Key words polycystic ovarian syndrome - impaired glucose tolerance - C-peptide - insulin
Introduction
Polycystic ovary syndrome (PCOS) is the most common endocrinological cause of infertility,
menstrual disorders and hirsutism in women of child-bearing age [1 ]. Hyperandrogenaemia as well as insulin resistance with a compensatory hyperinsulinaemia
that can develop into type 2 diabetes mellitus (DM) [2 ], [3 ] are the central pathomechanisms of PCOS. Although the presence of insulin resistance
is not necessary for a diagnosis of PCOS, it is clearly apparent that insulin resistance
plays a significant role in PCO syndrome [4 ]. The prevalence of insulin resistance in PCOS lies in the range of 50–70 % [2 ], [5 ], [6 ], [7 ] and occurs independently of obesity [8 ]. Slim women [9 ] and women for whom PCOS has been diagnosed according to the Rotterdam criteria,
appear to have less pronounced insulin resistance [10 ].
Women with PCOS carry a higher risk to develop an impaired glucose tolerance [IGT]
and also type 2 DM [4 ]. Impaired glucose tolerance is defined by 2-hour values of > 140 mg/dL (7.8 mmol/L)
and < 200 mg/dL (11.0 mmol/L) in the oral glucose tolerance test (OGTT) with 75 g
glucose [11 ]. In an American study up to 31.3 % of patients with PCOS had impaired glucose tolerance
and 7.5 % had type 2 DM, compared with 14 % and 0 % in an age- and weight-matched
control group without PCOS [12 ]. In addition, women with PCOS develop an impaired glucose metabolism earlier, and
their IGT also appears to progress more rapidly to type 2 DM [13 ].
IGT is also clinically relevant and its early detection and therapy improves the long-term
outcome [14 ]. It has been demonstrated in one study that IGT increases the risk for cardiovascular
diseases, mortality and type 2 DM [15 ].
The International Diabetes Federation has classified PCOS as a significant but immutable
risk factor that is associated with type 2 DM [16 ]. To date there are no long-term studies with robust results on IGT, type 2 DM and
cardiovascular diseases in PCOS but rather only studies with surrogate parameters.
There are hardly any longitudinal data on how the clinical and endocrinological symptoms
of PCO syndrome change over a longer time period. Also no studies have yet addressed
how pronounced the pre-diabetic metabolic situation is under everyday conditions in
women with PCOS compared to a control group. This was the aim of the present analysis.
Study Participants and Methods
Study Participants and Methods
Study collective
In the framework of the LIPCOS pilot study (L ifestyle I ntervention for Patients with Polycystic Ovary Syndrome [PCOS ]), 403 patients with oligo-amenorrhoea and/or hyperandrogenaemia were identified
from a large infertility database and requested to complete and return a questionnaire.
At the same time they were invited to participate in the prospective LIPCOS main study.
Further recruitment for the prospective LIPCOS main study took place through the outpatient
unit of the gynaecology department of Munich Technical University (director: Prof.
Marion Kiechle) in the “Klinikum rechts der Isar”, referring gynaecologists in the
Munich urban area, and the PCOS self-help groups in Munich.
The study was approved by the ethics committee of Munich Technical University (TUM).
Details of the conduct of the study have already been published [17 ], and are briefly delineated below.
Inclusion criteria
Patients with spontaneous (not post-pill) oligo-/amenorrhoea and/or clinical or biochemical
hyperandrogenemia (acne, hirsutism) were eligible if they fulfilled two of the following
three criteria according to the Rotterdam criteria of 2003 [5 ]:
anovulation,
hyperandrogenaemia, and
polycystic ovaries.
Oligomenorrhoea was defined as cycle duration of > 35 days and amenorrhoea as > 90
days. Eumenorrhoeic women with corresponding age and BMI from a cohort of the Institute
for Nutritional Medicine (director: Prof. Dr. Hans Hauner) served as control group.
