Contraception 76 (2007) 101 – 104

Original research article

Accuracy of a semi-quantitative urine pregnancy test compared to

serum beta-hCG measurement: a possible screening tool for ongoing
pregnancy after medication abortion
Daniel Grossman a,⁎, Karla Berdichevsky b , Fernando Larrea c , Jorge Beltran d
a
Ibis Reproductive Health, c/o Department of Obstetrics and Gynecology, San Francisco General Hospital, San Francisco, CA 94110, USA
b
Population Council, Regional Office for Latin America and the Caribbean, Mexico City, DF 04000, Mexico
c
Department of Reproductive Biology, Instituto Nacional de Ciencias Medicas y Nutricion Salvador Zubirán, Mexico City, DF 14000, Mexico
d
Instituto Nacional de Perinatologia Isidro Espinosa de los Reyes, Mexico City, DF 11000, Mexico
Received 10 October 2006; revised 16 April 2007; accepted 22 April 2007

Abstract

Purpose: Medication abortion protocols commonly rely on ultrasound or serum hCG measurement to confirm completion. In order to
explore the use of a urine-based test to screen for ongoing pregnancy, we compared the diagnostic accuracy of a recently developed semi-
quantitative urine pregnancy test to serum β-hCG testing.
Methods: We evaluated the urine test with 97 women in early pregnancy at a hospital and private clinic in Mexico City. The results of the
urine test (hCG level N or b1000 IU/L) were correlated with those of a serum quantitative β-hCG immunoradiometric assay.
Results: The sensitivity of the urine test to identify individuals with a serum β-hCG level N1000 IU/L was 88.6% (95% CI 74.6– 95.7%),
and its specificity was 71.7% (95% CI 57.4–82.8%).
Conclusion: The reasonably high sensitivity of this urine test suggests it might be useful as a screening test to detect ongoing pregnancy after
medication abortion. Future research should evaluate its utility in clinical follow-up protocols.
© 2007 Elsevier Inc. All rights reserved.

Keywords: Sensitivity and specificity; Urine pregnancy test; Abortion; hCG

1. Introduction Several studies have examined the utility of using serum

β-hCG levels to confirm completion after medication
Most randomized controlled trials of medication abortion abortion. Thonneau et al. [7] reported a significant decrease
rely on ultrasound to confirm completion of the procedure in serum β-hCG levels 2 weeks after mifepristone-
[1]. Cohort trials studying the mifepristone regimen in misoprostol medication abortion. Among the 25 women
developing countries such as India and Vietnam have used with a failed abortion either due to retained tissue or
protocols that rely on clinical assessment of completion continuing pregnancy, 23 had a β-hCG level of greater than
[2,3], but this has not been systematically studied. As more 500 IU/L [7]. Another study reported a 94% decline in
nongynecologists begin to provide medication abortion in β-hCG levels by Day 8, and by Day 14, the hCG level was
developed countries [4], and as the method becomes more found to have decreased to 1% of its original level [8]. In a
commonly used in resource-poor settings in developing third study of 34 women who underwent mifepristone-
countries [2,3,5], research is needed to reduce the reliance on misoprostol medication abortion, β-hCG levels had declined
ultrasound [6]. by more than 99% by Day 14 [9]. Only three subjects had
β-hCG concentrations above 1000 IU/L on Day 14, of which
one was the only abortion failure (serum β-hCG level of
3560 IU/L) [9].
⁎ Corresponding author. Fax: +1 415 206 4527. Commonly available highly sensitive urine pregnancy
E-mail address: [email protected] (D. Grossman). tests have a threshold of detection of approximately 10–25
0010-7824/$ – see front matter © 2007 Elsevier Inc. All rights reserved.
doi:10.1016/j.contraception.2007.04.008
102 D. Grossman et al. / Contraception 76 (2007) 101–104

