain (kg) Delivery type (%)
Vaginal Primary cesarean section Repeat cesarean section
Gestational age (wk) at entry Gestational age (wk) at delivery Birth percentile Birth weight (gm)
*P < o.o4.
Low-86 (n = 53)
32 ± 6 28 ± 7
43 4 2
2 43
15 ± 7
86 8 6
28 ± 5 39 ± I 53 ± 25
3174 ± 467
assigned a regimen of diet only or multiple injections
of mixed insulins on the basis of fasting plasma glucose
levels during an oral glucose tolerance test (OGTT) and
mean ambulatory blood glucose levels during preg
nancy.7· 8 All participants were taught self-monitoring
of blood glucose levels with a memory-based reflectance
meter. Study participants were requested to test their
blood seven times per day. The reflectance meter con
tained an onboard memory that permitted storage of
320 glucose readings with the corresponding times and
dates. The data from the meter were directly trans
ferred to a microcompu~er for rapid aggregation. An
ambulatory glucose profile was produced to reflect
Glycemic control level (mg! di)
87-105 106-High (n = 142) (n = 51)
31 ± 5 31 ± 6 28 ± 5 31 ± 7*
56 53 13 5 2 1
6 2 46 37
14 ± 7 16 ± 6
77 63 12 21 II 16
29 ± 5 28 ± 5 39 ± 2 39 ± 2 57 ± 23 59 ± 25
3250 ± 457 3261 ± 654
glucose patterns by the representation of all self
monitored blood glucose values as a series of curves
plotted as if they occurred on a single typical day (Fig.
1). Five properties that constitute the ambulatory glu
cose profile are mean, median, interquartile range (dis
tance between the twenty-fifth and seventy-fifth per
centiles), reversals in curve, and excursions from the
median." Participants were requested to return to the meta
bolic study unit at 2-week intervals to have adjustment
of insulin or diet made to optimize control. Assessment
of both maternal and fetal growth occurred at each
visit. Maternal size was determined bv bodv mass index
1480 Langer and Mazze
LO C<87l MIDC87·105l HJ. 0105l
December 1988 Am .J Obstet Gynecol
Fig. 2. Incidence of large infants by glucose cluster. Odd ratio between LO and MID = 2.41 (p < 0.05), odds ratio between MID and HI = 3.57 (p < 0.02), odds ratio between LO and HI = 7.95 (p < 0.002).
Table II. Gestational diabetes: Characteristics of glycemic control levels
Glycemic control levels (mg! dl)
Group I Group 2 Group 3 (Low-86) (87-105) (106-High)
Mean blood 80 ± 6 95 ± 5 116 ± 9 glucose
Interquartile 29 ± IO* 30 ± IO* 38 ± 10 range
Number of 6±4 4 ± 3* 3 ± 2* reversals
Average 17 ± II 28 ± 12 34 ± 13 excurs10n
Significance (p < .000 I) for all ambulatory glucose profile parameters between groups except those designated by an asterisk (*).
(body mass index = weight 7 height2 in kilograms pe1 meter). At delivery neonatal size was standardized according to a fetal growth curve. '0 LGA was defined as > the ninetieth percentile; macrosomia was defined as >4 kg.
Results
Participant characteristics. There were 246 women in this prospective study of gestational diabetes. For analysis purposes, at the completion of the study, the women were divided by hierarchical cluster analysis into three groups on the basis of overall mean blood glucose levels during the third trimester after stabilization of glycemia: Low (<87 mg/di); mid (87 to 105 mg/di); high (>105 mg/di). Demographic and biophysical data are presented in Table I. Overall, 9% of the participants had previous gestational diabetes, 22%
had previous large infants, and 5% had pregnancies that resulted in stillbirths. The overall cesarean section rate in the study was 24%, with 13% primary and I I% repeat rates. Of the 27 women with previous cesarean sections 89% had vaginal delivery in this study and the remaining I l % had repeat cesarean sections. The study population younger than 26 years of age (n = 32, 13%) differed from older participants with regard to the percentage treated with insulin (42% versus 66%, respectively, p = 0.00 l ).