Exclusion criteria
Patients on any medications that act on the hypothalamic-pituitary-gonadal axis, such
as hormonal contraceptives, oestrogens or progestins for hormone therapy, endocrine
therapy after breast cancer or GnRH analogues for endometriosis were excluded as well
as pregnant or breast-feeding women. Patients with hyperandrogenemia or oligomenorrhoea
due to other known endocrine diseases such as androgen-producing tumours, adrenal
hyperplasia, primary hypothalamic amenorrhoea or premature ovarian failure as well
as prolactinomas were also excluded from participation in the study.
Course of the study
After signing informed consent, a structured interview was carried out with all participants,
blood samples were taken and a vaginal ultrasound scan with ovary score was performed
[18 ]. All participants were also offered the standardised test meal study on a subsequent
study day.
Standardised test meal
For this part of the study, the participants presented for the 3-hour test in the
morning after a 10-hour fasting period. After placement of an indwelling venous cannula
in an antecubital vein, a blood sample was taken to determine the baseline values
of haemoglobin A1c (HbA1c ), glucose, insulin and C-peptide. The test consisted of four blood samplings at the
time points 0, 60, 120 and 180 minutes.
Test meal
Following the baseline blood sampling, each participant received a standardised, carbohydrate-rich
test meal. The meal was to be consumed within 10 minutes. It consisted of a 50 g white
flour bread roll, 25 g of jam as well as 10 g of butter and comprised 62 % carbohydrates,
32 % fat and 6 % protein or, respectively, 42 g of carbohydrates, 9 g of fat and 3.8 g
of protein, amounting to a total calorie count of 267 kcal.
Analyses
The samples for insulin and C-peptide (each 4.5 mL whole blood) were placed in small
plastic tubes that contained 500 µL of a mixture of 1.5 g of ethylenediamine tetraacetate
(EDTA) in 100 mL NaCl. They were stored at 4 °C. The samples for glucose were placed
in blood sampling tubes containing EDTA/sodium fluoride (NaF) and also stored in the
cold until centrifugation. HbA1c was taken up from EDTA-haemogram tubes and analysed by the Institute of Clinical
Chemistry at the “Klinikum rechts der Isar”.
After the test, the cooled samples were centrifuged at 3000 rpm for 15 minutes at
4 °C (Hettich Rotixa/P centrifuge, Tuttlingen, Germany). The separated plasma was
stored at − 26 °C until analysis. All samples from each patient were analysed at least
twice.
Insulin measurement
Insulin was measured using a radioimmunoassay (RIA) from Siemens Medical Solutions
Diagnostics (Los Angeles, California, USA) with < 20 % cross-reactivity to proinsulin
and subsequently with a gamma counter (type 1470 Wizard, Wallac, Freiburg, Germany).
The emitted radioactivity was recorded in counts per minute (cpm). Measured cpm values
were read as percentage binding and the hormone concentration originally present in
the plasma was calculated on the basis of the calibration values.
C-peptide measurement
The concentration of C-peptide in plasma was determined with the IRMA-CPEP test (CIS
Biointernational, Gif-sur-Yvette Cedex, France).
The C-peptide values of the samples were directly read off the standard curve. The
intra- and interassay variation coefficients amounted to 4.5 and 6.4 %.
Glucose measurement
Glucose was determined photometrically by means of the hexokinase method (Glucose-HK-Test
[100 + 1], Greiner Diagnostic GmbH, Bahlingen, Germany).
Calculations
The baseline insulin sensitivity was determined by the Homeostasis Model Assessment
Insulin Resistance (HOMA-IR) Index [19 ]. HOMA-IR was calculated as [fasting glucose (mg/dL) × fasting insulin (µU/mL)] ÷ 405.
The area under the curve (AUC) was calculated as Δ-AUC according to the trapezoid
method [20 ].
Statistics
Data analyses were carried out with the programmes SPSS and Microsoft Office Excel
for Windows in cooperation with the Institute for Medicinal Statistics and Epidemiology
(IMSE) of the TU Munich. Continuous variables are described with mean values (MV)
and standard deviation (± SD). Significance was tested by means of the Mann-Whitney
U test. Categorial values are described with absolute and relative frequencies and
were tested for significance by means of Fisherʼs exact test. The level of significance
was set at p < 0.05 [17 ], [21 ].
