Dynamic tests of ovarian reserve: a systematic review of diagnostic accuracy

RBMOnline - Vol 18 No 5. 2009 717-734 Reproductive BioMedicine Online; www.rbmonline.com/Article/3801 on web 13 March 2009

Review Dynamic tests of ovarian reserve: a systematic review of diagnostic accuracy Abha Maheshwari graduated in 1993 from the University of Delhi, India and then completed post-graduate studies there in Obstetrics and Gynaecology in 1998. In 2000 she moved to the United Kingdom and worked in a number of different hospitals. Currently she is a subspecialty trainee and clinical lecturer in Reproductive Medicine at University of Aberdeen, Scotland. She has a long-standing interest in infertility, focusing mainly on ovarian ageing and its impact on health services.

Dr Abha Maheshwari Abha Maheshwari1,3, Ahmed Gibreel1, Siladitya Bhattacharya1, NP Johnson2 Dept of Obstetrics and Gynaecology, University of Aberdeen, Aberdeen Maternity Hospital, Foresterhill, Aberdeen AB25 2ZD; 2Dept of Obstetrics and Gynaecology, University of Auckland, National Women’s Health, Auckland Hospital, Grafton, Auckland, New Zealand 3 Correspondence: e-mail: [email protected]

1

Abstract Despite a plethora of tests of ovarian reserve, there is no perfect test to predict pregnancy. Recent evidence points that antiMüllerian hormone and antral follicle count may be better than other tests, although other tests continue to be used and form the basis of exclusion of women from fertility treatments. This systematic review concentrated on dynamic tests of ovarian reserve [clomifene citrate challenge test (CCCT), gonadotrophin-releasing hormone agonist stimulation test (GAST) and exogenous FSH ovarian reserve test (EFORT)] and assessed their predictability in terms of fertility outcomes. The study did not restrict itself to women undergoing IVF. The diagnostic odds of abnormal CCCT for non-pregnancy were 2.11 (95% confidence interval, 1.04–4.29) at FSH >10 IU/l (day 3 or 10). The diagnostic accuracy of GAST and EFORT could not be determined due to inconsistencies in the way these tests were conducted. This systematic review and meta-analysis was limited by heterogeneity in terms of the population sampled and the index and reference tests. There is an urgent need for consensus on the performance of these tests and the definition of normality, if their use is to be continued. However, given the present level of evidence, these tests should be completely abandoned. Keywords: CCCT, dynamic test, EFORT, GAST, ovarian reserve

Introduction Ovarian reserve is a term used to describe the functional potential of the ovary and reflects the number and quality of oocytes. An ideal test of ovarian reserve should predict both ability and inability to have a live-born baby with or without treatment. In an ideal world, it should also predict preservation of current levels of ovarian activity. This is particularly relevant in the present social climate, when increasing numbers of women are deferring childbearing. However, the availability of a wide range of tests of ovarian reserve at the present time suggests that there is no ideal test. There are two types of tests of ovarian reserve: static and dynamic. Static tests assess specific parameters relating to ovarian reserve at a single point in time and involve both ultrasound and biochemical parameters. Dynamic tests assess ovarian response

to exogenous stimulation. Usually this involves measurement of hormonal concentrations in a serum sample before and after stimulating the ovaries using FSH, clomifene citrate (CC) or a gonadotrophin-releasing hormone (GnRH) agonist. There are a number of existing reviews on the tests of ovarian reserve (Scott and Hofmann, 1995; Broekmans et al., 1998, 2006; Shahara et al., 1998; Bukman and Heineman, 2001). The most comprehensive review in recent years was conducted by Broekmans et al. (2006) who restricted their population to women undergoing assisted reproduction treatment. The authors of the review did not indicate that contact had been made with primary authors in order to clarify possible duplication of patients in more than one report or to ascertain cut-off values for test results. Also, no actual meta-analysis (i.e. pooling of data, similar to meta-analysis of randomized controlled trials) was attempted.

© 2009 Published by Reproductive Healthcare Ltd, Duck End Farm, Dry Drayton, Cambridge CB23 8DB, UK

717

Review - Dynamic tests of ovarian reserve - A Maheshwari et al.

Due to the sheer number of tests advocated for ovarian reserve, it is difficult to present results of a detailed systematic review of these tests in one report. Ideally, they should be grouped into: (i) ultrasound-based tests; (ii) dynamic tests; and (iii) biochemical static tests (FSH, anti-Müllerian hormone [AMH] and other static tests). This systematic review has limited its evaluation to dynamic tests of ovarian reserve. Although, more expensive, invasive and associated with side effects, dynamic tests are expected, at least theoretically, to provide a better estimate of ovarian reserve in comparison with static tests. Recently, static tests such as antral follicle count (AFC) and AMH have been shown to be better predictors (Broekmans et al., 2006). Despite this, some centres still continue to use dynamic tests for prediction of fertility outcomes and determining access to assisted reproduction treatment.

Independent searches were performed by two reviewers (AM and AG). Any decisional conflicts were resolved after discussion with SB and NJ. Once the studies for inclusion were selected, full text versions of each were read and data extraction was performed by two reviewers (AM and AG) independently. The searches were performed in April 2007 and again in April 2008.

