Share this article on:

A High-Sensitivity Assay for Amniotic Fluid Insulin at 14–20 Weeks' Gestation


Original Research

Objective To examine sensitivity, precision, and sample stability of an immunochemiluminomimetric insulin assay in 14–20 week amniotic fluid (AF) and fetal age distribution of second-trimester AF insulin concentrations.

Methods We assayed fresh specimens from 576 gravidas who had amniocentesis at 14–20 weeks' gestation because of maternal age. In a preliminary study, samples were divided into aliquots and stored at 4C and −20C for 24 hours to assess freezing effect. Some samples stored at 4C were assayed repeatedly during a 14-day period and others, stored at −20C, were assayed after a 70-day period.

Results This assay reliably measured AF insulin to a detection limit of 0.03 μIU/mL. Insulin could be measured in all amniotic fluid samples and demonstrated a log10 Gaussian distribution, ranging from 0.24 to 7.41 μIU/mL. Interassay coefficients of variation ranged from 4.4 to 8.9% at concentrations of 0.4–2.0 μIU/mL. Linearity of dilution from 1.5 to 10 times was 99.2 ± 8.6%. Spike recovery of 10 μIU/mL was from 92–109%. Recovery after freezing to −20C for 24 hours (101%) and 70 days (97%) and after storage at 4C for 14 days (97%) demonstrated no significant loss.

Conclusion A two-site, dual monoclonal, immunochemiluminomimetric insulin assay was sufficiently sensitive and precise within the lower range of measured AF insulin concentrations to investigate clinical associations of 14–20 week AF insulin with maternal and fetal conditions. The insulin stability in this matrix suggests that assays can be reliable on specimens stored up to 70 days.

An immunochemiluminomimetric assay demonstrated log10 Gaussian insulin concentrations in 14–20 week amniotic fluid at greater than eight-fold the assay detection limit and stability for 70 days.

Departments of Obstetrics and Gynecology and Pathology, Brown University, Providence, Rhode Island, and Beckman Instruments, Chaska, Minnesota.

Address reprint requests to: Marshall W. Carpenter, MD, Women and Infants Hospital of Rhode Island, 101 Dudley Street, Providence, RI 02905; E-mail:

Financial Disclosure

Dr. Shellum was a salaried employee of Beckman Coulter, Inc. at the time of this research.

Received January 4, 1999. Received in revised form April 15, 1999. Accepted April 30, 1999.

In vitro insulin production by human fetal pancreatic tissue has been identified as early as 11 weeks' gestation and might be augmented in fetuses of diabetic mothers.1 However, the physiologic role of insulin and the pancreatic response to various stimuli in the early fetal period are poorly understood.

Insulin has been identified in amniotic fluid (AF) in trace amounts by radioimmunoassay as early as 12–16 weeks' gestation and has been measured at concentrations of 1–4 μIU/mL at 16 weeks.2 Amniotic fluid insulin appears to be of fetal origin and reaches the amniotic space from fetal urinary excretion.2 Amniotic fluid insulin concentrations are absent or lower in fetal death and lower in fetal growth restriction. Particularly among hyperinsulinemic fetuses, AF insulin concentration is associated with cord plasma insulin concentration.2,3 In the third trimester, increased AF insulin appears to be associated with maternal gestational diabetes, diabetic fetopathy (macrosomia, neonatal hypoglycemia),3 and increased birth weight in nondiabetic pregnancy.4

Both a case-control and a cohort study of women who had amniocentesis to determine fetal karyotype because of advanced maternal age found an association between high AF insulin concentration and subsequent gestational diabetes.5,6 However, in both studies the radio-immunoassay demonstrated an unacceptably high coefficient of variation at lower concentrations, which are characteristic of AF at this stage of pregnancy, and did not identify insulin in a significant proportion of specimens.

In the present study, we examined the sensitivity and precision of a new, two-site, monoclonal, immunochemiluminomimetric insulin assay (Klee G, Furlow B, Shellum C, Smith T. Development of an automated immunochemiluminometric insulin assay for the Access immunoassay system [abstract]. Clin Chem 1995;41:6, S34) on fresh and frozen AF and the distribution of second-trimester AF insulin concentrations at 14–20 weeks' gestation. The specificity of the assay was determined by specimen dilution and spike recovery. Stability of insulin in the AF matrix was observed at 4C for 2 weeks and at −20C for 70 days. Assay precision was determined at multiple insulin concentrations.