Results
Altogether 72 participants were recruited into the LIPCOS main study in the period
from 15.12.2008 to 24. 03. 2011 and invited to the test. 41 of the participants (PCOS)
and 68 BMI- and age-matched eumenorrhoeic controls (C) consumed the standardised,
carbohydrate-rich test meal and completed the 3-hour test.
Baseline characteristics
The baseline characteristics of both groups are listed in [Table 1 ]. HOMA-IR in the participants with PCOS was higher than that in controls but the
difference was not statistically significant (0.67 ± 0.95 vs. 0.45 ± 0.66; p = 0.144).
The HbA1c value in participants with PCOS was significantly higher compared to that in controls
(5.20 ± 0.29 [n = 40] vs. 4.98 ± 0.49 % [n = 58]; p = 0.016).
PCOS
C
p
Age (years)
33.61 ± 8.79
34.77 ± 9.49
n. s.
Weight (kg)
72.80 ± 17.18
71.01 ± 15.80
n. s.
BMI (kg/m2 )
25.68 ± 6.31
25.06 ± 5.38
n. s.
HOMA-IR
0.67 ± 0.95
0.45 ± 0.66
n. s.
HbA1c (%)
5.20 ± 0.29
4.98 ± 0.49
0.016
Baseline values and AUC
Baseline glucose of participants with PCOS was significantly elevated compared to
controls (92.9 ± 10.3 [PCOS] vs. 85.1 ± 9.4 mg/dl [C]; p < 0.001).
The area under the curve (Δ-AUC) for glucose in the controls was not significantly
higher than that for the participants with PCOS (1005.1 ± 2028.6 [PCOS] vs. 1127.7 ± 1956.0 mg/dL
× 180 min [C]; p = 0.755) ([Figs. 1 ] and [2 ]).
Fig. 1 Glucose, C-peptide and insulin courses after the carbohydrate-rich standardised
test meal in PCOS patients and age- and BMI-matched controls. * p = 0.055
Fig. 2 AUC for glucose, C-peptide and insulin for participants with PCOS and the age- and
BMI-matched controls. * p < 0.05
Baseline C-peptide in participants with PCOS amounted to 0.6 ± 0.3 pmol/L whereas
that of the controls were significantly lower with 0.5 ± 0.2 pmol/L (p = 0.019).
The average Δ-AUC for C-peptide in the participants with PCOS was significantly higher
than that for the control group (145.5 ± 68.4 [PCOS] vs. 115.3 ± 65.2 pmol/L × 180 min
[C]; p = 0.023) ([Figs. 1 ] and [2 ]).
Baseline values for the two groups were almost identical (2.8 ± 3.7 [PCOS] vs. 2.1 ± 2.9 µU/mL
[C]; p = 0.237).
Accordingly, the Δ-AUCs for insulin in both groups were almost the same (1685.8 ± 1248.3
[PCOS] vs. 1657.0 ± 1458.3 µU/mL × 180 min [C]; p = 0.916) ([Figs. 1 ] and [2 ]).
Postprandial values
The postprandial 60-minute glucose values for both groups were not significantly different
(109.2 ± 22.2 [PCOS] vs. 101.9 ± 21.0 mg/dL [C]; p = 0.089). The 120-minute elevation
for the PCOS patients was of borderline significance (95.4 ± 18.4 [PCOS] vs. 88.7 ± 17.0 mg/dL
[C]; p = 0.055). After 180 minutes the glucose in the PCOS participants was significantly
higher with 88.7 ± 12,5 mg/dL vs. controls (81.8 ± 11.8 mg/dL, p = 0.005) ([Fig. 1 ]).
In the postprandial period, after 60 minutes C-peptide increased to 2.0 ± 0.8 pmol/L
in the PCOS group and to 1.6 ± 0.7 pmol/L (p = 0.007) in controls. The increase in
the PCOS group was still statistically significant after 2 hours (1.6 ± 0.8 [PCOS]
vs. 1.2 ± 0.7 pmol/L [C]; p = 0.022). After 3 hours the higher C-peptide value for
the participants with PCOS was no longer statistically significant compared to controls
(0.9 ± 0.6 [PCOS] vs. 0.8 ± 0.5 pmol/L [C]; p = 0.093) ([Fig. 1 ]).
