Predictive value of plasma human chorionic gonadotrophin following assisted conception treatment

S.E. Sugantha, S. Webster, E. Sundar and E.A. Lenton1

Sheffield Fertility Centre, 24–26 Glen Road, Sheffield S7 1RA, UK


    Abstract
 Top
 Abstract
 Introduction
 Materials and methods
 Results
 Discussion
 References
 
A total of 429 pregnancies after assisted conception treatment was analysed, using receiver operator characteristic curves. The best balance between sensitivity and specificity for predicting viable (single and multiple births) and non-viable (fetal heart positive abortions, ectopic and biochemical pregnancies) outcomes was human chorionic gonadotrophin (HCG) 50 IU/l on day 14 and 200 IU/l on day 21 after treatment. Utilizing these indices all pregnancies could be classified into one of four groups. In group A (day 14 HCG <50 IU/l and day 21 <200 IU/l), the probability of a birth was 0%, pregnancy loss 72% and ectopic pregnancy 28%. Conversely for group D (day 14 HCG >50 IU/l and day 21 >1000 IU/l), the likelihood of a birth was 90%, pregnancy loss 8% and ectopic pregnancy only 1%. Between groups A and D there was, as expected, a gradually shifting balance in favour of a reduction in ectopic (28, 13, 3, 1%) and biochemical pregnancies (70, 36, 33, 2%) and an increase in fetal heart positive pregnancy losses (2, 6, 13, 7%) and births (0, 44, 50, 90%). The majority of multiple pregnancies (98%) occurred in group D. Two accurately linked HCG measurements allowed a greater predictive accuracy of pregnancy outcome than could be obtained using either alone.

Key words: assisted conception/chorionic gonadotrophin/in-vitro fertilization/outcome/predictive value


    Introduction
 Top
 Abstract
 Introduction
 Materials and methods
 Results
 Discussion
 References
 
Pregnancy test kits are widely available and extensively used by both health professionals and the public. Modern test kits are sensitive and specific with low false positive rates (Chard, 1992Go). However pregnancy test kits are not quantitative and cannot distinguish between different outcomes of conception. The ability to predict outcome as soon as possible after assisted conception treatment is important for clinic staff and patients.

A variety of hormones has been evaluated for their ability to predict outcome (Ledger et al., 1994Go; Long et al., 1994Go) but beta human chorionic gonadotrophin (ß-HCG) alone provided the best sensitivity and specificity for detection of pathological pregnancies (Buylos et al., 1992).

Other studies have shown that the daily increase in plasma HCG is significantly greater in viable pregnancies than in ectopic or biochemical pregnancies (Fridstrom et al., 1995Go; Bjercke et al., 1999Go). Similarly HCG concentrations are significantly higher in multiple compared with singleton pregnancies. This observation has been used to predict the probability of a multiple birth (Guth et al., 1995Go).

The present study was undertaken to see if two consecutive linked HCG measurements could give a better prediction of outcome following assisted conception than a single isolated measurement.


    Materials and methods
 Top
 Abstract
 Introduction
 Materials and methods
 Results
 Discussion
 References
 
Patients in the study
All records from patients who conceived following assisted conception treatment between January 1994 and December 1998 were examined. A total of 429 patients fulfilled the criteria of two plasma HCG measurements, exactly 14 and 21 days after egg collection, intrauterine insemination or day of luteinizing hormone (LH) surge, plasma HCG >10 IU/l at least on one occasion and menstrual delay of at least 7 days (i.e. menstrual bleeding delayed 7 days beyond expected post-treatment period). The patients' ages ranged from 24–46 years (including egg recipients) with a mean of 34 years.

Treatment protocols and clinical procedures
The assisted conception treatments resulting in pregnancy included intrauterine insemination (IUI) 115 cycles, in-vitro fertilization (IVF) 157 cycles, intracytoplasmic sperm injection (ICSI) 69 cycles, frozen embryo replacements (FER) 54 cycles and donated oocyte recipients (OR) 34 cycles. The number of embryos transferred depended on factors such as patient age, outcome in previous assisted conception attempts and patient preference. No more than three embryos were transferred in any cycle. Where intrauterine insemination was performed, the number of follicles present varied between one and four. Pregnancy detection was by routine plasma HCG measurement on day 14, followed by further sampling on day 21 for all patients whose day 14 HCG was >10 IU/l with ultimately an ultrasound scan on or around day 28 to document fetal heart activity.

