Key words
infertility - polycystic ovary syndrome - ovarian hormone - assisted reproductive
techniques
Introduction
Ovarian hyperstimulation syndrome (OHSS) is a serious and completely iatrogenic complication
of controlled ovarian hyperstimulation [1]. It is an important issue in the field of IVF (in vitro fertilization). However,
the exact pathogenesis of the OHSS remains less clear, vascular permeability mediated
by human chorionic gonadotropin (hCG) appears to be the most probable mediator. OHSS
is categorized into four classes, including mild, moderate, severe, and critical forms
of the syndrome on the basis of the severity of symptoms, signs, and laboratory parameters.
Mild OHSS is characterized by the enlargement of bilateral ovaries (up to 8 cm) with
abdominal bloating and mild abdominal pain. The moderate form of OHSS is described
by the enlargement of the ovaries (up to 12 cm), as well as ultrasound finding of
ascites. The severe form of OHSS is defined by the observation of large ovarian cysts
(>12 × 12 cm), clinical manifestation of ascites with or without hydrothorax, with
abnormality findings like sodium, potassium, and osmolarity serum leading to decreased
urine output and hypovolemic shock.. Critical OHSS is characterized when there is
severe ascites on ultrasound examination or hydrothorax, hematocrit of over 55%, WBC
over 25 000/ml, oliguria or anuria, creatinine ≥1.6 mg/dl, creatinine clearance less
than 50 ml/min, thromboembolism, or acute respiratory distress syndrome [1]
[2]. The reported incidence of the severe and mild form of OHSS ranges from 0.5–33%,
respectively [2]
[3]
[4]. Polycystic ovary syndrome (PCOS) appeared to be the major risk factor for OHSS
in many studies [2]
[5]
[6]
[7]. PCOS increases the risk of OHSS because of the hyper-response to ovarian stimulation,
since too many antral follicles are already present at the beginning of the cycle.
Hyperinsulinemia and hyperandrogenism are identifying features of PCOS and both promote
early folliculogenesis and frequently a multi-folliclular response following the ovulation
induction [2]
[8]
[9]. Hyperinsulinemia and hyperandrogenism are causes rather than consequences of polycystic
ovarian morphology (PCOM), therefore high insulin and/or testosterone levels may increase
the risk of OHSS by inducing PCOM. Insulin-like growth factor 1 (IGF-1) and insulin
can stimulate vascular endothelial growth factor (VEGF) production by luteinized granulosa
cells in vitro [10], but this is insufficient evidence to suggest that insulin resistance is directly
responsible for OHSS. OHSS is prevalent among PCOS infertile women, however, it is
also observed among infertile non-PCOS women. Therefore, we assessed the probability
of the presence of hyperinsulinemia and hyperandrogenemia in the incidence of OHSS
in the non-PCOS women undergoing ART cycles and to evaluate if a threshold level is
present.
Subjects and Methods
This prospective follow-up study was conducted on 144 infertile and non-PCOS women
with regular menstrual cycle undergoing IVF or intracytoplasmic sperm injection (ICSI)
cycles from December 2012 to December 2014. Exclusion criteria were patients with
PCOS diagnosis based on Rotterdam criteria [11], previous ovary surgery, ovarian insufficiency, endometriosis, some medical conditions
like (diabetes mellitus, thyroid dysfunction, hyperprolactinemia, Cushing’s syndrome
[12]), being smoker and use of the drugs like spironolactone, aspirin, corticosteroid,
and metformin. The selection of the participants was made from the two referral infertility
clinics at Ahvaz, Iran. The study protocol complies with Helsinki Declaration and
was approved by the Research Ethics Committee of Ahvaz Jundishapur University of Medical
Sciences. Informed consent was obtained from all participants. The blood samples of
the participants were collected on 1–3 days of their cycles after a fasting period
at least 8 h. The samples were centrifuged at 3000 g for 10 min, and the serums were stored at −20 °C until assaying. Fasting blood sugar
(FBS) level was measured by Autoanalyzer (Selectra, Flexer Model, Netherlands) using
glucose oxidase method. Testosterone was assayed using a direct, competitive chemiluminescent
immunoassay (CLIA) performed on the manufacturer’s analyzer (Liaison Analyzer, Diasorin
Inc., USA). The method for quantitative determination of insulin was a sandwich chemiluminescent
immunoassay performed on the manufacturer’s analyzer (Liaison Analyzer, Diasorin Inc.,
USA). The level of SHBG was detected by an enzyme-linked immunosorbent (ELIZA) kits
(IBL International, Hamburg, German). The free index (FAI) was determined as the total
testosterone × 100/SHBG. The IR was calculated using the homeostasis model assessment
(HOMA) [HOMA-IR=(insulin × glucose)/22.5] [13]. Anti-Müllerian hormone (AMH) was assayed using Elecsys Cobas e 411, Roche Kiet,
Germany. Patients underwent vaginal ultrasonography scan for polycystic ovary morphology
(PCOM) and antral follicle count (AFC) in 1–3 days of their cycles and then based
on the indication of the treatment underwent long protocol with the use of GnRH agonist
(bucerelin) and gonadotropins. The number of preovulatory follicles was recorded.