Hence, the purpose of this review was to determine the diagnostic accuracy of three dynamic tests of ovarian reserve described in the literature, clomifene citrate challenge test (CCCT), GnRH agonist stimulation test (GAST) and exogenous FSH ovarian reserve test (EFORT), and to determine whether they have any role in the present circumstances. The review has not been limited to women undergoing assisted reproduction treatment.

Exclusion criteria

First described by Navot et al. (1987), CCCT involves the administration of 100 mg CC on days 5–9 and measurement of serum FSH on days 3 and 10. An abnormal test is defined as abnormally high FSH on day 3 and/or on day 10. First described in 1990 in two separate studies (Garcia et al., 1990; Padilla et al., 1990), Winslow et al., (1991) advocated GAST because of the dose–response nature of the results compared with the normal/abnormal results obtained with other ovarian reserve screening. GAST, though not dependent on steroid hypothalamic feedback, does depend on the pituitary production of gonadotrophins, as well as the response of the ovary to stimulation (i.e. follicle reserve). EFORT, first described by Fanchin et al. (1994), involves measurement of basal FSH and oestradiol, and the oestradiol response following FSH injection.

Materials and methods

718

The protocol for this review has previously been published (Johnson et al., 2006). Searches were performed using Ovoid Medline (1996–2007), EMBASE (1980–2007), Cochrane Library (2007:4), PubMed (1980–2007) EBM Review (1980– 2007) and CAB Abstract (1990–2007). A combination of medical subject headings (MeSH) and keywords were used to generate citations: for CCCT, keywords were CCCT or clomifene citrate challenge test; for GAST, keywords were GnRHa stimulation test, GAST, gonadotrophin agonist stimulation test; for EFORT, keywords were EFORT, exogenous FSH ovarian reserve test, exogenous follicle stimulating hormone reserve test. In addition, ovarian reserve (‘ovary’, ‘ovarian’) and dynamic test were used. Where necessary, this search strategy was adapted for use in the other electronic databases. The reference lists of all known primary and review articles were cross-checked to identify articles missed by electronic searches. There were no language restrictions.

Inclusion criteria Studies were selected in which accuracy of GAST, EFORT and CCCT (or indeed any other dynamic test of ovarian reserve) was evaluated for predicting ovarian reserve using fertility outcomes as the reference standard.

Primary studies were excluded where outcomes other than fertility outcomes were used as reference standards, e.g. concentration of hormones. Reviews were excluded. Where data from the same cases were included in more than one publication, the most recent version was selected. From each selected paper, information on study characteristics, quality and accuracy of results was extracted. Accuracy data were used to construct 2 × 2 tables (the test was considered to be positive if the result was abnormal, reflecting diminished ovarian reserve; and the test was negative if the result was normal, reflecting normal ovarian reserve. Disease positive meant absence of the desired fertility outcome (non-pregnancy or cycle cancellation); disease negative was indicated by the presence of the fertility outcome, e.g. live birth, ongoing pregnancy. All manuscripts meeting the selection criteria were assessed for methodological quality. Quality assessment involved scrutinizing study design, period of recruitment, relevant features of the population, test, reference standard and outcomes of the study. A study was considered to be of good quality if it used a prospective design, consecutive enrolment, full verification of the test result with reference standard, and had adequate test description. Meta-analysis was performed only if data from at least two studies could be combined (using same experimental and reference tests), on a similar population. Studies were pooled and sensitivity, specificity, likelihood ratio of positive and negative test results as well as diagnostic odds ratios were calculated using Meta Disc software (version 1.4) (Zamora et al., 2006). The main analysis was restricted to studies in which 2 × 2 tables of the diagnostic test results and the relevant pregnancy outcomes could be extracted. The summary receiver operating characteristic (ROC) curve was still considered the preferred method of pooling dichotomous test result from primary studies (Deeks, 2001; Khan et al., 2001). Because there were considerable differences in the cut-off concentration for an abnormal test included in the studies, this study initially tested for a high positive correlation (i.e. correlation coefficient >0.6) between their true positive rates and the false negative rates using the Spearman correlation test to determine if a summary ROC curve could be generated (Midgette et al., 1993). This study also determined, wherever possible, posttest probability using Bayes’ theorem using the formula:

RBMOnline®

Review - Dynamic tests of ovarian reserve - A Maheshwari et al.

likelihood ratio × pretest probability/[1 – pretest probability × (1 – likelihood ratio)].

Results Clomifene citrate challenge test (CCCT) A total of 158 citations were identified, of which 40 full text publications were extracted for further evaluation. Of these, 18 studies were excluded as shown in the flow chart (Figure 1). Two sets of studies (Tanbo et al., 1989, 1990; Scott et al., 1993, 1995) had overlap of patients (confirmed after correspondence with the authors), hence only the later studies (Tanbo et al., 1990; Scott et al., 1995) were included. In three studies (Corson et al., 1999; Hendriks et al., 2005a; Vladimirov et al., 2005), a 2 × 2 table could not be constructed, hence authors were contacted via e-mail or letters but replies were not received any reply. A total of 20 studies remained (Table 1). All the studies included in this review performed CCCT in a uniform manner, as originally described (Navot et al., 1987). However, the definition of an abnormal test varied among individual studies. Seven studies had a similar definition for an abnormal test (FSH >10 IU/l on day 3 or 10) (Table 1). Five studies tested women undergoing IVF, one study tested women undergoing intrauterine insemination (IUI) and the last tested women attending general infertility clinic. All of these reported pregnancy rates while three reported cycle cancellation rates.