Characterization of glycemic control by (OGTT) and ambulatory glucose profile. Evaluation of metabolic control began with the (OGTT) to determine the severity of glucose intolerance. No significant difference in (OGTT) results was found among the groups (e.g., fasting: low= IOI ± 19; mid = IOI ± 14; high = 103 ± 16 mg/di; as well as at each hourly determination). Analysis of the relationship between each (OGTT) determination and mean blood glucose values for the study period was not significant. We used the ambulatory glucose profile to characterize metabolic control in the three groups (Table II). Metabolic profiles of the three groups showed that the 53 women in the low group and the 142 women in the mid group had a significantly lower interquartile range (variability) than the 51 women in the high group. In contrast, the low group had a significantly higher number of reversals (oscillation in stability) than the other groups. The magnitude of variability (excursions) indicated the high group had the greatest change in control levels after a reversal (34 ± 13 mg/di), compared with the mid group (28 ± 12 mg/di) and the low group (17 ±II mg/di). Thus these groups appeared to have significant differences in metabolic characteristics related to sta-
Volume 159 Number 6
Table III. Association among large infants
(>4000 gm or >90%) treatment modalities
and glyce~ic control levels
Control level (mgldl) Diet Insulin
Low-86 % 2 0 (n) (I /21) (0/29)
87 to I05 % 10 IO (n) (5/49) (9/93)
106-High % 27 17 (n) (6/22) (5/29)
Totals % 13 9 (n) (12/92) (14/154)
Total
2 (I I 53)
IO (14/ 142)
22 (ll/51)
11 (26/246)
bility and variability as well as overall control as mea
sured by mean blood glucose levels.
Level of ambulatory glycemia, treatment modality,
and neonatal size. We then examined the infant size
by group and by treatment modality within the groups.
Fig. 2 shows the frequency distribution by group for
macrosomic and LGA infants. A significant difference
was found for the combined incidence of macrosomia
and LGA infants between each group. The high group
had a nearly tenfold increase in the incidence of ma
crosomic and LGA infants compared with the low
group and an incidence twofold higher than the mid
group. In the low group no case of macrosomia was
found, whereas half of the large infants were macro
somic in the mid group and two thirds of the large
infants were macrosomic in the high group.
We pooled all patients and correlated the birth per
centiles and each ambulatory glucose profile property.
Mean blood glucose values and number of reversals in
the curves were positively correlated (r = 0.16, p =
0.01). However, analysis of the relationship between
each ambulatory glucose profile characteristic and LGA
and macrosomic infants revealed that mean blood
glucose values and the degree of excursion were not
correlated, whereas the interquartile range and the
number of reversals were significantly related (r =
0.13, p = 0.05; r = 0.50, p = 0.04) to LGA and mac
rosomic infants. Thus a linear relationship between in
fant size and blood glucose level appears to occur at
low- and mid-range blood glucose levels, whereas the
high-range blood glucose level per se does not correlate
with infant size. Instead, infant size appears to be re
lated to the degree of fluctuations (reversals and inter
quartile range) in blood glucose.
Investigation of treatment modality within each
group revealed that, whe.reas no significant overall dif
ference was found for diet versus insulin in terms of
large infants, the percentage of LGA and macrosomic
LGA infants and glycemic control 1481
Table IV. Association among maternal size,
large infants, and glycemic control levels
Obese Nonobese
Low-86 % 4 0 Subtotal (I /23) (0/30)
87-I05 % 15* 3 Subtotal (12179) (2/63)
>106-High % 19 25 Subtotal (5/27) (6/24)
Totals % 14 7 (n) (18/ 129)* (81117)
*Odds ratio 5.46 (p < 0.04).
infants was highest among those treated with diet in
both the mid and high groups. Within the low group
no difference in risk of LGA and macrosomic infants
was found with regard to treatment. It was found, how
ever, that as the mean blood glucose levels increased
(within the group) the risk of LGA infants and macro
somia within the same treatment category increased
(Table Ill).
Maternal size and neonatal size. We then evaluated
the impact of obesity on neonatal outcome. Analysis by
ambulatory glucose profile revealed no significant dif
ference with regard to the five parameters when the
obese and nonobese participants were pooled. How
ever, separate analysis within each group for the obese
and nonobese women revealed significant differences
in interquartile range and glucose stability. As glucose
level increased within the group, both the interquartile
range and the magnitude of the excursions throughout
the day increased. Examination of the risk of LGA and
macrosomic infants associated with obesity showed a
twofold greater incidence among obese women. When
level of metabolic control was considered, the obese
patient in the mid group was found to be at a fivefold
risk (odds ratio, 5.5, p = 0.03), compared with the non
obese counterpart at the same level of metabolic control
(Table IV). The impact of the three factors (control level, treat
ment type, and maternal size) on neonatal size was
considered (Fig. 3). Four categories were identified.