Triggering of the ovulation was performed by hCG and oocytes retrieved were recorded,
then the patients were followed-up for the signs and symptoms of OHSS (mild, moderate,
and severe), then they were divided into two groups with and without OHSS (mild, moderate,
and severe) for the analysis. We used 10 000 IU hCG for trigger of ovulation, but
for the patients with probability OHSS based on the number of preovulatory follicels
and the level of estradiol or rapid increased level of estradiol, we choose the coasting
method without further gonadotropin stimulation and delaying the use of hCG for 1–3
days until estradiol levels flatten or decline, and used lower dose of hCG (5000 IU).
The total number of ART cycles was 1225 and finally of those, 330 patients developed
OHSS (27%, including 20, 5, and 2% for mild, moderate and severe OHSS), respectively,
then 66 non-PCOS patients with mild, moderate or severe OHSS were used for final analysis.
The data were analyzed using Statistical Package for the Social Sciences (SPSS, version
11.5, Chicago, IL, USA). To compare quantitative variables, independent sample t-tests were applied. Chi-square test was used to test dichotomous variables between
groups. The forward stepwise multiple logistic regression model was constructed for
OHSS. We used the absence or presence of OHSS as the dependent variable. The univariate
analyses were conducted on the variables including age, BMI, HOMA-IR, FAI, AMH, AFC,
preovulatory follicles, and oocytes retrieved, and dichotomous variable of PCOM. Then,
the variables assessed significant in univariate regression analysis were entered
as independent variables in multivariable logistic regression analysis. The receiver
operating characteristics (ROC) curve was used to identify the best cut-points value
for those variables that were independent predictors of OHSS development in non-PCOS
patients.
Results
A total of 144 patients were followed-up in this study. The incidence of mild, moderate,
and severe OHSS in two centers were 20, 5, and 2%, respectively. There was no significant
difference between patients who developed OHSS and those who did not, regarding age
and oocytes retrieved using univariate logistic regression analyses ([Table 1]). Only two patients were admitted with sever OHSS (0.14%). Patients with OHSS had
higher chance to have ovaries with polycystic morphology (74%), about three times
more than patients who did not develop OHSS (29%) (p<0.001) ([Table 1]). In addition, univariate logistic regression analysis showed that mean scores of
BMI, FAI, HOMA-IR, AMH, AFC, PCOM and preovulatory follicles were significantly higher
in OHSS compared with non-OHSS patients ([Table 1]). Of the 9 variables: BMI, HOMA-IR, FAI, AFC, AMH, PCOM, and preovulatory follicles
were risk factors, while the age and retrieved oocytes were not ([Table 1]). The 7 variables that provided significant in the univariate analyses were chosen
as independent variables included in the multivariable logistic regression analysis,
as a result, a total of 5 risk factors, BMI, HOMA-IR, FAI, PCOM, and preovulatory
follicles entered the equation. The maximum contribution was HOMA-IR followed by PCOM,
FAI, preovulatory follicles, and BMI ([Table 2]). To assess the optimal cut-points for HOMA-IR,FAI, AFC, and AMH to predict the
development of OHSS in non-PCOS patients who undergo ART, ROC curves were obtained
The best cut-points for HOMA-IR, FAI, AMH, AFC and preovulatry follicles were 2.36
(AUC=0.78, sensitivity=75%, specificity=70%), 3.9 (AUC=0.67, sensitivity=54%, specificity=80%),
3.3 ng/ml (AUC=0.79, sensitivity=80%, specificity=71%), 8 (AUC=0.85, sensitivity=86%,
specificity=70%), and 10 (AUC=0.86, sensitivity=80%, specificity=88%), respectively
([Figs. 1] and [2]).