HCG immunoassay
Plasma total HCG measurement was performed using the Abbott IMX Microparticle Enzyme Immuno Assay (MEIA) system (Abbott Diagnostics; Maidenhead, Kent, UK). The reference preparation was the World Health Organization (WHO) Third International Standard 75/537. The upper limit of detection for undiluted samples was 1000 IU/l. Samples which fell outside this upper limit were simply reported as >1000 IU/l and were not analysed further.

Pregnancy definitions
For the purpose of this study pregnancies were classified as `viable' or `non-viable'. Viable pregnancies included multiple pregnancies (where two or more babies were born or where only one baby was born but two fetal hearts had been demonstrated or where there was a late pregnancy loss of a multiple gestation) and singleton pregnancies (where one baby was born (including stillbirths) from a pregnancy in which only one fetal heart had been demonstrated). Non-viable pregnancies included clinical abortions (fetal heart positive pregnancy loss of single or multiple implantations), ectopics (confirmed tubal abortions) or biochemical pregnancies (all pregnancies with two positive HCG measurements, with or without a gestation sac but with no detectable fetal heart pulsation).

Statistical methods
Retrospective analyses of the day 14 and day 21 HCG concentrations were undertaken. Using the scatter plots as a basis to determine probable cut-off limits of HCG values, receiver operator characteristic (ROC) curves were generated (one each for day 14 and day 21 HCG concentrations and one for day 14 HCG values in multiple pregnancies). ROC curves plot sensitivity (true positives) for a given concentration of HCG (y axis) and 1 – specificity (false positives) (x axis). Thus co-ordinates could be obtained which defined the sensitivity and specificity of all HCG values available in order to predict outcome. The co-ordinates were joined to produce ROC curves. A hypothetical curve with high sensitivity and specificity would rise vertically and then turn sharply horizontal at the upper end of the y axis. The point at which the curve deviates to the horizontal defines the point of greatest sensitivity and specificity.

All statistical analyses were performed on Microsoft ExcelTM (Microsoft Corporation, Seattle, WA, USA) using Analyse-it (www.analyse-it.com) statistical package for Microsoft ExcelTM.


    Results
 Top
 Abstract
 Introduction
 Materials and methods
 Results
 Discussion
 References
 
The geometric mean day 14 HCG concentrations for each type of pregnancy or pregnancy loss are shown in Table IGo. HCG concentrations in biochemical and ectopic pregnancies were lower than in successful singleton pregnancies but not significantly so. However HCG concentrations were significantly higher (P < 0.01) in multiple compared with singleton pregnancies.


View this table:
[in this window]
[in a new window]
 
Table I. Geometric means and 68% confidence intervals for human chorionic gonadotrophin (HCG) concentrations on day 14 from egg recovery in 429 assisted conception pregnancies
 
ROC curves were used to determine the points of inflexion (specificity and sensitivity) to predict potential viability. From the ROC curves shown in Figure 1Go, it is evident that day 21 HCG concentrations gave a better prediction of potential viability than day 14 concentrations. This was because the day 21 curve was steeper and deviated later than the day 14 curve.



View larger version (25K):
[in this window]
[in a new window]
 
Figure 1. Receiver operator characteristic (ROC) curve showing the sensitivity on the y axis and 1 – specificity on the x axis of human chorionic gonadotrophin (HCG) concentrations on day 14 and day 21 to predict a non-viable pregnancy.

 
The inflection co-ordinates of these ROC curves were 0.2, 0.8 (day 14) and 0.01, 0.55 (day 21) corresponding to a day 14 HCG concentration of approximately 50 IU/l for prediction of a non-viable outcome (sensitivity 85% and specificity 79, 95%, confidence intervals 77–91% and 74–84% respectively) with positive and negative predictive values of 66 and 92% respectively. For day 21 the equivalent co-ordinates corresponded to 200 IU/l (sensitivity 53%, specificity 99, 95%, confidence intervals 44–62% and 93–100% respectively) with positive and negative predictive values of 99% and 82% predicting a non-viable outcome.

