Prediction of late failure after medical abortion from serial {beta}-hCG measurements and ultrasonography

C. Rørbye1,3, M. Nørgaard2 and L. Nilas1

1 Department of Obstetrics and Gynaecology, H:S Hvidovre Hospital, University of Copenhagen, 2650 Hvidovre and 2 Department of Obstetrics and Gynaecology, H:S Frederiksberg Hospital, University of Copenhagen, 2000 Frederiksberg, Denmark

3 To whom correspondence should be addressed at: Department of Obstetrics and Gynecology 537, H:S Hvidovre Hospital, University of Copenhagen, Kettegaard Alle 30, 2650 Hvidovre, Denmark. e-mail: christina.roerbye{at}hh.hosp.dk


    Abstract
 Top
 Abstract
 Introduction
 Materials and methods
 Results
 Discussion
 References
 
BACKGROUND: Surgical treatment of failed medical abortion may be performed several weeks after initiation of the abortion. There are no recognized methods for early identification of these late failures. We assessed the prognostic values of {beta}-hCG and ultrasonography in predicting late failure after medical abortion. METHODS: A total of 694 consecutive women with gestational age (GA) <63 days were followed up with {beta}-hCG measurements initially on the day of treatment and on day 8 and day 15, after the medically induced abortion with mifepristone and gemeprost. Measurement of the endometrial thickness by ultrasonography was performed on day 15 after induction of medical abortion. Failures diagnosed after day 15 and within 15 weeks were identified and classified as late failures. All interventions in this group were due to bleeding problems. The predictive values of different absolute and relative {beta}-hCG levels and different endometrial thickness were analysed. RESULTS: Of all failures in the study period, 65% (32) were diagnosed after the 2 week follow-up, and categorized as late failures. The absolute and the relative {beta}-hCG values on days 8 and 15 as well as the endometrial thickness on day 15 were greater among late failures than successes. Used as a predictive test, the positive predictive values of these variables were low. CONCLUSION: Neither {beta}-hCG nor endometrial thickness can be used clinically as diagnostic tests in predicting late failure after medical abortion.

Key words: {beta}-hCG/diagnostic test/failure/medical abortion/prediction


    Introduction
 Top
 Abstract
 Introduction
 Materials and methods
 Results
 Discussion
 References
 
The frequency of failure after medically induced abortions, defined as subsequent surgical intervention, is 2–6% for regimens using mifepristone/vaginal prostaglandin analogue (UK Multicentre Trial, 1990Go; Ashok et al., 1998Go, 2002; Schaff et al., 1999Go, 2000a,b). A noticeable number of failures are diagnosed after the usual 2 week follow-up (Ashok et al., 1998Go; Schaff et al., 2000bGo; Knudsen, 2001Go; Rørbye et al., 2003Go). Compared with surgical termination, medical abortion seems to be associated with a lower risk of pelvic inflammatory disease, probably due to the avoidance of instrumentation of the uterus (Henshaw et al., 1994Go; Winikoff et al., 1997Go; Jensen et al., 1999Go; Child et al., 2001Go). Incomplete medical abortion may, however, increase the risk of infection and is furthermore associated with discomfort as persistent or recurrent bleeding and pain. Early identification of women with retained tissue would allow early intervention. The risk of failure is related to high gestational age (McKinley et al., 1993Go), high body mass index (Grimes et al., 1990Go) and multiparity (Bartley et al., 2000Go). Few prospective studies have addressed prediction of failure after medical abortion, and no prospective tests have been suggested.

Failures diagnosed after the 2 week follow-up, defined as late failures, are associated with retained products of conception and heavy or prolonged bleeding. The absolute and/or relative {beta}-hCG might therefore be higher in failures than successes, and could perhaps be an indicator of late failure. This hypothesis is supported by Honkanen et al. (2002Go), who found that the decline in {beta}-hCG after a medically induced abortion is inversely correlated with the time taken to abort. The thickness of the endometrium is a useful parameter in diagnosing incomplete abortion after first trimester spontaneous abortion (Wong et al., 2002Go), and is also expected to be useful in diagnosing failure after medical abortion.

