Oestradiol and FSH in gestational trophoblastic disease:a prospective analysis of their predictive value

Robert Macdonald,a John Kirwan,a Catherine Thomson,b Barry Hancock,c

Jan Everard,c John Tidyd

Objective To determine whether oestradiol or follicle-stimulating hormone (FSH) measured prior to treatmentcould improve the predictive value of the risk score used to determine treatment in cases of persistentgestational trophoblastic disease (GTD).

Design Prospective observational study.

Setting Tertiary referral centre for GTD.

Population All women referred to Weston Park Hospital, Sheffield, with GTD between 1st January 1996 and31st December 2000.

Methods Blood was taken prior to the initiation of treatment to measure oestradiol and FSH. The results wereanalysed with respect to the time taken for the h-hCG to return to normal.

Main outcome measures Time taken to reach a normal h-hCG.

Results Data on 118 women were collected. Three women died of GTD during follow up. Using Cox’sproportional hazards regression analysis, division of the risk scores into high and low risk groups (�7, >7)demonstrated a significant difference with regard to the length of time taken to reach a normal h-hCG level(hazard ratio 0.32; 95% CI: 0.18, 0.57) comparing high risk relative to low risk. However, addition ofneither oestradiol nor FSH to the Cox regression analysis produced a significant hazard ratio (Ln oestradiol,0.95; FSH 0.99). Division of the patients’ risk scores into three groups of low (0–4), intermediate (5–7) andhigh (>7) risk groups produced similar results.

Conclusions The measurement of neither oestradiol nor FSH appears to improve the prediction of outcome inpersistent GTD when added to the risk score.

INTRODUCTION

Gestational trophoblastic disease (GTD) is an uncom-

mon disease; the quoted incidence is between 0.71 and 3.27

per 1000 pregnancies,1–3 with the most recently reported

incidence in Britain being 1 per 714 live births.4 Chorio-

carcinoma occurs after 1 in 50,000 live births but can arise

following any previous pregnancy; a full-term delivery, an

abortion or miscarriage, an ectopic pregnancy or most

commonly following a complete hydatidiform molar preg-

nancy. Persistent trophoblast occurs in up to 15–20% of

complete molar pregnancies and 21–55% of twin (normal

and molar) pregnancies.5,6 Indications for the diagnosis and

treatment of persistent trophoblastic disease are based on

the criteria outlined in Table 1.7 Treatment is medical, not

surgical, with chemotherapy regimens in specialist centres

producing a cure rate of 100% and 76–82% for low and high

risk cases, respectively.8 Clinically, treatment protocols are

based on the risk assessment at the outset of treatment. At

Weston Park Hospital in Sheffield, the Charing Cross prog-

nostic scoring system, with the Sheffield modification, has

been used to determine treatment (Table 2).9

The normal placenta is known to produce a wide variety

of hormones during pregnancy. The same hormones are

also produced in cases of gestational trophoblastic tumours.

The placenta produces oestrogens (oestriol, oestrone and

oestradiol), human chorionic gonadotrophin (hCG), human

placental lactogen (HPL) and placental alkaline phospha-

tase. Hydatidiform moles and choriocarcinomas are derived

from the syncytiotrophoblast, which is the main source of

oestradiol, oestrone and hCG. In vitro cultures of chorio-

carcinoma cells have been shown to produce oestrogen10

and contain oestrogen receptors.11 There is also an associ-

ation demonstrated between the presence of theca lutein

cysts (a known risk factor for requiring chemotherapy) and

an increase in oestrogen levels measured.12 The trophoblast

production of hCG is used in the monitoring of gestational

trophoblastic tumours as hCG levels give an accurate

BJOG: an International Journal of Obstetrics and GynaecologyJuly 2005, Vol. 112, pp. 977–980

D RCOG 2005 BJOG: an International Journal of Obstetrics and Gynaecology www.blackwellpublishing.com/bjog

aDepartment of Gynaecological Oncology, Liverpool

Women’s Hospital, Liverpool, UKbTrent Cancer Registry, Sheffield, UKcSheffield Trophoblastic Disease Centre, Weston Park

Hospital, Sheffield, UKdDepartment of Gynaecological Oncology, Royal Hallamshire

Hospital, Sheffield, UK

Correspondence: Dr R. Macdonald, Liverpool Women’s Hospital, Crown

Street, Liverpool, L8 7SS, UK.

DOI: 10.1111/ j .1471-0528.2005.00612.x

estimation of the tumour volume remaining, hence the

necessity for further treatment or otherwise.