On the day 21 ROC curve the inflection point of 0.1 on the x axis (1 – specificity) and 0.8 on the y axis (sensitivity) gave the best combination of sensitivity and specificity, being the point at which the curve deviated sharply to the right. However, in order to maximize the specificity to offer near 100% specificity in the prediction of non-viable pregnancies, the inflection point at 0.01 which corresponded to 200 IU/l was chosen. This enabled a near certain prediction that day 21 HCG concentrations below 200 IU/l would lead to a non-viable pregnancy and offered some practical advantages in terms of specific counselling of this group of patients.

Using the two inflection points obtained from the ROC curves (50 IU/l on day 14 and 200 IU/l on day 21) as well as the 1000 IU/l (upper limit of assay detection), four combinations of HCG concentrations were apparent:

The proportion of cycles represented by each of these combinations is shown in Table IIGo.


View this table:
[in this window]
[in a new window]
 
Table II. Distribution of pregnancy outcomes according to the paired HCG concentrations obtained on day 14 and day 21 after egg collection in 429 assisted conception pregnancies
 
Patients in group A had a poor prognosis. There were no live births and only one fetal heart positive (FH+) pregnancy (2%) in this group. Ectopic pregnancies were common (28%) as were biochemical pregnancies (70%).

In group B, the outcomes were more evenly distributed between births (44%) and biochemical pregnancies (36%). Ectopic pregnancies were significantly reduced (P < 0.05) at only 13% although there were a small number (7%) of FH+ losses.

Overall outcomes in group C were not dissimilar to group B with 50% of patients having a birth and 33% a biochemical pregnancy. However ectopic pregnancies were relatively uncommon (3%) although FH+ pregnancy losses had become appreciable at 13%. Despite the rising proportion of births in group C, these with one exception were all singletons.

The picture for group D was quite different, as here the prognosis was excellent with 90% of patients achieving a birth. Virtually all (98%) of the multiple pregnancies were also in group D. The incidence of ectopic pregnancy (1%) and biochemical pregnancy (2%) was, reassuringly, almost negligible.

Incidence of biochemical pregnancies
Biochemical pregnancies occurred throughout the groups but were more common (74/89 or 83% of all biochemicals) in groups A and B. This observation probably reflects the mixed aetiology of those pregnancies which are lost prior to detection of a fetal heart or where no fetal heart could be demonstrated even with frequent scanning (anembryonic pregnancies). Those that fall into group A (35/89 or 40%) showed a low initial HCG concentration followed by an equally low day 21 value. These implantations were probably hopelessly compromised from the start. More interestingly the group B biochemicals (39/89 or 43%), although exhibiting a similarly low initial HCG, did seem then to progress more normally such that HCG concentrations exceeded 200 IU/l by day 21. It is possible that some of these pregnancies were not inherently abnormal but were simply delayed or late implantations relative to a normally evolving pregnancy. A small number of biochemical losses (5/89 or 6%) occurred despite completely normal HCG profiles (group D). These were unexpected and could not be predicted but could theoretically have resulted from some sort of implantation catastrophe occurring between day 21 and day 28.

It is interesting to note that in group A in only two out of 19 instances where the day 21 HCG concentration was lower than the day 14 HCG was a diagnosis of an ectopic pregnancy made. The remainder of the 17 gestations were all biochemical pregnancies. A falling HCG concentration therefore offers some reassurance against the possibility of an ectopic pregnancy. The rate of HCG decline in the two diagnosed cases of ectopic pregnancy did not differ from the rate of decline in biochemical pregnancies, suggesting that perhaps these ectopics would have been self-limiting.

Predicting an ectopic implantation
The probability of an ectopic was high (28/32 or 88% of all ectopics and 28/157 or 18% of all pregnancies) if the day 14 HCG was less than 50 IU/l (groups A and B). Above 50 IU/l the odds were more favourable with only 4/272 or 1.5% of all pregnancies presenting as an ectopic implantation. Ectopic pregnancies typically presented with a low HCG on day 14 but within a week concentrations may have risen dramatically and in 9% a good HCG concentration (>1000 IU/l) was observed by day 21. Without the preceding day 14 result these pregnancies would not have given cause for concern.

Predicting a clinical abortion (FH+)
The pattern of HCG in the pregnancies destined to end as FH+ abortion was more variable. Approximately 73% of these pregnancies (22/30) initially presented with good day 14 HCG concentrations (greater than 50 IU/l, groups C and D) but subsequently HCG concentrations rose more slowly in the FH+ loss group compared with the majority of equivalent pregnancies destined to end in a birth. It is difficult to quantify precisely FH+ pregnancy losses because loss of one fetus in a multiple implantation does not necessarily result in loss of the pregnancy. In this series there were only two twin pregnancies in which both fetuses were lost (group D, both second trimester losses).