IU/L. One study of standard urine hCG testing after (Orchid Biomedical Systems, Goa, India) was performed
medication abortion reported that the mean number of days within 10 min by one of four trained clinicians. Urine strip
until the test was negative was 31 days after starting the tests were dipped in the sample for 10–15 s and placed
procedure, which was significantly longer than the mean horizontally on a flat surface. They were interpreted 5 min
duration of follow-up when completion was judged by after the test was conducted. There were four possible results
transvaginal ultrasound [10]. A less sensitive urine of the urine test: negative, hCG b1000 IU/L, hCG N1000 IU/
pregnancy test with a higher threshold for detection of L or indeterminate. The results of the serum β-hCG test were
hCG might be more useful to confirm completion at a not available at the time the urine test was interpreted. The
follow-up visit on Day 14 [11]. A recent study examined urine test result was not given to the clinicians managing the
the accuracy of low-sensitivity and high-sensitivity urine care of the woman.
pregnancy tests compared to ultrasonography in a medica- A whole blood sample was collected immediately after
tion abortion trial and found that both tests had very high obtaining the urine sample and refrigerated. At the end of
negative predictive values (96–100%), but due to high each day, all specimens were centrifuged and serum samples
proportions of false-positive results, they had very low were frozen at −20°C. Determination of serum β-hCG was
positive predictive values (1–2%) [12]. performed in the Department of Reproductive Biology at the
A semi-quantitative urine dipstick pregnancy test (Orchid Instituto Nacional de Ciencias Medicas y Nutrición Salvador
Biomedical Systems, Goa, India) was recently developed Zubirán using a commercially available immunoradiometric
that has a positive control, a band sensitive to 10 IU/L and assay (IZOTOP Institute of Isotopes Co., Ltd., Budapest,
another sensitive to 1000 IU/L. The objective of this Hungary) in batches with standard controls. A single
prospective study was to compare the diagnostic accuracy laboratory technician with extensive experience with this
of this semi-quantitative urine test to serum β-hCG test performed all β-hCG measurements. The laboratory did
measurement to determine whether it merited future not have access to the urine test results. Each test was
evaluation as a screening tool for detecting ongoing performed in duplicate, and the mean of the two values was
pregnancy after medication abortion. reported as the result. The participant's attending physician
was notified of the quantitative β-hCG result as soon as it
was available. Patients were either reimbursed for their
2. Materials and methods clinical visit costs (up to US$5) or given the equivalent of US
$5 to reimburse them for travel costs.
Women were recruited for this study who presented The manufacturer donated 100 urine tests for evaluation,
during early pregnancy at a hospital and private clinic in and we anticipated that 10% of the tests might be lost during
Mexico City. Women were recruited at their first prenatal training or due to specimen errors. We hypothesized that the
visit who were assessed by the physician to have a normal sensitivity of the urine test to identify an individual with a
pregnancy of 8 weeks' gestation or less, as determined by serum β-hCG N1000 IU/L was 90% and that its specificity
report of last menstrual period (LMP) and confirmed by was 95%. With approximately even numbers in the true-
physical exam and in some cases ultrasound. Women were positive and true-negative groups and alpha of 5%, recruiting
also recruited during a clinical visit for treatment or follow- a sample size of 90 women would result in 95% confidence
up after a spontaneous or induced abortion or otherwise intervals around the measurement of sensitivity of ±8.8%
abnormal pregnancy. Women undergoing abortion were and around the measurement of specificity of ±6.4%. Ninety-
recruited not more than 3 weeks after the process began. five percent confidence intervals were calculated according
Participants were recruited consecutively on the days the to the efficient-score method [13].
study clinician was present at the facility. All participants
were at least 18 years old. The study was approved by the
institutional review boards of the Population Council and the 3. Results
hospital where women were recruited.
After informed consent was obtained, information on the One urine test was lost due to specimen error, leaving 99
woman's age, gravidity, parity, LMP and relevant informa- tests available for the study. Ninety-nine women in early
tion on her current pregnancy or abortion (i.e., threatened,
complete or incomplete, post-abortion follow-up) was
obtained. If the woman underwent a curettage or aspiration Table 1
Serum β-hCG results according to the clinical presentation of participants
or had taken misoprostol to induce an abortion at the current
visit or previously for this pregnancy, this was also recorded. Serum β-hCG result
Because abortion is legally restricted in Mexico, it was not b1000 IU/L N1000 IU/L
always possible to ascertain whether the abortion cases were Normal pregnancy ≤8 weeks' gestation 16 30
spontaneous or induced. Abnormal pregnancy or spontaneous abortion 15 9
Immediately after recruitment, a urine sample was self- Post-abortion care 22 5
Total 53 44
collected, and the semi-quantitative urine pregnancy test
 D. Grossman et al. / Contraception 76 (2007) 101–104 103