Of three studies, which reported a cut-off value for FSH ≥10 IU/l (Table 1), two reported pregnancy as the reference standard and one reported cycle cancellation. Hence meta-analysis was possible only for the two studies reporting pregnancy rates. Two studies from the same centre (Tanbo et al., 1990, 1992) had defined an abnormal result as FSH >12 IU/l (on cycle day 3 or 10) and reported cycle cancellation and pregnancy as the reference standard. There was no overlap of patients (confirmed by authors). With any other cut-off value, there were no two studies that used the same cut-off values or compared the result with same reference standard; hence meta-analysis was not possible. Wherever authors have calculated pregnancy rate per cycle as well as pregnancy rate per woman, the latter was used in the meta-analysis. The pooled sensitivity, specificity, positive and negative likelihood ratio and diagnostic odds ratio of an abnormal CCCT to predict non-pregnancy and cycle cancellation are presented in Table 2. Data was only combined if studies were clinically homogeneous. Since none of the relevant subgroups of studies generated a sufficiently high positive correlation between their true positive and the false negative rates using the Spearman correlation test [the Spearman correlations, on threshold analysis for predicting cycle cancellation and non-pregnancy were low (0.02 and 0.02), hence ROC curve was not deemed appropriate], the analyses were conducted using likelihood and diagnostic odds ratio as

Figure 1. Study selection process for systematic review of clomifene citrate challenge test to predict fertility outcome. ART = assisted reproduction treatment; IUI = intrauterine insemination. RBMOnline®

719

720

131 women undergoing IVF. Exclusion criteria: women aged >40 years, poor visualization of ovaries, presence of an ovarian cyst of ≥20 mm and the presence of polycystic ovaries on scanning

198 women undergoing ICSI for (oligoasthenoteratozoospermia or obstructive or non-obstructive azoospermia). Inclusion criteria: at least one of the following criteria; over 35 years of age, previous removal of one ovary or a history of ovarian surgery, presence of ovarian endometrioma and previous poor response to ovarian stimulation 140 women/279 IVF cycles. Inclusion criteria: age ≥35 years with a diagnosis of unexplained infertility, any age with history of a poor response to FSH stimulation, or any age with history of intra-abdominal surgery or endometriosis. Exclusion criteria: age >38 years 142 women undergoing IVF/ICSI. Inclusion and exclusion criteria not mentioned

Ng et al. (2005)

Kahraman et al. (1997)

51 women undergoing IVF/ICSI. Inclusion and exclusion criteria not mentioned

146 women undergoing IUI or ovulation induction. Inclusion criteria: regular menstruating (23–35 days), normal ultrasound scan, normal hysterosalpingography (HSG) and seminal analysis of the partner. Exclusion criteria: not mentioned

1200 patients undergoing infertility evaluation; 588 women met inclusion criteria. Inclusion criteria: not mentioned. Exclusion criteria: patients with tubal factor, moderate or severe endometriosis, pelvic adhesive disease, or a male factor

Van der Stege et al. (2001)

Magendzo et al. (2006)

Scott et al. (1995)

van Swieten et al. (2005)

Csemiczky et al. (2002)

Population

Study

Table 1. Characteristics of included studies for clomifene citrate challenge test (CCCT).

Continued on page 721

Pregnancy

FSH >10 IU/l (combination of Pregnancy rate per cycle; day 3 or 10 or both analysed); pregnancy rate per woman abnormal test not predetermined; electrochemiluminescence in ELECYSY 2010 (Roche Diagnostics, Basel, Switzerland)

FSH >10 IU/l (day 3 or 9); RIA Cycle cancellation; (Eleesys immunoassay, Roche pregnancy per woman Diagnostic Nederland, Almere, The Netherlands)

July 1989 to July 1993 FSH >10 IU/l (day 3 or day 10); RIA (Becton-Dickinson, Orangeburg, NY)

May 2004 to Aug 2005; prospective

Period of recruitment not mentioned; consecutive and prospective recruitment Period of recruitment not mentioned; consecutive recruitment

Cycle cancellation; pregnancy rate per woman; live birth

Dose of HMG; duration of HMG; number of oocytes; pregnancy rate/ cycle; pregnancy rate/ transfer; implantation rate; multiple pregnancy rate; ongoing pregnancy rate Cycle cancellation; pregnancy rate per started cycle; ongoing pregnancy per started cycle

Reference standard

FSH >10 IU/l (day 3 or 9); assay Pregnancy per woman not mentioned

FSH >10 IU/l (before or after CC stimulation); enzyme immunometric assay (Immulite FSH, Diagnostic Products Corporation, LA, USA) FSH >10 IU/l (day 10); RIA (Farmos Group, Oulu, Finland)

Oct 1994 to Dec 1995; prospective and consecutive recruitment Jan 1995 to Dec 1998; prospective recruitment

FSH >10 IU/l (basal or stimulated); Automated Chemiluminescence ACS-180 System (Bayer Corporation, USA)

Cut-off/Assay

Jan 2002 to Nov 2003; prospective recruitment

Study design

Review - Dynamic tests of ovarian reserve - A Maheshwari et al.