With this background, we wanted to compare the absolute and the relative {beta}-hCG values and the endometrial thickness after 1–2 weeks in successes and late failures after medical abortion, and to analyse the prognostic value of these variables.


    Materials and methods
 Top
 Abstract
 Introduction
 Materials and methods
 Results
 Discussion
 References
 
A total of 871 women aged >18 years had a medical abortion in the period of August 1999 to May 2001. The women were included at H:S Copenhagen University Hospital; Hvidovre and Frederiksberg. Contraindications for medical abortion in our facility and exclusion criteria were: asthma bronchiale requiring daily medication, lactation, suspicion of or verified ectopic pregnancy, coagulation disorder, insufficient language skills, multiple gestation and impairment of liver or kidney function. The medical regimen was mifepristone (Mifegyne®) 600 mg day 1 and vagitorium gemeprost (Cervagem®) 1 mg day 3, offered to women with gestational age ≤63 days, confirmed by ultrasonography. Pain and nausea were treated with morphine/metoclopramide supplemented with paracetamol/codeine if necessary. Serum {beta}-hCG was measured with a Sandwich Immuno Assay measuring total {beta}-hCG using the World Health Organization 3rd international standard reference preparation (detection limit <5 IU/ml).

In all, 694 women were followed-up with a clinical visit 2 weeks after the medical abortion. A transvaginal ultrasonography was performed (day 14–16) on a Siemens Sonoline SI-400 or SI-200 (5–7 mHz) measuring the maximum thickness of the endometrium in a sagittal dimension (double-layer including tissue in the cavity). Serum {beta}-hCG was measured initially and on day 15 (day 14–16). In one of the departments an additional measurement of {beta}-hCG was performed in 133 women on day 8 (day 7–9). In all, 177 women were excluded due to one or more of the following reasons: {beta}-hCG measurements on incorrect days (n = 90), performance of ultrasonography on incorrect days (n = 79) or surgical intervention before day 15 (n = 8). Late failure was defined as surgical intervention after day 15 and within 15 weeks. No interventions on women included in the study were performed as a consequence of either serum {beta}-hCG values or ultrasonographic images.

All women were followed up by a unique personal identification number in a computer system 15 weeks after initiation of the abortion, allowing identification of all surgical interventions anywhere within the country due to abortion-related complications. No women were lost to follow-up in this study.

The median values of the different variables for late failures and successes were compared by Mann–Whitney U-test. The positive and negative predictive values as well as sensitivity and specificity were calculated and analysed at different threshold limits for absolute {beta}-hCG values on days 1, 8 and 15 and for relative {beta}-hCG values on days 8 and 15, and for different endometrial thicknesses on day 15. To characterize each parameter with respect to its accuracy to detect failures, sensitivity and specificity were calculated from 2x2 contingency tables for each possible threshold value and plotted as receiver operating characteristic (ROC) curves. The areas under the curves (AUC) were calculated as a test for the hypothesis that AUC is not significantly different from 0.5, the area under the straight line (sensitivity = 1 – specificity) that represents no discrimination between failure and success.

The local ethics committee approved the study, and all patients gave informed consent.


    Results
 Top
 Abstract
 Introduction
 Materials and methods
 Results
 Discussion
 References
 
Women excluded from the study had more often had prior, induced abortions than women who were followed up as intended, but the two groups were otherwise comparable (Table I).