Table 2 them and would miss 11. If 100,000 women were to use
Results of the urine semi-quantitative pregnancy test compared to serum the misoprostol alone regimen, approximately 5000
β-hCG result
ongoing pregnancies would result, and this test would
Urine test Serum β-hCG result miss 550 of them. Using this test alone might be
result
b1000 IU/L N1000 IU/L acceptable with the mifepristone regimen but might not
b1000 IU/L 38 5 be for the less effective misoprostol alone regimen,
N1000 IU/L 15 39 provided ultrasound follow-up is feasible. If ultrasound is
Total 53 44 not available in developing country settings, due to
distance or cost, the imperfect accuracy of the urine test
might be more acceptable.
However, it might be possible to increase the sensitivity
pregnancy were recruited into the study from March to May
of urine testing by performing sequential tests. Although still
2005, and the mean age of participants was 27 years. One
unproven, one possible protocol would be to perform the
recruited patient was excluded from the study due to
semi-quantitative test 2 weeks after medication abortion and
difficulties in obtaining a blood sample, and data from
refer women with a positive result for ultrasound. Women
another woman were eliminated because her blood sample
with a negative initial urine test but who do not experience
was lost, resulting in 97 women analyzed here. Forty-six
normal menses 1 month after the first test could then undergo
women entered the study at 8 weeks or less of normal
a standard highly sensitive urine pregnancy test. Highly
gestation, 50 participants were undergoing (or had com-
sensitive urine tests usually turn negative by 1 month after
pleted) a spontaneous or induced abortion, and one
medication abortion [10], and a persistent positive result
participant had an ectopic pregnancy. In no case was the
would need to be evaluated with ultrasound.
urine test result indeterminate, although when present, the
This protocol would likely reduce the need for ultrasound
color of the 1000 IU/L band always appeared lighter than
after medication abortion. Assuming that the semi-quantita-
the other bands. There were no adverse events from
tive urine test's specificity is at the lowest range of the
collecting the urine or blood samples.
reported confidence interval, 57% of women with serum
Table 1 shows the distribution of patients according to
β-hCG levels below 1000 IU/L would have a negative urine
their serum β-hCG results and clinical diagnosis at recruit-
test. If 10% of women at Day 14 after medication abortion
ment. The sensitivity and specificity of the urine pregnancy
have a serum β-hCG level above 1000 IU/L [9], approxi-
test to determine whether an individual had a serum β-hCG
mately half of women would have a negative semi-
value N1000 IU/L were 88.6% (95% CI 74.6–95.7%) and
quantitative urine test and would not be referred for
71.7% (95% CI 57.4–82.8%), respectively (see Table 2).
ultrasound. Reducing the need for such an expensive
Forty-four of 97 samples had serum β-hCG results above
technology could significantly improve access to a safe
1000 IU/L. Of these, the urine test result corresponded to the
abortion technique in low resource settings.
serum result in 39 cases (true positives) but failed to detect
A recent study examined the accuracy of low-sensitivity
five cases in which the serum β-hCG level was N1000 IU/L
and high-sensitivity pregnancy tests to predict the ultra-
(false negatives). These five cases had serum β-hCG levels
sound presence of a gestational sac at a follow-up visit on
of 1252, 2173, 3566, 13,230 and 16,034 IU/L.
Days 6–8 or Days 12–16 after mifepristone [12]. Both
tests were found to have a high proportion of false-positive
4. Discussion results [12]. Because this study was performed in the
context of a clinical trial of the mifepristone regimen in
The results of the semi-quantitative urine pregnancy test which there were very few cases of ongoing pregnancy, the
correlated reasonably well with serum β-hCG measure- authors were unable to assess the sensitivity of urine testing
ments among women in early pregnancy, approximating to detect this outcome [12]. The authors conclude that urine
the a priori estimate for sensitivity but failing to meet the testing after medication abortion is of limited clinical
estimate for specificity. The reasonably high sensitivity of utility, although they note that a negative test result could
the test suggests that it could be a useful screening tool to be used to exclude a proportion of women from further
detect ongoing pregnancy after medication abortion. But evaluation [12]. While this study questions the utility of
what minimum level of sensitivity is acceptable for such a urine testing in the context of medication abortion with the
screening test? After the mifepristone regimen, approxi- mifepristone regimen in the United States, it should be
mately 0.1% of pregnancies is ongoing [14], while the noted that there are no published validation data for the
misoprostol alone regimen, which is commonly used in low-sensitivity urine test used by these authors. Although
developing countries where mifepristone is not available studied in a different context with a different standard, we
[5], results in approximately 5% of women having an found the Orchid semi-quantitative urine test to have a
ongoing pregnancy [15]. If 100,000 women were to use higher measured specificity (71%) than that reported for
the mifepristone regimen, approximately 100 ongoing the low-sensitivity test used in the prior study (39% at
pregnancies would result; this test would detect 89 of Days 12–16) [12].
104 D. Grossman et al. / Contraception 76 (2007) 101–104