RBMOnline®

Reference standard

RBMOnline®

91 women undergoing IVF. Inclusion criteria: at least one of these criteria: women ≥35 years, previous removal of one ovary or resection of both, ovarian endometriosis ≤30 mm as diagnosed by laparoscopy. Exclusion criteria: women with PCO 43 women undergoing IVF. Inclusion criteria: patients with ovulatory cycles, who previously had their first CC/HMG cycle for IVF cancelled due to a poor ovarian response. Exclusion criteria: PCO 692 women. Inclusion criteria: woman aged ≥35 years or unexplained infertility or with one ovary or a history of significant ovarian surgery or endometriosis or a history of a poor response to HMG

Tanbo et al. (1992)

107 women seeking fertility evaluation. Inclusion criteria: woman aged ≥35 years or unexplained infertility or with one ovary or a history of significant ovarian surgery or endometriosis or a history of a poor response to HMG. Exclusion criteria not mentioned

219 women seeking fertility evaluation. Inclusion criteria: woman aged ≥35 years or unexplained infertility or with one ovary or a history of significant ovarian surgery or endometriosis or a history of a poor response to HMG. Exclusion criteria not mentioned

114 women undergoing IVF. Inclusion and exclusion criteria not mentioned

Hofmann et al. (1995)

Hofmann et al. (1996)

Loumaye et al. (1990)

Hofmann et al. (2000)

Tanbo et al. (1990)

123 women from infertility clinic. Inclusion criteria: infertile women aged 22 to 50 years with a history of shortened menstrual cycles, grey hair, age 35 years or older, or a familial history of early menopause. Exclusion criteria: women who did not complete the test

Hicks et al. (2003)

Erdem et al. (2004)

353 women undergoing IVF, 483 oocytes retrievals. Inclusion and exclusion criteria not mentioned

Period of recruitment not mentioned; Prospective recruitment

Sum of cycle days 3 and 10 FSH values >26.03 IU/l; immunoradiometric assay (Madgenix, Fleurus, Belgium)

Period of recruitment FSH ≥ 25 IU/l (cycle day 3 or 10); IMx kit (Abbott not mentioned; Laboratories, Abbott Park, IL, prospective USA)

Continued on page 722

Poor response; pregnancy per woman

Pregnancy per woman

Period of recruitment Progesterone ≥1.1 ng/ml elevated Pregnancy per woman not mentioned; and ≤0.9 ng/ml normal; RIA prospective (Diagnostic Products, Los Angeles, CA, USA)

Period of recruitment FSH ≥ 25 IU/l (day 3 or 10); IMx Recurrent pregnancy loss not mentioned; kit (Abbott Laboratories, Abbott retrospective Park, IL, USA)

Cycle cancellation; pregnancies per cycle

Cycle cancellation; pregnancies per cycle

FSH ≥10 IU/l (day 3 or day 10); Pregnancy per woman assay not mentioned

FSH >12 IU/l (day 9, 10 or 11); Delfia FSH kit (JKB, Wallace, Turku, Finland) Jan 1987 to Dec 1989; FSH >12 IU/l; Delfia FSH kit prospective (JKB, Wallace, Turku, Finland)

1990 to 1991; prospective

1999 to 2002; retrospective

Pregnancy per cycle Jan 1998 to Dec 1999; Cut-off point FSH ≥10 IU/l (day 3 or day 10, day 3 and 10); retrospective Technicon Immuno 1 system (Bayer Corp., Tarrytown, NY, USA) Poor response; cancelled 32 women undergoing IVF/ICSI. Inclusion criteria: regular menstrual cycles (21–35 Jan 2000 to Dec FSH ≥10 IU/l (basal or day cycle days), no endocrine disorders and ovaries visible on ultrasonography. Exclusion 2000; prospective 10); chemoluminescence recruitment criteria: patients with basal FSH concentrations >18 mIU/ml or clinical and immunometric assay (Diagnostic sonographic signs of polycystic ovaries (i.e. >10 follicles, peripherally located and Products Corporation, Los smaller than 10 mm and increased ovarian stromal density) Angeles, CA, USA)

Cut-off/Assay

Yanushpolsky et al. (2003)

Study design

Population

Study

Table 1 continued from page 720

Review - Dynamic tests of ovarian reserve - A Maheshwari et al.

721

722

110 women undergoing first IVF. Inclusion criteria: age 18–39 years, unexplained infertility for >3 years and/or due to a male factor and/or cervical hostility. Women with regular menstrual cycles, two ovaries, and at least one patent Fallopian tube. Exclusion criteria: polycystic ovary syndrome, severe male factor, insufficiently corrected endocrinopathies, clinically relevant systemic diseases, or a body mass index >28 kg/m2.

Jun 1997 to Dec 1999; randomized prospective

FSH >14,>16, >18, >20, >22 IU/l Poor responders (day 10); commercially available immunometric assay (Amerlite, Amersham, UK)

CC = clomifene citrate; HMG = human menopausal gonadotrophin; ICSI = intracytoplasmic sperm injection; IUI = intrauterine insemination; PCO = polycystic ovaries; RIA = radioimmunoassay.