View this table:
[in this window]
[in a new window]
 
Table I. Baseline characteristics for women excluded from the study and women followed-up as intended
 
The overall success rate for all women in the study period was 94.4% (95% CI: 92.9–96.1) (n = 822/871). Efficacy was significantly higher at gestational age (GA) <49 days: 97.5% (95.0–99.0) than at 49<GA<56 days: 93.3% (90.5–96.1) and at 56<GA<63 days: 89.7% (84.8–94.6) (P < 0.01). Among the women included for further analysis in this study, parity and gestational age were higher in the failure group than in the success group. The remaining baseline characteristics were similar (Table I). Sixty-five per cent (32/49) of all failures were diagnosed after day 15 and defined as late failures. Of the 32 late failures, six interventions were due to acute heavy bleeding and 26 due to prolonged bleeding. There were no ongoing pregnancies and no interventions on patient request not related to bleeding. Late failures were diagnosed with a median interval of 23.5 days (range 15–103) after mifepristone administration. Twenty of the 32 failures were diagnosed after 30 days. Tissue from 20 women with late failure was examined histologically. Products of conception were found in 19 of these.

The absolute as well as the relative values of {beta}-hCG 1–2 weeks after the abortion were significantly higher in the late failure group than in the success group—with a wide range within each group and a great overlap of values between the two groups (Figures 1 and 2). Two women had higher {beta}-hCG values on day 8 compared with day 1, but both had successful abortions without surgical intervention. The median endometrial thickness measured by ultrasonography on day 15 was greater in the late failure than in the success group: 16 mm (13–18) versus 10 mm (7–13) (lower and upper quartiles, P < 0.0001).



View larger version (17K):
[in this window]
[in a new window]
 
Figure 1. Absolute serum {beta}-hCG (IU/l) day 1, 8 and 15 for failures (grey boxes) and successes (black boxes) after medical abortion shown on a log scale. The two women with higher {beta}-hCG values on day 8 compared with day 1 are excluded. The box indicates the lower and upper quartiles and the central line marks the median. The points at the end of the whiskers represent the range of the values. *P < 0.05, **P < 0.01 for failures and successes compared by Mann–Whitney U-test on days 8 and 15 respectively.

 


View larger version (14K):
[in this window]
[in a new window]
 
Figure 2. Relative serum {beta}-hCG (%) on days 8 and 15 for failures (grey boxes) and successes (black boxes) after medical abortion shown on a log scale. The two women with higher {beta}-hCG values on day 8 compared with day 1 are excluded. The box indicates the lower and upper quartiles and the central line marks the median. The points at the end of the whiskers represent the range of the values. *P < 0.01 for failures and successes compared by Mann–Whitney U-test.

 
Absolute and relative {beta}-hCG as well as endometrial thickness were poor predictors of late failure after medical abortion due to low positive predictive values in the range of 0.10–0.69. Plotted as ROC curves, AUC was 0.55 for initial {beta}-hCG, rising to AUC = 0.67 using absolute day 8 values and to AUC = 0.82 using absolute day 15 values (Figure 3). The discriminative ability of the relative {beta}-hCG values on days 8 and 15 gave similar curves (AUC = 0.88 and 0.78 for days 8 and 15 respectively). Using endometrial thickness on day 15 as predictive parameter resulted in an analogous curve (AUC = 0.73). For threshold levels representing the optimal points on all the curves, the corresponding positive predictive values were low, ranging from 0.07 to 0.30.



View larger version (21K):
[in this window]
[in a new window]
 
Figure 3. Sensitivity as a function of (1 – specificity) presented as receiver operating characterisic curves for absolute serum {beta}-hCG values on day 1 (n = 694), day 8 (n = 132) and day 15 (n = 694). Grey line: absolute values day 1, area under the curve (AUC) = 0.56; black line: absolute values day 15, AUC = 0.82; dotted line: absolute values day 8, AUC = 0.67.

 

    Discussion
 Top
 Abstract
 Introduction
 Materials and methods
 Results
 Discussion
 References
 
We have prospectively followed a large group of consecutive women undergoing medical abortion, and our failure rate is similar to other studies using the same regimen (UK Multicentre Trial 1990Go; Ashok et al., 1998Go; Schaff et al., 1999Go, 2000b; Ashok et al., 2002Go). Retained products of conception in 95% (19/20) of the curettages examined histologically indicate that the surgical interventions were medically justified. We do not know, however, how many of these would have resolved if left untreated. It is also possible that some of the cases categorized as successful would have had similar histological findings if a curettage had been performed. More than half of the failures were identified after day 15, reflecting a long and tiring course of bleeding and/or pain. Being able to diagnose these failures earlier would optimize the medical abortion procedure.