Among the diet-treated obese participants, 42% in the
high, 30% in the mid, and 20% in the low groups had
large infants. In contrast, obese women treated with
insulin in both the mid and high groups had the same
incidence of LGA and macrosomia (approximately
10%). Among the nonobese participants, treatment re
sulted in a significant difference in incidence of LGA
and macrosomia among those with mean blood glucose
values >I 05 mg/ di. Those treated with insulin in the
1482 Langer and Mazze
50 PERCENT
40
20
10
0 DIET l'OOESE INS/a3ESE DIET /l{);·CJ:ESE lt-EYtDHl3ESE
llllllBH! 0105)
=MID (87-105)
December 1988 Am J Obstet Gynecol
Fig. 3. Incidence of large infants by treatment, obesity, and glucose cluster.
Table V. Association between glycemic control level and neonatal outcome
Blood glucose level
<87 >86 Total n = 53 n = 193 N = 246
(%) (%) (%)
Polycythemia 3 6 5 (>60%)
Hyperbilirubinemia 0 9 7 (>12 mg/di)
Hypoglycemia 5 9 8 (<35 mg/di)
H ypocalcemia 3 (<7.5 mg%)
Respiratory distress 3 3 3 syndrome
Anomalies 0 2 2 Neonatal intensive care 5 9 8
unit
high group had a 12 % higher incidence of large infants than their diet-treated counterparts.
Neonatal outcome. Comparison of diet- and insulintreated participants showed no significant difference in neonatal metabolic measurements. Overall, 8% of the infants experienced some type of metabolic disturbance. When the incidence of neonatal intensive care admissions was compared no difference was found (4% diet-treated and 11 % insulin-treated). Further analysis by group according to level of maternal glycemic control showed a slightly higher (nonsignificant) difference (2% versus 6%) for all neonatal metabolic disturbances. Examination of fetal anomalies revealed a 2% rate, which was comparable with the general population served in the medical center (Table V).
Comment
The study showed that among women with comparable glucose intolerance after initiation of treatment for gestational diabetes, those who achieved the lowest range of blood glucose values throughout pregnancy (the low group with mean blood glucose levels <87 mg/ di) appeared to have a significantly lower incidence of LGA and macrosomia than the mid and high groups. Average blood glucose levels of 87 to 105 mg/di increased the incidence twofold for the development of LGA and macrosomia. When average blood glucose values increased to > 105 mg/ di, the incidence of large infants quadrupled when compared with the first group and doubled when compared with the second group. When examined independently of level of control, obesity was found to cause a twofold increase in the incidence of LGA and macrosomia when compared with women of normal body mass. Within each group, however, obesity did not significantly increase the risk of adverse neonatal outcome except in the mid group (87 to 105 mg/di), where the increased risk was nearly fivefold. Treatment modality did not increase the risk of LGA and macrosomia within any group.
Overall, the findings suggest glucose intolerance that existed before initiation of treatment need not be a barrier to achievement of euglycemia (to the extent that patients are assigned a treatment modality designed to optimize control); nor does the OGTT predict neonatal outcome. Instead, it appears that morbidity is related to control of blood glucose levels in pregnancy as shown in the case of the highest-risk group comprised of women with mean blood glucose levels> 105 mg/di and body mass index >27, whereas at lowest-risk group was made up of women with mean blood glucose levels <87 mg/di and body mass index <27. Between these two
Volume 159 Number 6
groups, the increase in risk appeared to be related to blood glucose levels in the following manner. There
appear to be two points of demarcation, 87 and 105 mg/di. Increases in blood glucose levels within each group did not appear to be in a linear relationship to outcome, whereas between groups the blood glucose levels appear to define the levels of risk.
Because the severity of glucose abnormalities measured by OGTT before treatment did not appear to be related to outcome, we targeted tight metabolic control as a possible factor in neonatal outcome. To achieve tight glycemic control, we collected verified selfmonitoring of blood glucose data to assure proper analysis of glucose levels. With the ambulatory glucose profile we were able to demonstrate that measures other than the mean could be used to characterize control
and were sensitive to subtle changes that might explain the incidence of adverse neonatal outcome. For ex
ample, with an increased risk of LGA and macrosomic infants we found that both the interquartile range and the size of excursions in the median curve increased and that there was less stability (greater fluctuation) in the median curve as reflected in the increased number of reversals. Furthermore, there were significant differences in mean blood glucose levels and interquartile range when ambulatory glucose profiles of obese (14% LGA) infants and the nonobese (7% LGA) infants were compared. Use of the ambulatory glucose profile revealed that the higher incidence of LGA infants in the obese group was reflected in a wider interquartile range and that this parameter may need greater clinical attention in the treatment of gestational diabetes.