Considering only those pregnancies in groups C and D where a single fetal heart was demonstrated, 4/18 (22%, group C) compared with 16/181 (9%, group D) failed to progress suggesting that a low day 21 HCG (<1000 IU/l) is about twice as predictive of failure. Despite this 14/18 (78%, group C) and 165/181 (91%, group D) singleton pregnancies did continue.

Predicting a multiple implantation
At the other end of the scale, the probability of a multiple pregnancy was negligible if day 14 HCG concentrations were less than 50 IU/l. This became highly likely when HCG concentrations were rising strongly such that by day 21, concentrations were in excess of 1000 IU/l (98% of all multiples were in group D).

In an attempt to discriminate between normal singleton and multiple pregnancies within group D, a third ROC was developed comparing day 14 HCG concentrations between single and multiple implantations (Figure 2Go). This curve deviated at 0.3, 0.8 which corresponded to about 100 IU/l (sensitivity 85%, specificity 73, 95%, confidence intervals 72–93% and 66–78% respectively). Positive and negative predictive values at 100 IU/l were 57 and 92% respectively.



View larger version (27K):
[in this window]
[in a new window]
 
Figure 2. Receiver operator characteristic (ROC) curve of day 14 HCG and day 21 HCG showing sensitivity on the y axis and 1 – specificity on the x axis to predict a multiple pregnancy.

 
Re-analysis of group D revealed that the majority, 44/51 (86%) of all multiple deliveries, presented with a day 14 HCG greater than 100 IU/l while in the majority of singleton births, 97/165 (59%) showed an initial day 14 HCG which fell between 50 and 100 IU/l.


    Discussion
 Top
 Abstract
 Introduction
 Materials and methods
 Results
 Discussion
 References
 
One of the more recent papers looking into the predictive ability of plasma HCG came from Norway (Bjercke et al., 1999Go). This paper looks at the predictive ability of plasma HCG taken 12 days after embryo transfer in 417 IVF pregnancies. The authors showed that an HCG concentration of 55 IU/l (sensitivity of 0.73 and specificity of 0.83) discriminated best between early pregnancy losses and vital pregnancies, which was remarkably similar to the 50 IU/l (sensitivity 0.85 and specificity 0.79) cut-off described here. Their early pregnancy loss group included biochemical pregnancies, ectopic gestations and spontaneous abortions in the first trimester and vital pregnancies included single and multiple gestations delivered or aborted any time during or after the second trimester. Their two groups correspond closely to the viable and non-viable groups described here, though the viable pregnancy group in the current study included FH+ pregnancy losses.

The study by Bjercke (Bjercke et al., 1999Go) restricts itself to pregnancies occurring after IVF treatment. This was not so in our study as the results were derived from a wide variety of assisted conception treatments and therefore may have more universal applicability.

Qasim (Qasim et al., 1996Go) reported in their study of 153 pregnancies that they were able to distinguish between normal and abnormal pregnancies using a single ß-HCG cut-off 42 mIU/ml (sensitivity 0.79 and specificity 0.83), again very similar to the present study.

A study from Liverpool looking at free ß-HCG in 554 pregnant women demonstrated a significantly lower free ß-HCG in women who had a clinical abortion, ectopic or biochemical pregnancy (Al-Sebai et al., 1996Go). Schmidt et al. (1994) in their study of 391 patients also found significantly lower HCG concentrations in patients with non-viable compared to the viable pregnancies (P < 0.0001). The non-viable group in their study also included ectopic gestations, spontaneous abortions and biochemical pregnancies.

Various studies have shown that HCG can be predictive after assisted conception treatment. A study (Heiner et al., 1992Go) looked at 134 pregnancies where HCG concentrations were measured once between days 14 to 16 after embryo transfer (2–4 days later than in the present study). An HCG concentration of <100 IU/l was shown to discriminate between viable and non-viable outcomes. The incidence of a live birth was only 6% if the HCG was <100 IU/l but 82% if it was >100 IU/l. However a similar study demonstrated that a single HCG measurement on day 9 after treatment was too early to differentiate between biochemical pregnancies, abortions and ongoing pregnancies (Legro et al., 1995Go).