Although there is a use of (indeed reliance on) the mea-

surement of hCG production from trophoblastic tumours,

other associated hormones [oestradiol, follicle-stimulating

hormone (FSH) and luteinizing hormone (LH)] are mea-

sured but are of undetermined significance. There is a re-

ported association linking oestrogen (particularly combined

oral contraceptive use) to a poorer outcome,13 although this

is disputed.14 We are not aware, however, of any data

examining the prognostic value of oestradiol or FSH. Thus,

the aim of this study is to investigate the potential signif-

icance of pretreatment levels of these hormones and com-

pare them to present prognostic measures and clinical

outcome.

METHOD

All cases of persistent GTD referred to the Sheffield

Trophoblastic Disease Centre at Weston Park Hospital

between 1st January 1996 and 31st December 2000 were

part of the study. The information collected on each patient

is shown in Table 3. All those with complete blood results

available were included. The data come from the Clinical

Oncology Database from Weston Park Hospital, Sheffield,

where the demographic data, risk score, antecedent preg-

nancy, histological diagnosis, h-hCG levels and any relevant

outcome information were recorded prospectively. All data

were transferred to a Microsoft Excel database, and any

missing data were tracked down individually.

At present, treatment decisions are based initially on the

risk score (Table 2). If any of the measured hormones were

to have any relevance to the clinical decision making in

patients with GTD, the risk score plus the additional infor-

mation would be required to provide a better prediction of

outcome than the risk score alone. As the mortality was

expected to be so low in this group of patients, the time

taken for the measured h-hCG at staging to reach a normal

value was taken as the endpoint representing a measure of

the severity of disease.

Cox’s proportional hazards models were fitted to the risk

score split into two groups (�7, >7 as is done in Sheffield)

and the standard three groups (low, intermediate and high

risk; 0–4, 5–7 and >7) to examine how well the risk score

predicted the risk of developing persistent GTD on aver-

age within the groups. To examine whether the addition of

oestradiol or FSH could improve the risk score, the natural

log of oestradiol (Ln E2) (the continuous variable E2 was

highly skewed) was used. There was no such skewed distri-

bution for FSH. This was assessed by stratifying the anal-

yses by either the two or three levels of the risk score and

averaging the effect of Ln E2 or FSH over these levels to see

whether either Ln E2 or FSH was associated with the time to

return to normal h-hCG, for each given risk score.

Additionally, the Harrell’s c statistic15 was obtained to

assess the discrimination at an individual level to determine

how well the risk score predicted an individual woman’s

chances of returning to a normal h-hCG level. With Harrell’s

c statistic, values close to 1 indicate near-perfect concor-

dance of predicted and actual ranking of the times taken to

return to normal h-hCG level on an individual basis (the

chance expected value is 0.5). For a binary outcome,Harrell’s

c statistic is the area under the ROC plot. All statistical

analyses were undertaken using the statistics package Stata,

version 7.0. (StataCorp, College Station, TX, USA).

Table 1. Criteria for the initiation of chemotherapy for persistent tropho-

blastic disease.

h-hCG levels >20,000 iu/L after one or two uterine evacuations

Static or rising hCG levels after one or two uterine evacuations

Persistent hCG elevation six months post uterine evacuation

Persistent uterine haemorrhage with raised hCG levels

Pulmonary metastases with static or rising hCG levels

Metastases in liver, brain or GI tract

Histological diagnosis of choriocarcinoma

Table 2. Charing Cross prognostic scoring system used in Sheffield in 1990s (from WHO system).

Score

Risk Factors 0 1 2 4

Age �39 >39 – –

Antecedent pregnancy Molar pregnancy Abortion/miscarriage Term pregnancy –

Interval (months) <4 4–6 7–12 >12

Pre-Rx hCG (iu/L) <103 103–104 104–105 >105

ABO blood group

(female � male)

– A � O, O � A, O or

A � unknown

AB � A or O –

Largest tumour (cm) <3 3–5 >5 –

Metastatic site None, lungs, vagina Spleen, kidney GI, liver Brain

No. of metastatic sites 0 1–4 4–8 >8

Previous chemotherapy – – Single agent

chemotherapy

Two or

more agents

Low risk: 0–4; intermediate risk: 5–7; high risk: >7.

Sheffield modification: low risk: �7; high risk: >7.

978 R. MACDONALD ET AL.

D RCOG 2005 BJOG: an International Journal of Obstetrics and Gynaecology 112, pp. 977–980

RESULTS

A total of 145 women were recorded as being referred to

Sheffield between October 1996 and December 2000 and

having blood samples, from which the results of 118 oes-

tradiol and FSH measurements were available. These

118 women were the focus of this paper. The median age at

referral was 30 years (range 15–52). There were some

incomplete data; one woman’s past obstetric history (grav-

idity and parity) was unknown, one unknown time from

evacuation to staging (due to a previous hysterectomy), 15 un-

known gestations, mainly due to no LMP dates recorded,

and two risk scores were unavailable. Oestradiol data were

available for all 118 women, while for FSH there were

110 records. The mean oestradiol level prior to treatment

was 568 pmol/L [954] and FSH 5.1 iu/L [11.0]. Division into

the low and high risk groups for the oestradiol data gave a

mean of 363 pmol/L [292] and 1707 pmol/L [2036], re-

spectively. For FSH the same figures for the low and high

risk groups were 4.6 iu/L [2.98] and 7.28 iu/L [26.9].