The overall percentage decline in {beta}-hCG is consistent with earlier findings (Walker et al., 2001; Honkanen et al., 2002Go). Both the absolute and the relative values of {beta}-hCG as well as the endometrial thickness after medical abortion were higher in women who turned out to be late failures than in successfully treated women. However, we conclude that none of these variables can be used as diagnostic tests in predicting late failure after medical abortion.

The demands for a diagnostic test depend on the clinical situation, the consequences of a positive test result, and the risks connected with an overlooked condition. Diagnosing a failed medical abortion leads to surgical intervention often performed under general anaesthesia. To avoid unnecessary interventions, the positive predictive value of the applied test must be high. The specificity should not be <0.95, which is the approximate chance of a successful course using this medical abortion regimen.

We found that threshold levels of {beta}-hCG and endometrial thickness with high positive predictive values had a low sensitivity, leading to identification of only a minority of the failures. As an example: using {beta}-hCG >5000 on day 15 as a threshold, the specificity was 0.99, the positive predictive value was 0.69, but the sensitivity only 0.28. Used in a clinical situation, this means that 31% of performed interventions would be unnecessary, while 72% of the failures would be missed. Higher positive predictive values were achieved by combining changes in {beta}-hCG and endometrial thickness, but still at the expense of sensitivity. As an example: a {beta}-hCG value >5000 on day 8 combined with an endometrial thickness >10 mm on day 15 gave the positive predictive value 0.75, but the corresponding sensitivity was 0.28 and the specificity 1. Similar values were seen with other combinations of different threshold levels.

We found that none of the analysed parameters was acceptable as a diagnostic test because of the low positive predictive values. The consequences of overlooking a failure are limited, as the risk of serious morbidity associated with retained tissue is minimal and because failures ultimately will be revealed clinically. With this background we conclude that the analysed variables used as diagnostic tests would lead to an unacceptable number of unnecessary interventions. If a positive test result could lead to a simple, risk-free, definitive test, a higher false positive rate could be acceptable, but such a definitive test has not been proven.

Since we did not intervene on either {beta}-hCG values or on specific endometrial thicknesses, but followed the spontaneous courses after a 2 week follow-up, we were able to analyse the prognostic value of {beta}-hCG and ultrasonography as predictors of late failure. Serum {beta}-hCG was measured according to the internationally accepted standard reference, but the absolute values may vary with different assays. The use of relative changes allows comparison between centres using different assays. Resident doctors under the supervision of a specialist doctor performed the ultrasonography at follow-up. We have not analysed the validity of the measurements of the endometrial thickness and therefore cannot exclude a possible inter-observational discrepancy. To minimize this, all residents had to qualify before authorization was given to perform ultrasound examinations at our department. Furthermore, doctors were unaware of the final outcome. There is no reason to believe that a systematic error has been made, but rather a random incorrectness that does not influence the conclusion. Moreover, our findings are consistent with endometrial thicknesses after a tamoxifen/misoprostol regimen (Harwood et al., 2001Go).

Some of the cases excluded in this study were unavoidable, because surgical intervention was performed before follow-up on day 15, but the majority of exclusions was caused by a lack of serial {beta}-hCG values or ultrasonography performed on the correct days. Similar frequencies of lost or incorrect follow-up after induced abortion are a recognized problem (Thonneau et al., 1998Go). However, the baseline characteristics were similar in the women included and those women excluded, and it is therefore unlikely that results from the excluded women would affect the overall conclusion. Comparing women in the failure group with women in the success group revealed more parous women among failures, which correspond to the increased risk of failure with increased parity (Bartley et al., 2000Go).