Back to Top | Article Outline

Materials and Methods

Assays were performed on AF samples of 576 gravidas who had amniocentesis at 14–20 weeks' confirmed gestation who chose amniocentesis for fetal karyotype based solely on their advanced maternal age (over 34 years at the expected date of delivery). Fresh AF specimens received by the Women and Infants Hospital cytogenetics laboratory from June 13, 1997, to July 8, 1998 that were not contaminated by blood and were available for immediate handling were accepted for study. Patients were excluded from the study if the indications for amniocentesis were prior abnormal serum studies that increased risk of karyotypic abnormalities or open fetal defects or if maternal diabetes was noted on the laboratory request form. Multifetal pregnancies were excluded. Amniotic fluid was stored at 4C within 6 hours of amniocentesis. Within 24 hours, aliquots were also stored at −20C. In a preliminary study, 11 samples were divided into aliquots stored for 24 hours at 4C and −20C, to assess AF insulin freeze and thaw stability. The stability of insulin in the AF matrix was examined in eight specimens assayed after storage at 4C for 1, 2, 5, 8, and 14 days. Another sample subset (n = 4) was assayed after storage for 70 days at −20C. Thereafter, assays were performed after storage at −20C for 9 ± 3 days from amniocentesis.

We used an automated immunochemiluminomimetric assay (Access Immunoassay System, Beckman Coulter, Inc., Chaska, MN). Two monoclonal antibodies (clones 19 and P10) directed to epitopes specific to insulin7 form a solid phase sandwich.8 One antibody is bound to paramagnetic particles. The other antibody is bound to an alkaline phosphatase, which acts on a chemiluminescent substrate.

The assay was validated by examining linearity of dilution and recovery of insulin in AF samples spiked with a known concentration of insulin. The assay's lower limit of detection, the concentration of insulin at two standard deviations above zero calibrator, and the interassay coefficient of variation at concentrations between 0.4 and 2.0 μIU/mL were documented.

The effect of gestational age on AF insulin concentration was determined by linear regression of weighted weekly median log AF insulin values on gestational age. The positive association of AF insulin concentration with gestational age was addressed by calculating medians of AF insulin concentration for each completed week, then giving a numeric weight to each median corresponding to the number of samples in each completed week. The equation of the linear least square regression on the weighted medians was then used to calculate multiples of the gestational age-specific medians (MoM) for each AF insulin concentration. A probability plot was used to illustrate the normal distribution of log AF insulin MoMs. The D'Agostino test9 for skewness-kurtosis was used to test for normality. Statistical significance was as accepted at the P = .05 threshold. Stata 5.0 (Stata Corp., College Station, TX) software was used for statistical tests. This investigation was approved by the Women and Infants Hospital's institutional review board.

Back to Top | Article Outline


Table 1 shows demographic characteristics of the subjects. By design, only patients with age at expected date of delivery of over 34 years were included for study. Median maternal age was 38 years. Racial distribution was consistent with that of the metropolitan Providence area. During the 12 months after July 1, 1997, roughly the time of sample acquisition for the study, the proportion of patients with gestational diabetes was 5.5% (479 of 8710 births) and with chronic diabetes was 0.8% (72 of 8710 births) at Women and Infants Hospital, the site of delivery of 56% of the pregnancies sampled in this study.

Table 1

Table 1

Table 2 shows the assay characteristics. The insulin assay's lower limit of detection in AF was 0.03 μIU/mL, similar to that described in validation studies in serum (Klee G, et al. Clin Chem 1995;41:6, S34). Insulin was measured in all 576 AF specimens, with a range of measured values of 0.24 to 7.41 μIU/mL. The total coefficient of variation was calculated from assays performed in duplicate, 21 times over a period of 30 days. Amniotic fluid samples used for this experiment had mean insulin concentrations of 0.40, 0.69, 1.24, and 2.00 μIU/mL. There was no loss in precision at lower concentrations.

Table 2

Table 2

Compatibility of the assay for insulin in AF was demonstrated by recovery after sample dilution and spiking with a known concentration of insulin. Dilution of three samples 1.5-fold to tenfold resulted in insulin recovery of 99.2 ± 8.6%. Four AF samples were spiked with 10 μIU/mL of insulin and 92–109% recovery was observed. The specificity of the assay for insulin and lack of reactivity to proinsulin and C-peptide has been described earlier (Klee et al, above).

The stability of insulin in the AF matrix was tested by comparing samples frozen at −20C and thawed to those stored for the same duration at 4C within 24 hours of amniocentesis. Frozen samples assayed within 24 hours of amniocentesis had a 101% recovery of insulin compared with samples stored for the same duration at 4C. After 14 days of storage at 4C, recovery was 97% (Figure 1). After 70 days of storage at −20C, recovery was also 97%.

Figure 1

Figure 1

A weighted regression of the median AF insulin concentration at each week (14–20) of gestation on gestational age (Figure 2) showed a significant positive association (log AF insulin = 0.104 weeks' gestation − 1.665, P < .001). Insulin values, expressed as MoM, are normally distributed on the log10 scale (Figure 3). Log AF insulin MoM values conformed to a Gaussian distribution (P > .07, indicating a nonsignificant difference from normality) as illustrated by the probability plot in Figure 4.

Figure 2

Figure 2

Figure 3

Figure 3

Figure 4

Figure 4

Back to Top | Article Outline


Prevention of fetal macrosomia in women with glucose intolerance is justified by the association of fetal macrosomia in diabetic pregnancy with operative delivery and birth trauma and subsequent obesity and diabetes during the offspring's adulthood. Screened gravidas found to be glucose intolerant are at a lower risk of fetal macrosomia with glucose surveillance or prophylactic insulin. There is significant controversy about the criteria that should be used to treat pregnancies at risk for fetal macrosomia.