Fig. 1 The best cut-points for HOMA-IR and FAI were 2.36 (AUC=0.78, sensitivity=75%, specificity=70%),
and 3.9 (AUC=0.67, sensitivity=54%, specificity=84%), respectively.
Fig. 2 The best cut-points for AMH and AFC were 3.3 ng/ml (AUC=0.79, sensitivity=80%, specificity=71%),
and 8 (AUC=0.85, sensitivity=86%, specificity=70%), respectively.
Table 1 Risk factors related to ovarian hyperstimulation syndrome in 144 patients: univariate
analysis.
|
Numerical Variable
|
OHSS (n=66)
|
Non-OHSS (n=78)
|
95% CI
|
p-Value
|
|
Age (years)
|
31.07±5.63
|
29.85±5.68
|
0.980–1.102
|
0.827
|
|
BMI (kg/m2)
|
26.21±2.20
|
23.21±1.96
|
1.517–2.260
|
<0.001
|
|
HOMA- IR
|
3.21±1.6
|
1.58±O.79
|
2.263–5.251
|
<0.001
|
|
FAI
|
5.52±4.68
|
2.43±1.53
|
1.235–1.748
|
<0.001
|
|
AMH (ng/ml)
|
6.7±5.6
|
3.1±2.9
|
1.013–1.529
|
0.012
|
|
AFC (number)
|
13.73±6.11
|
6.7±5.6
|
1.108–1.723
|
<0.001
|
|
PCOM
|
49 (74)
|
23 (29)
|
3.303–14.384
|
<0.001
|
|
Preovulatory follicles (number)
|
15.3±5.7
|
8.3±2.5
|
1.147–2.063
|
<0.001
|
|
Oocytes retrieved (number)
|
14.3±8.6
|
9.9±5.5
|
0.991–1.208
|
0.06
|
Data are expressed as means±standard deviation or number (percentage). OHSS: Ovarian
hyperstimulation syndrome; BMI: Body mass index; HOMA- IR: Homeostasis model assessment-
insulin resistance; FAI: Free androgen index; AMH: Anti-Müllerian hormone; AFC: Antral
follicle count; PCOM: Polycystic ovary morphology.
Table 2 Risk factors related to ovarian hyperstimulation syndrome in 144 patients: multivariable
logistic regression analysis (Forward, LR, α=0.05).
|
Women underwent ART
|
Odds ratio (OR)
|
95% CI
|
p-Value
|
|
BMI
|
1.434
|
1.070–1.921
|
0.016
|
|
HOMA-IR
|
4.704
|
2.422–9.137
|
<0.001
|
|
FAI
|
1.613
|
1.181–2.204
|
0.003
|
|
PCOM
|
3.161
|
1.041–9.599
|
0.042
|
|
Preovulatory follicles (number)
|
1.539
|
1.147–2.063
|
0.004
|
OHSS: Ovarian hyperstimulation syndrome; ART: Assisted reproductive technology; BMI:
Body mass index; HOMA-IR: Homeostasis model assessment-insulin resistance; FAI: Free
androgen index; AMH: Anti-Müllerian hormone; AFC: Antral follicle count; PCOM: Polycystic
ovary morphology.
Discussion and Conclusions
Discussion and Conclusions
PCOS is known as a major primary risk factor for OHSS in a large number of studies
[2]
[7]
[14]
[15]. On the other hand, vascular endothelial growth factor (VEGF) is proved to be the
principal mediator in OHSS pathophysiology [16]. In addition, it has been reported that an increased expression of VEGF’s mRNA within
the hyperthecal stroma of women with PCOS may be responsible for their higher risk
of OHSS [17]. It was shown that insulin and insulin-like growth factor (IGF) promote VEGF production
in luteinized granulosa cells [18]. It has been shown that the incidence of OHSS is related to the stimulation regimens
used [3]. PCOS isolated characteristics (i. e., PCOM, hyperinsulinemia, and hyperandrogenism)
were revealed to be possible predisposing factors for OHSS. In the current study,
BMI, HOMA-IR, FAI, preovulatory follicles and the presence of PCOM retained as risk
factors for OHSS based on multivariate logistic regression analysis. In some moderate-quality
evidences, the use of insulin-sensitizing agents (i. e., metformin) decreased the
risk of OHSS in PCOS women who underwent ART [14]
[18]. They reported that in the metformin group, there was a reduction in the testosterone
concentration and in the free-androgen index [14]. Salamalekis et al., suggested that there is no association between insulin levels
or IR levels and the development of OHSS in women, with or without PCOS, undergoing
ART. The fasting glucose/insulin ratio (FGIR) was used for the evaluation of insulin
resistance (IR) in their study [19]. One possible explanation for the observed discrepancy is the difference in IR measurement
methods while in our study the HOMA-IR test was used. Considering simple surrogates
for HOMA-IR are among the best and most extensively validated, they are probably more
reliable than FGIR [20]
[21]. The other explanation is the difference in sample size, while our sample size was
1.5 times that of the aforementioned study [19]. Moreover, the age and the number of retrieved oocytes were not retained as independent
risk factors in the multivariable logistic regression analysis. Several studies evaluate
the impact of BMI on the development of OHSS and have reported contradictory results
[2]
[22]
[23]. In contrast to our finding, body weight/BMI did not appear to be a useful marker
for predicting risk of OHSS according to the study of Fiedler et al. [22]. PCOM is a criterion for identifying PCOS but not necessarily predicting severity
or presence of endocrine dysfunction [11]
[24]. However, we found that PCOM is a risk factor for OHSS in multivariable analysis.