Kwee et al. (2006)

53 women undergoing IVF. Inclusion criteria: women (27–38 years old) who had Period of recruitment failed to achieve pregnancy after 2 years of tubal surgery, regularly menstruating and not mentioned; underwent prior to surgery a complete investigation for infertility. Exclusion criteria prospective not mentioned

Csemiczky et al. (1996)

Pregnancy

FSH: >10 IU/l on day 3 and day Pregnancy rate 10 and inhibin 35 years or had a history of oligo-ovulation or a poor prior response to ovarian stimulation, FSH >15

Corson et al. (1999)

FSH >7 IU/l (separate 2 × 2 tables for day 3 and day 10); RIA (Diagnostic Products Corporation, Los Angeles, CA, USA)

FSH >26 IU/l on day 9–11; RIA Pregnancy (Amersham Radiochemical Centre, Amersham, UK)

June 1983 to Jan 1986; prospective recruitment

Reference standard

51 women from general infertility clinic. Inclusion criteria: ≥35 years, regular periods, normal semen analysis, normal HSG

Cut-off/Assay

Navot et al. (1987)

Study design

Population

Study

Table 1 continued from page 721

Review - Dynamic tests of ovarian reserve - A Maheshwari et al.

RBMOnline®

RBMOnline®

197 229

5

2

2

Nonpregnancy

Nonpregnancy

Nonpregnancy

Cycle 2 cancellation Cycle 2 cancellation

≥10 (day 3 or 10) >12 (day 9, 10 or 11) >12 (day 9, 10 or 11) >10 (day 3 or 10)

Sensitivity (95% CI)

All (assisted reproduction treatment + general infertility) Assisted reproduction treatment

Assisted reproduction treatment

0.65 (0.56– 0.74) 0.43 (0.29– 0.58)

0.40 (0.33– 0.48)

0.35 (0.31– 0.40)

0.26 (0.23– 0.31)

All (assisted 0.27 (0.24– reproduction treatment 0.30) + general infertility + women undergoing IUI)

Population

Assisted reproduction treatment Kahraman et al., 1997; Van der Stege Assisted reproduction et al., 2001 treatment

Tanbo et al., 1990, 1992

Tanbo et al., 1990, 1992

Scott et al., 1995; Kahraman et al., 1997; Van der Stege et al., 2001; Csemiczky et al., 2002; Ng et al., 2005; van Swieten et al., 2005; Magendzo et al., 2006 Kahraman et al., 1997; Van der Stege et al., 2001; Csemiczky et al., 2002; Ng et al., 2005; van Swieten et al., 2005; Hicks et al., 2003; Yanushpolsky et al., 2003

CI = confidence interval; DOR = diagnostic odds ratio; IUI = intrauterine insemination; LR = likelihood ratio.

249

607

661

1395

>10 (day 3 or 10)

7

Nonpregnancy

>10 (day 3 or 10)

Number Number Studies included of studies of cycles

Reference standard

Cut-off value (IU/l)

Table 2. Summary statistics for the clomifene citrate challenge test.

0.77 (0.69– 0.84) 0.77 (0.70– 0.83)

0.94 (0.70– 0.10)

0.83 (0.77– 0.88)

0.87 (0.81– 0.91)

0.92 (0.89– 0.94)

Specificity (95% CI)

DOR (95% CI)

2.96 (0.53– 16.66) 1.83 (1.21– 2.77)

0.44 (0.34– 0.58) 0.75 (0.58– 0.97)

6.52 (1.21– 35.10) 2.44 (1.26– 4.71)

4.02 (0.91– 0.65 (0.55– 7.17 (1.30– 17.80) 0.77) 46.17)

2.06 (1.48– 0.77 (0.70– 2.67 (1.75– 2.88) 0.84) 4.10)

1.77 (1.01– 0.86 (0.74– 2.11 (1.04– 3.11) 0.99) 4.29)

2.48 (1.23– 0.82 (0.75– 3.08 (1.38– 5.01) 0.90) 6.91)

+LR (95% –LR (95% CI) CI)

Review - Dynamic tests of ovarian reserve - A Maheshwari et al.

723

Review - Dynamic tests of ovarian reserve - A Maheshwari et al.

the measure of the diagnostic attribute of a test. ROC planes were constructed (both for prediction of cycle cancellation and non-pregnancy) (Figures 2 and 3). Only five studies were suitable for inclusion in ROC plane for cycle cancellation. The sixth study (Csemiczky et al., 2002) reported cancellation rate per cycle for which a 2 × 2 table could not be constructed. Of 17 studies reporting pregnancy as the outcome measure, we could only 14 studies could be included in the ROC plane for the prediction of non-pregnancy as the definition of an abnormal test included different combinations of hormones: progesterone measurement (Hofmann et al., 1995); sum of cycle day 3 and 10 FSH concentrations (Loumaye et al., 1990); FSH + inhibin (Corson et al., 1999). It was not considered appropriate to combine them with rest of the studies using only FSH at either day 3 or 10 to define an abnormal test. The positive predictive values were calculated using the available indices for determination of non-pregnancy and cycle cancellation at various cut-offs using Bayes’ theorem. The post-test probability for cycle cancellation at cut-off of FSH 10 IU/l was 0.23 whereas at FSH 12 IU/l, it was 0.69. There were only two studies at each cut-off (data not shown). The post-test prediction of non-pregnancy was 0.97 at cut-off FSH of 7 IU/l; 0.81 (95% confidence interval [CI] 0.71–0.92) at cut-off FSH of 10 IU/l (five studies); 0.97 (95% CI, 0.7–1.2) at cut-off FSH of 12 IU/l (two studies, same author); and 0.94 at cut-off FSH of 24 and 26 IU/l (one study each). There was no pattern of increasing predictive value with higher cut-offs.