It is important to note that there were five cases in which References

the urine test indicated an hCG level of b1000 IU/L even
though the serum β-hCG measured N1000 IU/L. These false [1] von Hertzen H, Honkanen H, Piaggio G, et al. WHO multinational
negatives, as well as the light color of the 1000 IU/L band, study of three misoprostol regimens after mifepristone for early
medical abortion: I. Efficacy. BJOG 2003;110:808–18.
raise concerns about the test's performance and point toward [2] Coyaji K, Elul B, Krishna U, et al. Mifepristone-misoprostol abortion:
ways it could be improved. It is also possible that the false a trial in rural and urban Maharashtra, India. Contraception 2002;
negatives were related to urine dilution due to fluid intake, 66:33–40.
and future evaluations of this test should consider using the [3] Ngoc NT, Nhan VQ, Blum J, Mai TT, Durocher JM, Winikoff B. Is
home-based administration of prostaglandin safe and feasible for
first morning voided urine sample, which is generally the
medical abortion? Results from a multisite study in Vietnam. BJOG
most concentrated. Because samples were obtained in this 2004;111:814–9.
study whenever a woman presented for care, it was not [4] Prine LW, Lesnewski R. Medication abortion and family physicians'
possible to obtain first voided specimens. An additional scope of practice. J Am Board Fam Pract 2005;18:304–6.
limitation of this study is that accurate information about a [5] Lafaurie MM, Grossman D, Troncoso E, Billings DL, Chávez S.
history of induced abortion was not ascertainable due to the Women's perspectives on medical abortion in Mexico, Colombia,
Ecuador and Peru: a qualitative study. Reprod Health Matt
legal status of abortion in Mexico. 2005;13:75–83.
The data presented here suggest that the semi-quantita- [6] Clark W, Gold M, Grossman D, Winikoff B. Can mifepristone medical
tive urine pregnancy test should be further tested in clinical abortion be simplified? A review of the evidence and questions for
trials with medication abortion to evaluate its utility as a future research. Contraception 2007;75:245–50.
screening tool for ongoing pregnancy. Given the rarity of [7] Thonneau P, Fougeyrollas B, Spira A. Analysis of 369 abortions
conducted by mifepristone (RU486) associated with sulprostone in a
ongoing pregnancy with the mifepristone regimen, urine French family planning center. Fertil Steril 1994;61:627–31.
testing should be further evaluated in settings where [8] Walker K, Schaff E, Fielding S, Fuller L. Monitoring serum chorionic
misoprostol alone is the only available medication abortion gonadotropin levels after mifepristone abortion. Contraception
option. In addition to reducing reliance on ultrasound in 2001;64:271–3.
[9] Honkanen H, Ranta S, Ylikorkala O, Heikinheimo O. The kinetics of
clinic settings, urine testing also would give women who
serum hCG and progesterone in response to oral and vaginal
self-medicate with misoprostol an additional tool to help administration of misoprostol during medical termination of early
them know if their abortion has failed. As self-medication pregnancy. Hum Reprod 2002;17:2315–9.
becomes increasingly more common in Latin America and [10] Schaff EA, Eisinger SH, Franks P, Kim SS. Methotrexate and
elsewhere in response to restricted access to abortion [16], misoprostol for early abortion. Fam Med 1996;28:198–203.
it is critical that research address how to make this practice [11] Grossman D, Ellertson C, Grimes DA, Walker D. Routine follow-up
visits after first-trimester induced abortion. Obstet Gynecol 2004;
as safe as possible. 103:738–45.
[12] Godfrey EM, Anderson A, Fielding SL, Meyn L, Creinin MD. Clinical
Acknowledgments utility of urine pregnancy assays to determine medical abortion
outcome is limited. Contraception 2007;75:378–82.
[13] Clinical Calculator 1. From an observed sample: estimates of
We would like to thank Mariana Castillo for her role in population prevalence, sensitivity, specificity, predictive values, and
data collection and Mario Cardenas for performing the likelihood ratios. Available at: http://faculty.vassar.edu/lowry/clin1.
laboratory assays, as well as the participating clinic and html [Accessed 15 March 2006].
hospitals for their collaboration in patient recruitment, [14] Hausknecht R. Mifepristone and misoprostol for early medical
abortion: 18 months experience in the United States. Contraception
laboratory assistance and general logistical support. This
2003;67:463–5.
study was funded by Ibis Reproductive Health, Gynuity [15] Jain JK, Dutton C, Harwood B, Meckstroth KR, Mishell Jr DR. A
Health Projects and an anonymous donor. Orchid Biomedi- prospective randomized, double-blinded, placebo-controlled trial
cal Systems donated the urine tests used in the study but comparing mifepristone and vaginal misoprostol to vaginal misopros-
provided no further remuneration to the research institutions tol alone for elective termination of early pregnancy. Hum Reprod
2002;17:1477–82.
or the individual researchers. None of the research institu-
[16] Sherris J, Bingham A, Burns MA, Girvin S, Westley E, Gomez PI.
tions or individual researchers involved in the study has any Misoprostol use in developing countries: results from a multicountry
financial interest in Orchid Biomedical Systems. study. Int J Gynaecol Obstet 2005;88:76–81.