The breakdown of cases is shown in Table 4. Overall,

three women died due to GTD; one following a metastatic

placental site tumour, and two following stage IIIa chorio-

carcinomas. One other woman died from an unrelated

illness two years after successful treatment.

Figure 1 shows the significant separation between the

time taken to return to a normal h-hCG level for the two risk

score groups. Cox’s proportional hazards modelling gave a

hazard ratio of 0.32 (95% CI: 0.18, 0.57), indicating that as

a group, women with a high risk score at presentation took

longer to return to a normal h-hCG. On an individual level,

however, the Harrell’s c statistic was 0.57 demonstrating

only a limited ability of the risk score split into high and low

risk groups to predict the outcome (whether an individual

woman will return to a normal h-hCG value) in any specific

patient. When the Cox model was stratified by the two levels

of risk score, which averages the effect of either Ln E2 or

FSH over these levels, then the hazard ratio for the Ln E2

was 0.95 (95% CI: 0.75, 1.21; P ¼ 0.70) and for FSH 0.99

(95% CI: 0.98, 1.01; P ¼ 0.64), showing that the addition

of neither oestradiol nor FSH to the risk score improves the

prediction of the length of time taken to return to a normal

h-hCG level. The Harrell’s c statistic for Ln E2 and FSH

added to the risk score was 0.57 and 0.59, respectively, show-

ing that the addition of these two variables does not mate-

rially improve the prediction at an individual patient level.

When the risk score was divided into the three groups,

the hazard ratios for the intermediate and high groups

relative to the low group were, respectively, 0.63 (95%

CI: 0.40, 0.98) and 0.27 (95% CI: 0.15, 0.49), with a highly

significant test for linear trend (m2 ¼ 22.6, df ¼ 1, P <0.0001). This showed that its predictive value at a group

level of determining the chances of returning to a normal

h-hCG was good. The Harrell’s c statistic was 0.62, indi-

cating that it is still only moderately good at prediction on

an individual basis for this data set, although it was slightly

better than the binary risk score. Again neither the Ln E2

nor the FSH were significant in the Cox regression model

when stratified by division of the risk score into three groups.

DISCUSSION

Chemotherapy for GTD is one of the remarkable success

stories of medicine over the last 40–50 years. Recent

figures have shown a 100% cure rate for those with low

Table 3. Data collected on patients referred with GTD.

Patient demographic details Information involved in

staging/ risk assessment

Age FIGO stage

Parity h-hCG at staging

Gravidity Risk score (see Table 2)

Pregnancy factors Histology

Gestation at presentation No. of chemotherapy

cycles completed

Time from evacuation to staging Oestradiol

h-hCG level prompting treatment FSH

Follow up data

Time from staging to the first normal

h-hCG level

Table 4. Breakdown of GTDs presenting to Weston Park Hospital,

Sheffield 1.1.96–31.12.2000.

No. of cases

Complete hydatidiform mole 73

Partial hydatidiform mole 16

Choriocarcinoma 14

Placental site tumour 4

Unclassified molar pregnancy 11

Twin pregnancy with molar pregnancy 0

Fig. 1. Kaplan–Meier curves showing time to return to normal h-hCG.

OESTRADIOL, FSH AND GESTATIONAL TROPHOBLASTIC DISEASE 979

D RCOG 2005 BJOG: an International Journal of Obstetrics and Gynaecology 112, pp. 977–980

risk disease and 76–82% survival even of those with high

risk disease who fail first line chemotherapy.8 However,

multiple chemotherapy agents and multiple courses have

inherent risks; significant short term side effects and poten-

tial increased risks of future treatment induced malignancy,

such as myeloid leukaemia, colon or breast cancer.16 The

aim of scoring systems has been to distinguish between

those patients at high risk of disease progression from those

at low risk, hence minimising the exposure to potentially

toxic chemotherapy to those that are most likely to require

it. There have been many variations in proposed scoring

systems,8,9,17–19 both over time and worldwide, leading to

difficulties in comparing data sets and chemotherapy regi-

men. During the time scale of this study however, the risk

assessment and protocols for treatment in Weston Park

Hospital have been constant, thus allowing the group to be

looked at as a whole. The premise for this study was to

investigate whether the measurement of the oestradiol or

FSH could add further accuracy and refinement to the

prognostic value of the scoring system used in Sheffield in

the 1990s.