GnRH agonist stimulation test (GAST) A search revealed 28 abstracts from which 19 relevant full text articles were reviewed, including one in Mandarin Chinese (Zhu and Zhou, 2002). A total of 12 papers were included for final analysis (Figure 4). The full text of the Mandarin Chinese article was translated into English. Details of the included studies are given in Table 3, along with population, study design, methodology of the index test and along with the reference standard used. In eight papers, the study population comprised of women undergoing IVF. Of the remaining four studies, two included women undergoing IUI, one an infertile population and the last healthy women not trying to conceive. It was not considered appropriate to combine the results from the studies using different populations as poor response in IUI cycles is defined differently to that in IVF.

724

Of the eight studies on an IVF population, Table 3 shows that there was a wide variation in choice of GnRH agonist administered (dose and timing), hormones tested in initial samples (three) hormones tested in the final sample, cut-off points used and the reference standard. A response pattern of A, B, C and D were used by four studies; however, A, B, C and D were defined in different ways. These patterns were first described by Padilla et al. (1990): pattern A: serum oestradiol doubled the baseline value by cycle day 3 followed by at least a 10% drop by cycle day 4; pattern B: serum oestradiol showed a delayed elevation where doubling of baseline oestradiol concentration occurred by cycle day 4 or cycle day 5 followed by a fall; pattern C: serum oestradiol showed a prompt and persistent rise through cycle day 5 after leuprolide injection; pattern D: serum oestradiol remained low through cycle day 5. Five studies did not have clearly specified cut-off values. Where cut-off values were

available, most of them were not predetermined. The review could not find any two studies conducted in a similar manner that evaluated the same hormones pre- and post-GnRH agonist administration. It was, therefore, not possible to aggregate data by means of meta-analysis. Hence, with the present level of evidence, this review is unable to comment on the diagnostic accuracy of GAST, as a test, for any of the fertility outcomes.

Exogenous FSH ovarian reserve test (EFORT) Of the seven included studies (Figure 5), six did not mention the cut-off values for defining a normal and abnormal test. Poor and good responders were initially defined and the response to exogenous FSH was compared in both. The authors were written to and two replies were received (Yong et al., 2003; Iwase et al., 2005) to confirm that normal and abnormal tests were not defined. In this situation, a 2 × 2 table could not be constructed. As shown in Table 4, EFORT has been performed in a number different ways. Each reported method has varied in terms of choice of FSH (dose and preparation), hormones tested in initial serum samples, hormones tested in the final serum sample and timing of the final serum sample. A meta-analysis was therefore not possible. Hence, with the present level of evidence, this study is unable to comment on the diagnostic accuracy of EFORT, as a test, for any of the fertility outcomes.

Discussion Systematic reviews of evaluation of tests are undertaken for the same reasons as the systematic reviews of treatment intervention to produce estimates of test performance and impact based on all available evidence, to evaluate the quality of published studies and to account for variation in the findings between studies (Deeks, 2001). Assuming that a normal test result indicates favourable fertility prognosis, the sensitivity of an abnormal test result is the accuracy with which it can identify all women who cannot conceive, specificity is its ability to exclude women who are likely to conceive, positive predictive value is the probability of non-conception in women with abnormal tests, and negative predictive value provides the probability of conceiving if the test is normal. Likelihood ratio of a positive test quantifies how much more likely it is that an abnormal test will be found in a woman who cannot conceive than in a woman who can; likelihood ratio of a negative test indicates how much more likely it is that a normal test will be found in a woman who can conceive than in a woman who cannot conceive (Jain et al., 2004). Given that sensitivity, specificity and predictive value do not behave independently in meta-analyses, the most appropriate statistics for pooling the meta-analyses are diagnostic odds ratio and likelihood ratios of positive and negative tests. Likelihood ratios are most likely to be of clinical relevance, as there are accepted cut-offs for levels of test accuracy and knowledge of disease prevalence gives a clear estimate of the percentage likelihood of a disease being present once the diagnostic test result is known (Khan et al., 2001). For a test to be useful at ruling out a disease it must have high sensitivity, and for it to be useful for confirming the disease it must have high specificity. RBMOnline®

Review - Dynamic tests of ovarian reserve - A Maheshwari et al.

Figure 2. Receiver operating characteristic (ROC) plane demonstrating prediction of cycle cancellation due to poor ovarian reserve as measured by clomifene citrate challenge test.

Figure 3. Receiver operating characteristic (ROC) plane demonstrating prediction of non-pregnancy due to poor ovarian reserve as measured by clomifene citrate challenge test.

Figure 4. Study selection process for systematic review of gonadotrophin-releasing hormone agonist stimulation test to predict fertility outcome.