The postprandial insulin values increased almost in parallel up to the time point
60 minutes (22.1 ± 15.8 [PCOS] vs. 20.5 ± 16.6 µU/mL [C]; p = 0.607), and then declined
almost identically (120 minutes: 11.2 ± 11.2 [PCOS] vs. 10.7 ± 14.6 µU/mL [C]; p = 0.836).
Also after 180 minutes, the insulin values of both groups had declined further but
still almost identically (3.6 ± 4.9 [PCOS] vs. 3.2 ± 5.7 µU/mL [C]; p = 0.735) ([Fig. 1 ]).
IFG und DM
An impaired fasting glucose [IFG] with baseline glucose values > 100 mg/dL was detected
in 17.07 % (7/41) of the PCOS patients and 5.88 % (4/68) of the controls. None of
the patients had baseline glucose levels > 126 mg/dL and thus there was no evidence
for diabetes mellitus (DM) [21 ].
Discussion
Previous studies that examined diabetological parameters in PCOS patients used either
the euglycaemic hyperinsulinaemic clamp method, that represents the gold standard
test for determining insulin sensitivity [22 ], or the OGTT [23 ], [24 ], [25 ]. In this study we specifically decided in favour of the standardised carbohydrate
test meal [26 ] in order to generate a submaximal, “more physiological” insulin stimulation, similar
to that to be expected also in everyday conditions. The 75 g glucose in the OGTT leads
to a maximal insulin stimulation whereas, by comparison, the standardised carbohydrate-rich
test meal contains 42 g carbohydrates that still have to be degraded.
Due to test duration of three hours instead of the two hours in the OGTT, we were
able to verify the decline in the measured values more precisely with our test procedure.
Because of this strategy, the results presented here can only be compared to a limited
extent with those from studies based on the OGTT. However, there are studies with
similar objectives that can be compared with our work [2 ], [27 ], [28 ]. For the participants, the nutritional medicine part of the study meant an additional
time load of about four hours for the visit on a subsequent day. This could not be
arranged in any other way for the participants. For this reason only 41 of the total
of 72 participants did in fact partake in the standardised, carbohydrate-rich test
meal study.
In this study, patients with PCOS exhibited higher baseline glucose values and identical
peripheral insulin concentrations compared to controls. We also detected a formally
higher insulin resistance in the patients with PCOS. Strikingly in patients with PCOS,
higher C-peptide concentrations both at baseline and postprandial existed as indication
for an increased insulin secretion. This could mean that the insulin clearance in
these patients (uptake of insulin by the liver from the portal vein system) is higher
than in controls.
Glucose tolerance in patients with PCOS was first investigated systematically in 1987
by Dunaif et al. [2 ], [27 ]. Obese and slim patients with PCOS were compared both with obese and slim ovulatory
hyperandrogenic patients and with age- and weight-matched control subjects of the
same gender. After administration of 40 g/m2 glucose orally, insulin values in obese PCOS patients increased significantly over
a period of 120 minutes compared with those of obese ovulatory hyperandrogenic women
and controls. Also, in the groups of slim patients those with PCOS exhibited significantly
higher insulin values in comparison to the other two groups. The glucose levels were
significantly increased only in the obese PCOS patients from 30 minutes onward after
administration of the glucose. The conclusion from these results was that hyperinsulinaemia
is a feature of PCOS and is not dependent on hyperandrogenaemia.
In our studies the glucose levels were already elevated in the fasting state and also
remained significantly elevated in the postprandial phase. Worthy of note is that
the glucose values in the study of Dunaif et al. [2 ], [27 ] increased to about 150 mg/dL in obese patients with PCOS and to about 125 mg/dL
in controls. In our study, however, the glucose levels rose to 109 mg/dL in the patients
with PCOS and to 101 mg/dL in the controls. This could be explained by the lower amount
of glucose administered in our study since with a glucose dose of 40 g/m2 most of the patients probably received over 60 g glucose whereas in our study merely
42 g carbohydrate were administered.