Following mifepristone administration, {beta}-hCG continues to increase, but declines precipitously after administration of misoprostol (Somell et al., 1990Go; World Health Organization Task Force, 1991Go; Walker et al., 2001Go; Honkanen et al., 2002Go). The finding of significantly higher absolute and relative {beta}-hCG values among late failures compared with successes on days 8 and 15 after a mifepristone/prostaglandin regimen has, to our knowledge, not been published previously. The only published predictive analysis concluded that a decrease in {beta}-hCG of <50% within 24 h after misoprostol administration in methotrexate pre-treated women was associated with an increased risk of ending as a failure, but the positive and negative predictive values were not given (Creinin, 1996Go).

In conclusion, absolute and relative {beta}-hCG values as well as endometrial thickness were significantly higher in late failures than successes after medical abortion, but none of these parameters is clinically useful as a diagnostic test in predicting late failure after medical abortion. {beta}-hCG measurement and ultrasonography can be used to identify ongoing pregnancies, and might be helpful in a clinical situation of doubt, but are merely supplements to the general clinical evaluation. After ongoing pregnancy is excluded, only the clinical course will reveal the remaining failures. The women must therefore be informed to return for a clinical examination in cases of acute or prolonged bleeding, pain or fever, as these signs may be indicators of failure or other complications that need treatment.


    References
 Top
 Abstract
 Introduction
 Materials and methods
 Results
 Discussion
 References
 
Ashok PW, Penney GC, Flett GM et al (1998) An effective regimen for early medical abortion: a report of 2000 consecutive cases. Hum Reprod 13,2962–2965.[Abstract/Free Full Text]

Ashok PW, Templeton A, Wagaarachchi PT et al (2002) Factors affecting the outcome of early medical abortion: a review of 4132 consecutive cases. Br J Obstet Gynecol 109,1281–1289.

Aubeny E, Peyron R, Turpin CL et al (1995) Termination of early pregnancy (up to 63 days of amenorrhea) with mifepristone and increasing doses of misoprostol [corrected]. Int. J. Fertil. Menopausal Stud., 40(Suppl 2),85–91.[Medline]

Bartley J, Tong S, Everington D et al (2000) Parity is a major determinant of success rate in medical abortion: a retrospective analysis of 3161 consecutive cases of early medical abortion treated with reduced doses of mifepristone and vaginal gemeprost. Contraception 62,297–303.[CrossRef][ISI][Medline]

Bartley J, Brown A, Elton R et al (2001) Double-blind randomized trial of mifepristone in combination with vaginal gemeprost or misoprostol for induction of abortion up to 63 days gestation. Hum Reprod 16,2098–2102.[Abstract/Free Full Text]

Child TJ, Thomas J., Rees M. et al (2001) A comparative study of surgical and medical procedures: 932 pregnancy terminations up to 63 days gestation. Hum Reprod 16,67–71.[Abstract/Free Full Text]

Creinin MD (1996) Change in serum beta-human chorionic gonadotropin after abortion with methotrexate and misoprostol. Am J Obstet Gynecol 174,776–778.[ISI][Medline]

Grimes DA, Bernstein L, Lacarra M et al (1990) Predictors of failed attempted abortion with the antiprogestin mifepristone (RU 486). Am J Obstet Gynecol 162,910–915.[ISI][Medline]

Harwood B, Meckstroth KR, Mishell DR et al (2001) Serum beta-human chorionic gonadotropin levels and endometrial thickness after medical abortion. Contraception 63,255–256.[CrossRef][ISI][Medline]

Henshaw RC, Naji SA, Russell IT et al (1994) A comparison of medical abortion (using mifepristone and gemeprost) with surgical vacuum aspiration: efficacy and early medical sequelae. Hum Reprod 9,2167–2172.[Abstract]

Honkanen H, Ranta S, Ylikorkala O et al (2002) The kinetics of serum hCG and progesterone in response to oral and vaginal administration of misoprostol during medical termination of early pregnancy. Hum Reprod 17,2315–2319.[Abstract/Free Full Text]