The indication for measuring these hormones in the

context of persistent GTD would be if they provided further

information for the staging, risk assessment or prediction of

outcome (i.e., could they add information to the risk score,

hence predicting outcome better than the present scoring

system). Measurement of sex-related hormones has been per-

formed on a regular basis, but with no specific role defined

for the measurement. The findings from this study show

that oestradiol and FSH do not provide additional infor-

mation with regard to scoring systems for GTD and no

improvement in the prediction of outcome. The authors

would not recommend a continued measurement of oestra-

diol or FSH on a routine basis in the management of

persistent GTD.

References

1. Begent RHJ. Trophoblastic Disease. In: Shepherd JH, Monaghan JM,

editors. Chapter 14: Clinical Gynaecological Oncology. Oxford, UK:

Blackwell Scientific Publications, 1990.

2. Palmer JR. Advances in the epidemiology of gestational trophoblastic

disease. J Reprod Med 1994;39:155–162.

3. Kim SJ. Epidemiology. In: Hancock BW, Newlands ES, Berkowitz

RS, editors. Gestational Trophoblastic Disease. London, UK: Chapman

& Hall, 1997.

4. Tham BWL, Everard JE, Tidy JA, Drew D, Hancock BW. Gestational

trophoblastic disease in the Asian population of Northern England

and North Wales. Br J Obstet Gynaecol 2003;110:555–559.

5. Steller MA, Genest DR, Bernstein MP, Lage JM, Goldstein DP,

Berkowitz RS. Natural history of twin pregnancy with complete hy-

datidiform mole and coexisting fetus. J Reprod Med 1994;83:35–42.

6. Foskett MA, Seckl MJ, Paradinas FJ, et al. A review of 126 cases

registered at Charing Cross Hospital as twin multiple pregnancies

complicated by a complete hydatidiform mole (CHM). IXth World

Congress on Gestational Trophoblastic Disease, Jerusalem, 1998.

7. Gillespie AM, Coleman RE, Hancock BW. Gestational trophoblastic

disease: a guide to management practices at Weston Park Hospital.

UK: University of Sheffield, 2001.

8. Kendall A, Gillmore R, Newlands E. Chemotherapy for trophoblastic

disease: current standards. Curr Opin Obstet Gynaecol 2002;14:33–

38.

9. Sheridan E, Hancock BW, Smith SC. Gestational trophoblastic

disease: experience of the Sheffield (United Kingdom) supraregional

screening and treatment service. Int J Oncol 1993;3:149–155.

10. Ritvos O. Modulation of steroidogenesis in choriocarcinoma cells by

cholera toxin, phorbol ester, epidermal growth factor and insulin-like

growth factor I. Mol Cell Endocrinol 1988;59:125–133.

11. Jiang SW, Lloyd RV, Jin L, Eberhardt NL. Estrogen receptor

expression and growth-promoting functionin human choriocarcinoma

cells. DNA Cell Biol 1997;16:969–977.

12. Bevan BR, Savvas M, Jenkins JM, Baker K, Pennigton GW, Taylor NF.

Abnormal steroid excretion in gestational trophoblastic disease com-

plicated by ovarian theca-lutein cysts. J Clin Pathol 1986;39:627–634.

13. Stone M, Dent J, Kardana A, Bagshawe KD. Relationship of oral

contraception to development of trophoblastic tumor after evacuation

of a hydatidiform mole. B J Obstet Gynaecol 1976;83:913–916.

14. Lurain JR. When is it safe to get pregnant after having a complete or

partial mole, or persistent trophoblastic disease? VIIIth World Con-

gress on Gestational Trophoblastic Diseases, Seoul, 1996.

15. Harrell Jr FE, Lee KL, Califf RM, Pryor DB, Rosati RA. Regression

modelling strategies for improved prognostic prediction. Stat Med

1984;3:143–152.

16. Rustin GJS, Newlands ES, Lutz JM, et al. Combination but not single-

agent methotrexate chemotherapy for gestational trophoblastic tu-

mours increases the incidence of second tumours. J Clin Oncol 1996;

14:2769–2773.

17. Bagshawe KD. Risk and prognostic factors in trophoblastic neoplasia.

Cancer 1976;38:1373–1385.

18. World Health Organisation. Gestational trophoblastic diseases. WHO

Technical Reports Series 1983:692.

19. Ngan HY, Bender H, Benedet JL, Jones H, Montruccoli GC,

Percorelli S, FIGO Committee on Gynecologic Oncology. Gestational

trophoblastic neoplasia, FIGO 2000 staging and classification. Int J

Gynaecol Obstet 2003;83:175–177.

Accepted 29 November 2004

980 R. MACDONALD ET AL.

D RCOG 2005 BJOG: an International Journal of Obstetrics and Gynaecology 112, pp. 977–980