Compared to the study of Dunaif et al. [2 ], [27 ], the insulin values in our PCOS patients were not significantly increased. Due to
its higher glucose load in comparison to the standardised carbohydrate-rich test meal
used in the present study, the OGTT exerts a stronger effect with regard to the insulin
increase. This is useful for an interpretation of the results. However, the test meal
has a higher everyday relevance since a meal consisting of a bread roll with butter
and jam would be taken by a great many more patients in comparison to a drink with
75 g glucose.
An Indian case-control study by Kulshreshtha et al. [24 ] examined the glucose and insulin values after administration of an OGTT in 285 patients
with PCOS and in 27 slim controls (C) without diabetes in the family history. 62 %
of the PCOS patients had a normal glucose tolerance (NGT), 14 % elevated fasting glucose
values (impaired fasting glucose [IFG]), 17 % had IGT and 7 % type 2 DM. In this study
the glucose values of those PCOS patients with NGT were not significantly increased
in comparison to the values of controls (glucose 0 h: 84.8 ± 10.9 mg/dL [NGT-PCOS]
vs. 88.2 ± 7.2 [C]; glucose 1 h: 116.9 ± 26.2 mg/dL [NGT-PCOS] vs. 115.5 ± 27.5 [C];
glucose 2 h: 102.0 ± 18.2 mg/dL [NGT-PCOS] vs. 91.8 ± 20.5 [C]). However, baseline
and postprandial insulin values were significantly elevated compared to controls (insulin
0 h: 5.8 ± 11 [NGT-PCOS] vs. 15.0 ± 15.4 mIU/L [C]; insulin 1 h: 32.7 ± 26.5 [NGT-PCOS]
vs. 98.8 ± 81.8 mIU/L [C]; insulin 2 h: 14.6 ± 9.6 [NGT-PCOS] vs. 62.9 ± 49.3 mIU/L
[C]). Also the HOMA-IR values were significantly elevated in the PCOS patients with
normal glucose tolerance (3.1 ± 3.0 [NGT PCOS] vs. 1.2 ± 0.2 [C]).
Since the majority of our PCOS patients had a normal glucose tolerance, our results
can be compared with the results of the patients with normal glucose tolerance and
of the controls from the study of Kulshreshtha et al. [24 ]. With regard to BMI, the LIPCOS patients are comparable with the patients in the
Indian study (25.68 ± 6.31 [LIPCOS] vs. 26.5 ± 5.7 kg/m2 [Kulshreshtha et al.]), however, our patients are markedly older (34.77 ± 9.49 [LIPCOS]
vs. 22.6 ± 5.6 years [Kulshreshtha et al.]). In the Indian study the controls were
not matched for BMI and age as was done in LIPCOS, thus their characteristics are
markedly different (BMI 25.06 ± 5.38 [LIPCOS] vs. 19.7 ± 2.6 kg/m2 [Kulshreshtha et al.] and age 34.77 ± 9.49 [LIPCOS] vs. 22.8 ± 4.5 years [Kulshreshtha
et al.]).
In the Indian study fewer PCOS patients had impaired fasting glucose [IFG] than in
LIPCOS (14 vs. 17.07 %). Whereas in the Indian study the glucose values of PCOS and
control were almost identical, in the present work the glucose values of the PCOS
patients are significantly increased compared with controls. The baseline glucose
values of PCOS patients with normal glucose tolerance (NGT) in the study of Kulshreshtha
et al. [24 ] were even lower than that of the controls. The glucose values of PCOS patients with
IFG were, however, also significantly increased vs. controls, similar to LIPCOS. Comparability
is, however, limited since the baseline glucose value of the Indian controls already
shows that this was a group with impaired glucose tolerance [IGT] (108.3 mg/dL [Kulshreshtha
et al.]; 92.88 mg/dL [LIPCOS]). Accordingly, all postprandial glucose values were
markedly higher than those in LIPCOS. In the Indian study the insulin values were
significantly different between the PCOS patients and controls whereas in our investigation
there were no differences between the two groups in this regard. It is also striking
that the baseline insulin values in the present study are markedly lower both for
the patients with PCOS and for the controls (2.80 ± 3.66 [PCOS] and 2.05 ± 2.87 µU/mL
[C] in LIPCOS vs. 5.8 ± 1.1 [PCOS] vs. 15.0 ± 15.4 mIU/L [C] in Kulshreshtha et al.).