725 RBMOnline®

726

228 women undergoing IVF. Exclusion Period of criteria: single ovary, functional recruitment ovarian cyst, PCOS not mentioned; prospective and consecutive recruitment of eligible women

97 patients/100 cycles undergoing stimulation for IVF. Exclusion criteria not mentioned

Winslow et al. (1991)

Padilla et al. (1990)

Jan 1988 to Jun1988; retrospective

Oct 1993 to June 1994; prospective and consecutive recruitment

67 women/78 IVF cycles. Exclusion and inclusion criteria not mentioned

GaltierDereure et al. (1996)

Study design

Population

Study

Amount of human menopausal gonadotrophin used; length of stimulation; peak oestradiol concentration; number of oocytes; pregnancy rate Clinical pregnancy Four different serum oestradiol concentrations before and after gonadotrophin therapy were identified: doubling of oestradiol on day 3 followed by 33% rise in day 4; doubling of oestradiol on day 3 followed by relative plateau (33% decrease in day 4; less than a doubling of oestradiol on day 3. Oestradiol concentrations determined by Pantex, Santa Monica, CA, USA

Continued on page 727

Twice baseline oestradiol concentration by day 5; twice baseline FSH concentration by day 5. Four different oestradiol patterns were described from cycle day 2–5: doubling of oestradiol on day 3 followed by 10% drop by day 4; delayed elevation where doubling of oestradiol by day 4/5 followed by a fall; prompt and persistent rise through cycle day 5; serum oestradiol remained low through cycle day 5. Oestradiol concentration determined by double-antibody RIA (Diagnostic Products Inc., Los Angles, CA, USA); LH: RIA (Serono Diagnostics Inc., Norwell, MA, USA)

Poor response; number of oocytes; number of embryos

FSH1: 7.65 IU/l; FSH1 + FSH2: 24.6 IU/l RIA (Biomerieux, Lyon, France)

On day 1 of the stimulated cycle, serum FSH was measured, 0.3 mg buserelin were injected s.c. A second blood sample was obtained 2 h later. The following parameters were evaluated: FSH concentration before buserelin injection (FSH1), FSH concentration after buserelin injection (FSH2), sum and the difference between and percentage increase in FSH concentration. Patients were stimulated with 1 mg of leuprolide acetate s.c. on day 2 at 4.00 p.m and again on days 3 and 4 at 8.00 a.m. Baseline oestradiol, progesterone, luteinizing hormone and FSH concentrations were obtained at 8.00 a.m on day 2 and again on days 3 and 4 (just before leuprolide acetate)

All patients started the leuprolide acetate on menstrual cycle day 2, given every morning s.c. at a dose of 0.75 mg for patients under 70 kg and 1 mg for patients over 70 kg. The serum was assayed for oestradiol, FSH and LH on cycle days 2–5 and for oestradiol and LH thereafter

Reference standard

Cut-off/Assay

Index test

Table 3. Characteristics of studies evaluating gonadotrophin-releasing hormone agonist stimulation test (GAST).

Review - Dynamic tests of ovarian reserve - A Maheshwari et al.

RBMOnline®

RBMOnline®

Lashen et al. (1999)

84 women at high risk of cycle cancellation undergoing IVF/ICSI. Inclusion criteria: previously raised follicular phase FSH concentration >10 IU/l, >39 years of age, previous poor response to ovarian stimulation 81 women with unexplained infertility undergoing IUI, age limit not mentioned

McIlveen et al. (2007)

Oct 1996 to Sept 1997; prospective and consecutive recruitment of eligible women, who consented

Mar 2004 to Jun 2005; prospective recruitment

Jan 1996 to Feb 1997; prospective and consecutive recruitment

177 women undergoing IVF

Ranieri et al. (1998)

Blood samples were taken on cycle day 2 between 8.00 and 11.00 am (before and 24 h after starting GnRHa, 400 Mg twice daily). Samples were assayed for oestradiol and FSH

Buserelin nasal spray 100 Mg delivered into each nostril every 4 h during the day. The dose was doubled before bed time and administration of the drug was suspended during the night and started again next day at 8.00 a.m. Blood samples were collected on cycle day 2 to determine basal concentrations of FSH and oestradiol and the FSH/LH ratio, and on cycle days 3 and 4 to assess the increase in FSH and oestradiol after the commencement of GnRHa stimulation Blood sample was collected on day 2 of the menstrual cycle for measurement of oestradiol, inhibin B; 24 h later, second sample to measure oestradiol and inhibin B

Daily s.c. injection of GnRHa (100 Mg/ day). Blood samples were taken during the first 3 days in the morning, just before the injection of agonist; $oestradiol = oestradiol on day 2 – oestradiol on day 1

Nov 1991 to Dec 1992; prospective recruitment

140 cycles in 112 women, undergoing IVF cycles following pretreatment with progesterone. Inclusion criteria: 3 mature follicles) only. Oestradiol was measured by direct immunoassay (Sorin Biomedica Diagnostica, Wokingham, Berks, UK)

Day-3 oestradiol/day-2 oestradiol >2; day-3 oestradiol – day-2 oestradiol >195. Assay not mentioned.

$oestradiol ≥5 pg/ml. Four different patterns of oestradiol rise were described from day 2–5: day-3 serum oestradiol showing a secondary increase greater than one-third above that on day 2; no change larger than one-third of that on day 2; subsequent decrease greater than one-third below that on day 2; lack of doubling on day 2. Oestradiol measured by solid phase assay (Coat A Count DPC, Los Angeles, CA, USA) Increase in oestradiol 180 (day-3 to day-2 oestradiol concentration) + FSH 9.5. Oestradiol measured by RIA (Amersham International, UK)

Cut-off/Assay

Continued on page 728

Hyper-response

Poor response

Inadequate response; adequate response; hyper-response

Pregnancy rate per transfer

Reference standard

Review - Dynamic tests of ovarian reserve - A Maheshwari et al.