It has been recognized for some time that treatment for gestational diabetes has two principal functions, prevention of stillbirths and reduction in the incidence of LGA infants, which in turn will reduce maternal and neonatal morbidity.'' Despite significant decrease in the rate of stillbirths during the past decade, the incidence of large infants and the high cesarean section rate have remained virtually unchanged. 12
-14 Some researchers
have directed their attention to this public health problem by advancing the prophylactic or intensive use of insulin for all gestational diabetes."· 16 In those studies the incidence of large infants was substantially reduced to as low as 10%. Is the use of insulin the answer? We believe that conventional treatment augmented with self-monitoring of blood glucose levels with verified data and analysis of glucose patterns with ambulatory glucose profiles can produce the same results without unnecessary hazard to women with gestational diabetes. In this study, 80% of the women had a mean blood glucose range throughout pregnancy of < 105 mg I di, with 50% treated on diet-only regimens. Among those women in the low glucose group, this resulted in no macrosomia and 2% LGA infants. Even among our
LGA infants and glycemic control 1483
midrange participants, rates of macrosomic and LGA infa,nts were low (10%).
With regard to maternal morbidity, the cesarean rate was found to be lowest among the low group and highest among those women with mean blood glucose levels > 105 mg I di. Perhaps most important, we were able to identify and characterize an ambulatory glycemic profile defined by a glucose range (<87 mg/di) and level of stability (29 ± 10 mg I di interquartile range) that could be reached with conventional treatment and to associate this with successful reduction in the incidence of neonatal and maternal morbidity.
REFERENCES
I. Gyves MT, Schulman PK, Merkatz IR. Results of individualized intervention in gestational diabetes. Diabetes Care 1980;3 :495-6.
2. Philipson EH, Kalhan SC, Edelberg SC, Williams TG. Maternal obesity as a risk factor in gestational diabetes. Am J Perinatal 1985;2:268-70.
3. Lavin JP, Lovelace DR, Miodovnik M. Clinical experience with one hundred seven diabetic pregnancies. AM J OBSTET GYNECOL 1983;147:742-52.
4. Langer 0, Mazze R. Diabetes in pregnancy: evaluation of self-monitoring performance and glycemic control with memory-based reflectance meters. AM J 0BSTET GYNECOL 1986; 155:635-7.
5. Goldberg JD, Franklin B, Lasser D, et al. Gestational diabetes: impact of home glucose monitoring on neonatal birth weight. AM J OBSTET GYNECOL 1986; 154:546-50.
6. National Diabetes Data Group. Classification and diagnosis of diabetes mellitus and other categories of glucose intolerance. Diabetes 1979; 1039-57.
7. Mazze RS, Langer 0. Primary, secondary, and tertiary prevention: program for diabetes in pregnancy. Diabetes Care 1988;ll(suppl 3):263-8.
8. Langer 0, Anyaegbunam A, Brustman L, Guidetti D, Mazze R. Gestational diabetes: insulin requirements in pregnancy. AM J OBSTET GYNECOL 1987; 157:669-75.
9. Mazze RS, Lucido D, Langer 0, Hartmann K, Rodbard D. Ambulatory glucose profile: representation of verified self-monitored blood glucose data. Diabetes Care 1987; IO( suppl I): 111-2.
10. Brenner WE, Edelman DA, Hendricks CH. A standard of fetal growth for the United States of America. AM J 0BSTET GYNECOL 1976; 126:555-64.
11. Karlsson K, Kjellmer I. The outcome of diabetic pregnancies in relation to the mother's blood sugar level. AM J 0BSTET GYNECOL 1972;112:213-20.
12. Gabbe SG, Mestman JG, Freeman RK, Anderson GV, Lowensohn RI. Management and outcome of Class A diabetes mellitus. AM J OBSTET GYNECOL 1977; 127 :465-9.
13. Metzger BE, Phelps RL, Freinkel N. Navickas IA. Effects of gestational diabetes on diurnal profiles of plasma glucose, lipids, and individual amino acids. Diabetes Care l 980;3:402-9.
14. Persson B, Stangenberg M, Hansson U, Nordlander E. Gestational diabetes mellitus (GDM). Comparative evaluation of two treatment regimens, diet versus insulin and diet. Diabetes l 985;34(suppl 2): 101-5.
15. Coustan DR, lmarah J. Prophylactic insulin treatment of gestational diabetes reduces the incidence ofmacrosomia, operative delivery, and birth trauma. AM J OBSTET GYNECOL 1984; 150:836-42.
16. Roversi GD, Garbiulo M, Nicolini U, et al. Maximal tolerated insulin therapy in gestational diabetes. Diabetes Care !980;3:489-94.