The present study was restricted to women conceiving by assisted conception where samples had been obtained exactly on days 14 and 21 after egg collection or intrauterine insemination and by this stage differences between the outcomes were obvious (Table IGo).

As expected the mean HCG was greatest in the multiple pregnancies. This observation is in agreement with other studies (Heiner et al., 1992Go; Schmidt et al., 1994Go; Glatstein et al., 1995Go; Bjercke et al., 1999Go) where sampling was performed between days 12 and 16 after embryo transfer. However again if sampling was performed earlier (days 7–10 after embryo transfer) neither qualitative nor quantitative differences between singleton gestations and multiple gestations could be detected (Check et al., 1992Go).

The importance of the timing of the HCG sample is reinforced by the ROC curve analyses shown here (Figure 1Go). It was clear that day 21 measurement gave a better sensitivity and specificity. At a sensitivity of 0.5, the day 21 HCG ROC curve offered a false positive rate of <1%, compared with the day 14 curve which at the same sensitivity gave a false positive rate of 5%. Similar studies (Glatstein et al., 1995Go; Bjercke et al., 1999Go) used ROC curves to derive HCG cut-off concentrations for all sampling times between day 12 and day 20 after embryo transfer in women who had undergone IVF.

Glatstein et al. (1995) did not analyse their results serially. Although it appeared that subjects had been tested on at least two occasions, within-patient correlations were not made. The present study attempts to go further and relate the day 14 HCG concentration to the paired concentrations obtained on day 21. As a result better predictability was achieved (Table IIGo). If the day 21 HCG was <200 IU/l, then the prognosis was poor and no births resulted irrespective of day 14 concentrations. Subjects in group A had a 70% chance of a biochemical pregnancy (FH loss), a 28% chance of an ectopic pregnancy, and only a 2% chance of a positive FH abortion.

In group B, where day 14 HCG was <50 and a day 21 HCG was >200 IU/l the prognosis improved considerably. Patients in this group had a 50% chance of a fetal heart being detected and of these pregnancies, 94% resulted in a birth. The quality of pregnancies reaching the fetal heart positive stage in this group was excellent, indicating that the low HCG observed on day 14 (<50 IU/l) might have simply been due to a slightly delayed but normal implantation which was ultimately capable of a normal rate of HCG secretion (concentrations rising on target to exceed 200 IU/l by day 21). No multiple births were observed in this group.

In group C, where subjects had a good day 14 HCG >50 IU/l but a lower than expected day 21 HCG of <1000 IU/l, the outcome overall was rather similar with 53% chance of fetal heart detection although now only 79% chance of delivery. The main difference between groups B and C was that although the proportion of ectopic pregnancies had declined (13–3%), these had been replaced by an increased number of fetal heart positive losses (6–13%). This was probably not surprising considering that paired HCG measurements in group C identified implantations which had initially started well (HCG >50 IU/l) to be followed by a slowing rise in HCG. It is likely that many of these losses were of chromosomally incompetent embryos. It was suggested (Schmidt-Sarosi et al., 1998Go) that there were differences in the distribution of chromosomal abnormalities in fetuses that aborted before the detection of a fetal heart and those that did so afterwards although they did not attempt to describe HCG profiles. Again multiple births were uncommon (2% of all multiples and 3% of all group C subjects).

Group D contained most of the successful singleton pregnancies and virtually all of the multiple pregnancies. Nearly 97% of women in this group had potentially viable outcomes in that a fetal heartbeat was subsequently seen on ultrasound scan. Of this subset 92% continued to have a birth of which 22% were multiple. Figure 2Go shows the ROC curve for day 14 HCG concentrations discriminating between singleton and multiple births. The sensitivity at 100 IU/l to predict multiple births was 85% with a specificity of 73%. These results suggest that given a day 14 HCG less than 100 IU/l the chance of having a multiple birth is only about 14%.

In conclusion, it is possible to make reasonable predictions of outcome after two estimations of HCG taken on days 14 and 21 after the procedure (egg collection or IUI). A day 21 HCG >200 IU/l is associated with a reasonable prognosis with around 73% of such patients achieving a birth. Taking account of the day 14 HCG concentrations can further refine the probabilities. If the day 14 HCG concentration is <50 IU/l the birth rate drops to 44% whereas if it is >50 IU/l the birth rate increases to 84%. Thus providers of fertility services can reassure normally pregnant patients as well as filter and manage those with non-viable outcomes more efficiently.