Jensen JT, Astley SJ, Morgan E et al (1999) Outcomes of suction curettage and mifepristone abortion in the United States. A prospective comparison study. Contraception 59,153–159.[CrossRef][ISI][Medline]

Kahn JG, Becker BJ, MacIsaa L et al (2000) The efficacy of medical abortion: a meta-analysis. Contraception 61,29–40.[CrossRef][ISI][Medline]

Knudsen UB (2001) First trimester abortion with mifepristone and vaginal misoprostol. Contraception 63,247–250.[CrossRef][ISI][Medline]

McKinley C, Thong KJ and Baird DT (1993) The effect of dose of mifepristone and gestation on the efficacy of medical abortion with mifepristone and misoprostol. Hum Reprod 8,1502–1505.[Abstract]

Rørbye C, Nørgaard M and Nilas L (2003) Medical abortion—defining success and categorizing failures. Contraception, 68,247–251.[CrossRef][ISI][Medline]

Schaff EA, Eisinger SH, Stadalius LS et al (1999) Low-dose mifepristone 200 mg and vaginal misoprostol for abortion. Contraception 59,1–6.[CrossRef][ISI][Medline]

Schaff EA, Fielding SL, Westhoff C et al (2000a) Vaginal misoprostol administered 1, 2, or 3 days after mifepristone for early medical abortion: a randomized trial. J Am Med Assoc 284,1948–1953.[Abstract/Free Full Text]

Schaff EA, Fielding SL, Eisinger SH et al (2000b) Low-dose mifepristone followed by vaginal misoprostol at 48 hours for abortion up to 63 days. Contraception 61,41–46.[CrossRef][ISI][Medline]

Somell C, Olund A, Carlstrom K et al (1990) Reproductive hormones during termination of early pregnancy with mifepristone. Gynecol Obstet Invest 30,224–227.[ISI][Medline]

Spitz, I.M., Bardin C.W., Benton, L. et al (1998) Early pregnancy termination with mifepristone and misoprostol in the United States. N Engl J Med 338,1241–1247.[Abstract/Free Full Text]

Thonneau P, Fougeyrollas B, Ducot B et al (1998) Complications of abortion performed under local anesthesia. Eur J Obstet Gynecol Reprod Biol 81,59–63.[CrossRef][ISI][Medline]

UK Multicentre Trial (1990) The efficacy and tolerance of mifepristone and prostaglandin in first trimester termination of pregnancy. Br J Obstet Gynecol 97,480–486.[ISI][Medline]

Walker K, Schaff E, Fielding S et al (2001) Monitoring serum chorionic gonadotropin levels after mifepristone abortion. Contraception 64,271–273.[CrossRef][ISI][Medline]

Winikoff B, Sivin I, Coyaji KJ et al (1997) Safety, efficacy, and acceptability of medical abortion in China, Cuba, and India: a comparative trial of mifepristone–misoprostol versus surgical abortion. Am J Obstet Gynecol 176,431–437.[ISI][Medline]

Wong SF, Lam MH and Ho LC (2002) Transvaginal sonography in the detection of retained products of conception after first-trimester spontaneous abortion. J Clin Ultrasound 30,428–432.[CrossRef][ISI][Medline]

World Health Organization Task Force (1991) Pregnancy termination with mifepristone and gemeprost: a multicenter comparison between repeated doses and a single dose of mifepristone. Fertil Steril 56,32–40.[ISI][Medline]

Submitted on September 23, 2002; resubmitted on August 20, 2003; accepted on September 23, 2003.





This Article
Abstract
FREE Full Text (PDF )
Alert me when this article is cited
Alert me if a correction is posted
Services
Email this article to a friend
Similar articles in this journal
Similar articles in ISI Web of Science
Similar articles in PubMed
Alert me to new issues of the journal
Add to My Personal Archive
Download to citation manager
Search for citing articles in:
ISI Web of Science (2)
Request Permissions
Google Scholar
Articles by Rørbye, C.
Articles by Nilas, L.
PubMed
PubMed Citation
Articles by Rørbye, C.
Articles by Nilas, L.