In addition, in the current study, other risk factors including BMI, HOMA-IR, FAI,
and preovulatory follicles are identified as a risk factor for OHSS.
In our study, ROC analysis revealed that HOMA-IR, FAI, AFC, AMH, and preovulatory
follicles could be good markers to predict the OHSS’ risk in non-PCOS women undergoing
controlled ovarian hyperstimulation. The best thresholds for HOMA-IR, FAI, AFC, AMH,
and preovulatory follicles in patients were 2.36, 3.9, 8, 3.3 ng/ml, and 10, respectively.
Similar to our study, Ocal et al. have shown that the cut-off value of AMH of 3.3 ng/ml
had a sensitivity and specificity of 90% and 71%, respectively, for predicting OHSS
[25]. Also, Nardo et al. showed a cut-off value for AMH 3.5 ng/ml with a sensitivity
and specificity of 88% and 70%, respectively, in both PCOS and non-PCOS together [26]. In contrast to our study, Vembu et al. described the cut-off value for AMH 4.85 ng/ml
in non-PCOS group provided a sensitivity of 85.7% and a specificity of 89.7% [27], which is higher than the value reported in our study. This could be attributed
to a different assay. Dickerson et al., found a greater total number of follicles
at the end of ovarian stimulation during IVF in patients with greatest IR, suggesting
that the relatively higher level of insulin stimulates follicle recruitment or development.
They suggested a positive correlation of HOMA-IR levels above a level of 2.5 and a
continuous positive correlation of FAI to total ovarian follicle count following medically-assisted
reproduction (MAR) in the non-PCOS women and they assessed the effect of IR and androgen
status in non-PCOS women on the follicular outcome of MAR [8]. In consistent with our study, the number of follicles on the day of hCG administration
appears to be a good prognostic indicator for the occurrence of OHSS in women undergoing
ART [28]. Hence, it could be hypothesized that HOMA-IR and FAI status in non-PCOS women are
good markers for predicting ovarian response, including an exaggerated one (i. e.,
OHSS). Surprisingly, we noticed that the threshold level of HOMA-IR above which the
OHSS could be predicted with good sensitivity and specificity (i. e., IR>2.36) was
very close to the level above which a good ovarian response (i. e., a greater total
number of follicles) could be expected (i. e., IR>2.5) [8]. The cut-off point used for defining biochemical hyperandrogenemia in the diagnosis
of PCOS is an FAI value of 8.5 [29]. Our result demonstrated that the effect of androgen on ovarian response starts
from much lower values of FAI (i. e., 2.25–3.9) than those are used to delineate classical
hyperandrogenemia. The same pattern is described by Dickerson et al., as they showed
a continuum of effect of androgen on ovarian response, even within those not classically
demonstrated to have hyperandrogenemia [8]. As OHSS treatment is largely supportive, prevention is crucial. So it seems logical
to assess IR and androgen status even in non-PCOS women and take preventive measures
including administration of insulin sensitizing agents or use of GnRH antagonist cycles
and triggering ovulation by GnRH agonist in ART cycles.
This study has some limitations. No GnRH antagonist cycles were in group ART and participants
were Caucasian, so our results cannot be extrapolated to other ethnic groups.
In conclusion, patients with a higher value of BMI, FAI, HOMA-IR, and preovulatory
follicles and the presence of PCOM are more likely to develop OHSS, which are not
confined to PCOS patients.