    Notes
 
1 To whom correspondence should be addressed

Back


    References
 Top
 Abstract
 Introduction
 Materials and methods
 Results
 Discussion
 References
 
Al-Sebai, M.A., Diver, M. and Hipkin, L.J. (1996) The role of a single free beta-human chorionic gonadotrophin measurement in the diagnosis of early pregnancy failure and the prognosis of fetal viability. Hum. Reprod., 11, 881–888.[Abstract]

Bjercke, S., Tanbo, T., Dale, O.P. et al. (1999) Human chorionic gonadotrophin concentrations in early pregnancy after in-vitro fertilization. Hum. Reprod., 14, 1642–1646.[Abstract/Free Full Text]

Buyalos, R.P., Glassman, L.M, Rifka, S.M. et al. (1992) Serum beta-human chorionic gonadotrophin, estradiol and progesterone as early predictors of pathological pregnancy. J. Reprod. Med., 37, 261–266.[ISI][Medline]

Chard, T. (1992) Pregnancy tests: a review. Hum. Reprod., 7, 701–710.[Abstract]

Check, J.H., Weiss, R.M. and Lurie, D. (1992) Analysis of serum human chorionic gonadotrophin concentrations in normal singleton, multiple and abnormal pregnancies. Hum. Reprod., 7, 1176–1180.[Abstract]

Fridstrom, M., Garoff, L., Sjoblom, P. et al. (1995) Human chorionic gonadotrophin patterns in early pregnancy after assisted reproduction. Acta Obstet. Gynecol. Scand., 74, 534–538.[ISI][Medline]

Glatstein, I.Z., Hornstein, M.D., Kahana, M.J. et al. (1995) The predictive value of discriminatory human chorionic gonadotrophin concentrations in the diagnosis of implantation outcome in in-vitro fertilisation cycles. Fertil. Steril., 63, 350–356.[ISI][Medline]

Guth, B., Hudelson, J., Higbie, J. et al. (1995) Predictive value of HCG concentration 14 days after embryo transfer. J. Assist. Reprod. Genet., 12, 13–14.[ISI][Medline]

Heiner, J.S., Kerin, J.F., Scmidt, L.L. et al. (1992) Can a single, early quantitative human chorionic gonadotrophin measurement in an in vitro-fertilisation gamete intrafallopian transfer program predict pregnancy outcome? Fertil. Steril., 58, 373–377.[ISI][Medline]

Ledger, W.L., Sweeting, V.M. and Chatterjee, S. (1994) Rapid diagnosis of early ectopic pregnancy in an emergency gynaecology service – are measurements of progesterone, intact and free beta human chorionic gonadotrophin helpful? Hum. Reprod., 9, 157–160.[Abstract]

Legro, R.S., Paulson, R.J., Lobo, R.A., et al. (1995) Association of early beta-human chorionic gonadotrophin values with pregnancy wastage and multiple implantation in a donor oocyte programme. Hum. Reprod., 10, 3293–3296.[Abstract]

Long, C.A., Whitworth, N.S., Murthy, H.M. et al. (1994) First trimester rapid semiquantitative assay for urine pregnanediol glucoronide predicts gestational outcome with the same diagnostic accuracy as serial human chorionic gonadotrophin measurements. Am. J. Obstet. Gynecol., 170, 1822–1827.[ISI][Medline]

Qasim, S.M., Callan, C. and Choe, J.K. (1996) The predictive value of an initial serum beta human chorionic gonadotrophin concentration for pregnancy outcome following in vitro fertilization. J. Assist. Reprod. Genet., 13, 705–708.[ISI][Medline]

Schmidt, L.L., Asch, R.H., Frederick, J.L. et al. (1994) The predictive value of a single beta human chorionic gonadotrophin in pregnancies achieved by assisted reproductive technology. Fertil. Steril., 62, 333–338.[ISI][Medline]

Schmidt-Sarosi, C., Scwartz, L.B., Lublin, J. et al. (1998) Chromosomal analysis of early pregnancy fetal losses in relation to transvaginal ultrasonographic detection of fetal heart motion after infertility. Fertil. Steril., 69, 274–277.[ISI][Medline]

Submitted on September 13, 1999; accepted on November 11, 1999.