At some sites, AF insulin concentration is used as a criterion for treatment of glucose-intolerant gravidas to prevent diabetic fetopathy.2 Third-trimester AF insulin concentration falls with maternal insulin treatment, and the incidence of diabetic fetopathy appears to be reduced.

The role of fetal insulin in early fetal development and its association with maternal glucose intolerance is less understood. Fetal hyperglycemia before 20 weeks' gestation appears to stimulate fetal insulin release only in diabetic pregnancy.1,10 Quantification of AF insulin in the first half of pregnancy has not been reliable. Prior studies of AF insulin from early stages of pregnancy have been hampered by assays that were relatively insensitive and had unacceptably high coefficients of variation at lower concentration ranges. Studies of earlier, solid phase radioimmunoassays of insulin in AF, with an overall coefficient of variation of 13%, did not report assay performance for insulin concentrations before 20 weeks.2

More recent studies6,11 have used another solid phase radioimmunoassay (Insulin RIA 100; Pharmacia, Inc., Freiburg, Germany) to measure insulin in AF. Overall interassay coefficient of variation in the concentration range of 1–10 μIU/mL was 15% in one study.11 Another study identified a lower limit of detection of 0.35 μIU/mL for the assay and coefficients of variation from 4.9% among the higher concentrations to 32% at concentrations below 0.5 μIU/mL.6 The range of measured insulin concentrations in the latter study was 0.01 to 6.68 μIU/mL, but 14% of stored samples had no measurable insulin.

Because of the potential utility of a high-sensitivity insulin assay for examining both physiologic and clinical questions in human diabetic pregnancy, we studied a new immunochemiluminomimetric insulin assay already shown to have a lower detection limit of 0.03 μIU/mL in serum. We also sought to examine the stability of insulin and the assay in the AF matrix after freezing and storage. In contrast to earlier findings, this assay measured insulin in all samples (collected at 14–20 weeks' gestation) at concentrations an order of magnitude greater than the lower limit of assay detection. The association of insulin concentration with gestational age was confirmed; log10 insulin values expressed in gestational age–independent MoMs had a consistent Gaussian distribution, even at the extremes of measured concentrations.

The assay and the AF matrix appear to be stable through one freeze-thaw cycle and are not measurably affected by storage for 14 days at 4C or for 70 days at −20C. This degree of stability suggests that the assay can be used economically, in batched samples, for clinical studies.

Back to Top | Article Outline


1. Reiher H, Fuhrmann K, Noack S, Woltanski K, Jutzi E, Hahn H, et al. Age-dependent insulin secretion of the endocrine pancreas in vitro from fetuses of diabetic and nondiabetic patients. Diabetes Care 1983;6:446–51.
2. Weiss P, Pürstner P, Winter R, Lichtenegger W. Insulin levels in amniotic fluid of normal and abnormal pregnancies. Obstet Gynecol 1984;63:371–5.
3. Weiss P, Hofman H, Winter R, Pürstner P, Lichtenegger W. Gestational diabetes and screening during pregnancy. Obstet Gynecol 1984;63:776–80.
4. Hoegsberg B, Gruppuso PA, Coustan DR. Hyperinsulinemia in macrosomic infants of nondiabetic mothers. Diabetes Care 1993;16:32–6.
5. Star J, Canick J, Palomaki G, Carpenter M, Saller D, Sung C, et al. The relationship between second-trimester amniotic fluid insulin and glucose levels and subsequent gestational diabetes. Prenat Diagn 1997;17:149–54.
6. Carpenter MW, Canick JA, Star J, Carr SR, Burke ME, Shahinian K. Fetal hyperinsulinism at 14–20 weeks and subsequent gestational diabetes. Obstet Gynecol 1996;87:89–93.
7. Allauzen S, Joly S, Granier C, Molina F, Bouix O, Pan B, et al. Immunoanalysis of human insulin using monoclonal antibodies reveals antigenicity of evolutionarily conserved residues. Mol Immunol 1995;32:27–36.
8. Wide L, Porath J. Radioimmunoassay of proteins with the use of Sephadex-coupled antibodies. Biochim Biophys Acta 1966;130:257–60.
9. D'Agostino RB, Belanger A, D'Agostino RB Jr. A suggestion for using powerful and informative tests of normality. Am Statistician 1990;44:316–21.
10. Adam P, Teramo K, Raiha N, Gitlin D, Schwartz R. Human fetal insulin metabolism early in gestation. Response to acute elevation of the fetal glucose concentration and placental transfer of human insulin-I-I31. Diabetes 1969;18:409–16.
11. Crombach G, Mannerschmidt C, Schmitz-Rockerath B, Herrmann F, Seibolds M, Mies R, et al. Relationship between amniotic fluid insulin and maternal blood glucose concentrations in patients with carbohydrate intolerance during pregnancy. J Perinat Med 1996;24:77–84.
© 1999 The American College of Obstetricians and Gynecologists