Since the BMI values of both PCOS collectives are comparable, other reasons must be
taken into consideration. Beside methodical reasons like using a different kind of
insulin-assay a possible cause for these different insulin values could be the ethnic
differences between the two patient collectives. Verification of this possible ethnic
difference was the aim of the study by Mohan et al. [29 ], who compared the insulin values of Indians and Europeans with type 2 DM and controls.
It was noticed that not only in Indian patients with type 2 DM but also in Indian
controls both the baseline insulin value and the insulin response were significantly
higher in comparison to those of European patients and the European controls. The
authors concluded that ethnic differences could have contributed to the differing
values for the control groups.
The majority of insulin secreted from the pancreas is absorbed by the liver (insulin
clearance). Peripheral insulin concentrations are thus not suitable to evaluate the
secretion. C-peptide, consisting of 31 amino acids, as a polypeptide binds the two
chains of proinsulin and is enzymatically cleaved during transformation to insulin.
Together with insulin, C-peptide is emitted into the blood from the pancreas and can
be measured as a diabetic parameter [30 ]. Studies in the past few years have shown that C-peptide possesses specific binding
to cell membranes, influences numerous cell signalling pathways and has a protective
role in diabetic complications [31 ]. So far there are only few studies that included C-peptide in the investigation
of diabetic parameters in cases of PCOS. In our study not only baseline but also postprandial
C-peptide values were significantly higher in PCOS patients than in the controls.
In a study by Maciejewska-Jeske et al. [32 ] the baseline glucose and C-peptide levels of 5 overweight (27.2 years, BMI 27.3 kg/m2 ) and 60 obese patients with PCOS (26.2 years, BMI 35 kg/m2 ) were compared with those of 10 controls (28.8 years, BMI 21.2 kg/m2 ). In the group of overweight women with PCOS, the C-peptide concentration was somewhat
higher than that in the obese women with PCOS (1.39 and 1.31 nmol/L), but lower than
that in the controls (1.62 nmol/L). In the control group, C-peptide concentration
correlated negatively with glucose (R = − 0.71; p < 0.05), whereas in the group of
overweight women with PCOS there was a positive correlation between these two values
(R = 0.90; p < 0.05). Increased insulin clearance may explain these divergent correlations.
Due to the use of different assays and standard curves, the absolute values from the
study by Maciejewska-Jeske et al. [32 ] probably cannot be compared with the results of the LIPCOS study. However, on the
basis of the relative values, a trend can be seen that in patients with PCOS higher
glucose levels are associated with higher C-peptide concentrations.
In agreement with the present work, several studies have found significantly elevated
glucose values in patients with PCOS in comparison to weight- and age-matched controls.
Elevated insulin values, as have been found in a few studies on patients with PCOS,
could not be detected in our study. However, we did find significantly increased baseline
and postprandial C-peptide and HbA1c levels in patients with PCOS in comparison to controls.
In the present work we have shown for the first time that patients with PCOS have
a higher fasting insulin resistance than controls, not only with an OGTT as previously
published [2 ], [27 ], but also after administration of a standardised test meal. This insulin resistance
did not increase further in the postprandial phase despite of higher stimulated C-peptide
levels. This suggests a mechanism that is linked to a higher hepatic clearance in
PCOS patients [33 ], [34 ], [35 ]. However, the small number in our study is a limitation and our results need to
be replicated.
Petersen et al. [36 ] have shown that the insulin sensitivity in patients with coronary artery disease
could be improved by dieting in contrast to only physical training. Improved insulin
sensitivity was associated with a reduction in abdominal fat, hip circumference and
body weight. Whether a diet can affect the insulin sensitivity or the insulin clearance
in PCOS patients or whether there are differences in this clinical entity is, as far
as we are aware, not known and has not yet been investigated.