727

728

Scheffer et al. (2003)

144 women with unexplained infertility March 1999 undergoing first ovulation induction. to Dec 2002; Exclusion criteria: women over 40 prospective years of age, body mass index ≥30 kg/ random m2, anovulation or irregular cycles, selection ovarian cysts, ovarian endometriosis, uterine leiomyoma, PCOS, ovarian surgery, radiotherapy, chemotherapy, autoimmune disease, male factor infertility 162 healthy volunteers, only 40 women Period of underwent GAST. Inclusion criteria: age recruitment not mentioned; 25–46 years, proven normal fertility, prospective regular menstrual cycles, a biphasic body temperature chart, proven natural recruitment fertility by having at least one pregnancy carried to term, each of their pregnancies arisen spontaneously within one year after the start of unprotected intercourse, no evidence of endocrinological disease, no history of ovarian surgery, no ovarian abnormalities as assessed by transvaginal ultrasound, hormonal contraception stopped >2 months, before entering study protocol

Ozkaya et al. (2004)

A single s.c. injection of 100 Mg of triptorelin was administered at day 3 of the cycle. Blood samples were taken immediately before and 24 h after GnRH agonist administration for oestradiol, FSH and inhibin B.

Single s.c. injection of 1 mg buserelin acetate was given on day 2 of the cycle. Baseline oestradiol was measured concomitantly. Serum oestradiol was measured again after 24 h; day-3 oestradiol – day-2 oestradiol

27 clomifene-resistant women and Period of Patients were stimulated with s.c. 8 women with pituitary dysfunction recruitment injection of GnRHa 150 Mg/day between undergoing IUI. Inclusion criteria: not mentioned the 2nd and 4th day of menstruation negative progesterone withdrawal test, pituitary stimulation test showing pituitary dysfunction causing infertility or women under clomifene citrate treatment for 3 or more weeks, patent tubes and normal semen parameters

Index test

Zhu and Zhou (2002)

Study design

Population

Study

Table 3 continued from page 727

Young age 25– 34; middle age 35–40; old age 41–60

Number of mature follicles; number of ampoules of recombinant FSH used

Clinical pregnancy; ongoing pregnancy

Reference standard

Continued on page 729

Cut-off points not mentioned. Oestradiol concentrations were assayed with microparticle enzyme immunoassay (Abbott Laboratories, UK); inhibin B: immunoenzymometric assay (Serotec, Oxford, UK)

Increase in baseline serum oestradiol at least once on the 3rd day of menstruation, i.e. 24 h after first injection, was counted as a response. Three types of oestradiol responses were identified from day 2 to 6 of menstruation: exponential increase in oestradiol (daily increase >33%) or oestradiol concentration at plateau on the 4th to 6th day of menstruation (changes in oestradiol 33%); no response. RIA (Medical kits, Tianjin Depp Company) Cut-off points not mentioned. Assay not mentioned

Cut-off/Assay

Review - Dynamic tests of ovarian reserve - A Maheshwari et al.

RBMOnline®

RBMOnline®

57 women undergoing IVF/ICSI. Inclusion criteria: regular menstrual cycles (25–35 days), presence of both ovaries, no evidence of endocrine disorders, age 3 years duration, male factor, cervical hostility, regular menstrual cycle with documented ovulation (basal body oestradiol benzoate), 18–39 years of age, presence of both ovaries, at least one patent tube, no other pathology on diagnostic laparoscopy. Exclusion criteria: irregular menstrual cycles, oligo or amenorrhoea, severe male factor, insufficiently corrected endocrinopathies, clinically relevant systematic disease and body mass index >28 kg/m2

Poor response; hyperresponse

Pregnancy; cancellation

Poor response

Poor response

Reference standard

Continued on page 732

Cut-off point not mentioned. Assays: oestradiol: Amerlite (Amersham, UK and Sorin Biomedica, Saluggia Italy); inhibin B (Serotec Oxford, UK)

Cut-off point not mentioned; FSH: immunoradiometric assay (Immunotech International, Marseilles, France); oestradiol: RIA (BioMerieux, Marcyl’Etoile, France); inhibin: immunozymatic methods (Medgenix Diagnostics, Brussels, Belgium)

The test was performed during a menstrual cycle within 3 months of IVF treatment. Women received two ampoules (150 IU) of HMG i.m. for 5 days starting from day 2–3 of a spontaneous cycle. Blood samples were taken before and after the completion of HMG (day 7–8). Basal and post-HMG FSH, oestradiol and inhibin B was measured 300 IU recombinant FSH was administered s.c. Blood samples for the determination of FSH, oestradiol and inhibin B were drawn just before (basal values) and 24 h after (stimulated values). Analysis of EFORT included the following parameters: oestradiol increment, inhibin B increment 24 h after administration of FSH

Period of recruitment not mentioned; prospective

Blood samples were taken in early follicular phase (days 3–5) and 24 h after the s.c. administration of 225 IU recombinant human FSH

Jan 1995 to Dec 1998; prospective

46 women undergoing IVF. Inclusion criteria: 58 spontaneously ovulating women about to undergo IVF treatment. Exclusion criteria: endometriosis (other than minimal), presence of only one ovary, previous ovarian surgery, polycystic ovary syndrome or other endocrine disorders, exposure to drugs within 3 months of participation in the study, abnormal prolactin or thyroid function test 80 women undergoing first cycle of IVF. Inclusion criteria: 40 consecutive women having a basal (day 2–3) FSH ≥7 IU/l were initially selected, the remaining 40 patients were selected as follows: a patient having a basal FSH 42 years, severe male factor (