Contraception xx (2017) xxx–xxx

Original research article Serial multilevel urine pregnancy testing to assess medical outcome: a meta-analysis☆,☆☆ ⁎ Elizabeth G. Raymonda, , Tara Shocheta, Jennifer Bluma, Wendy R. Sheldona, Ingrida Plataisa, Hillary Brackena, Rasha Dabasha, Mark A. Weaverb, Nguyen Thi Nhu Ngocc, Paul D. Blumenthald, Beverly Winikoffa aGynuity Health Projects, 15 E 26th Street, Suite 801, New York, NY, 10010, USA bUniversity of North Carolina at Chapel Hill, Departments of Medicine and Biostatistics, Chapel Hill, NC, USA cCenter for Research and Consultancy in Reproductive Health, Ho Chi Minh City, Vietnam dStanford University Medical Center, 300 Pasteur Drive, Stanford, CA, USA Received 27 September 2016; revised 20 December 2016; accepted 22 December 2016

Abstract

Objectives: To summarize data on the accuracy of a strategy designed to exclude ongoing pregnancy after treatment by observing a decline in urine human chorionic gonadotropin (hCG) concentration as estimated by multilevel urine pregnancy tests (MLPTs) performed before and after treatment. Study design: We collated original data from seven studies performed by our organization that evaluated the accuracy of the MLPT strategy for assessment of outcome of medical abortion. Our first analysis included data from the five studies in which each participant was evaluated both with the MLPT strategy and with ultrasound or other clinical assessment. Our second analysis combined data from two randomized trials that compared the MLPT strategy to assessment by ultrasound. Both analyses included only participants treated at ≤63 days of gestation. Results: In the first analysis, 1482 (93%) of 1599 participants had a decline in hCG concentration after treatment. Twenty-one (1.3%) had an ongoing pregnancy, none of whom had a decline (predictive value 100%, 95% CI 93.3%, 100%). The remaining 96 women (6.0%) had no decline without an ongoing pregnancy. The second analysis, which included 3762 participants with follow-up, found no significant difference in the rates of ongoing pregnancy ascertained in the randomized groups (RR 0.88; 95% CI 0.50, 1.54). Nearly all of the post-treatment MLPTs in the seven studies (3484/3535; 99%) were performed by the participants themselves. Conclusions: Serial multilevel urine pregnancy testing is a highly reliable and efficient strategy for excluding ongoing pregnancy after medical abortion at≤63 days of gestation. Implications statement: Serial urine testing using MLPTs can obviate the need for routine ultrasound or examination after medical abortion treatment and can allow most women to avoid an in-person follow-up visit to the abortion facility. © 2017 Elsevier Inc. All rights reserved.

Keywords: Medical abortion; Diagnostic accuracy; Multilevel pregnancy test; Follow-up; Semiquantitative pregnancy test; Human chorionic gonadotropin

1. Introduction examination, or serum pregnancy testing. These tests are costly and may be inconvenient, particularly for women who Medical abortion providers commonly ask women to live far away or have competing commitments, and they return one or two weeks after taking the drugs must be performed by specially trained personnel. Moreover, to confirm pregnancy termination with ultrasound, pelvic they provide no clinical benefit to the vast majority of women who have successful uncomplicated [1]. Over the past decade, Gynuity Health Projects has ☆ investigated an alternative strategy for follow-up after early Disclosure of interests: Authors report no conflicts of interest. ☆☆ Funding: This research was funded by an anonymous charitable donor. medical abortion. In this strategy, a semiquantitative multilevel ⁎ Corresponding author. Tel.: +1 212 448 1230; fax: +1 212 448 1260. pregnancy test (MLPT) that measures the approximate E-mail address: [email protected] (E.G. Raymond). concentration of human chorionic gonadotropin (hCG) in http://dx.doi.org/10.1016/j.contraception.2016.12.004 0010-7824/© 2017 Elsevier Inc. All rights reserved. 2 E.G. Raymond et al. / Contraception xx (2017) xxx–xxx urine is performed before and after the woman takes the abortifacient drugs. A decline in concentration is interpreted to indicate that no ongoing pregnancy exists, whereas a stable or [7]

rising concentration indicates need for further evaluation. The 2012 strategy is not intended to identify persistent non-viable sacs or – Study G 63 2010 Group 1: MLPT plus ultrasound if indicated Group 2: exam, ultrasound if indicated Uzbekistan, Moldova incomplete abortions. The primary aim of our research was to 2 weeks evaluate the accuracy of this strategy using a specific MLPT, dBest® (AmeriTek, Seattle WA, USA), which estimates the urinary hCG concentration asb25, 25–99, 100–499, 500–1999, – ≥ 2000 9999, or 10,000 mIU/ml. Here we summarize the [6] 2011 resultsofthiswork. – Study F 63 Group 1: MLPT plus ultrasound if indicated Group 2: ultrasound Vietnam

2. Material and methods

Gynuity Health Projects has completed eight studies that [5] 2014 2010 evaluated the MLPT strategy using the dBest test for – Study E 63 2013 Group 1: MLPT plus ultrasound Group 2: HSPT plus ultrasound Vietnam assessment of medical abortion outcome, six of which have 3, 7, 14 days 2 weeks been published [2–7]. We excluded one unpublished study from this report because it was not designed to ascertain ongoing pregnancy as a specific reason for intervention after abortifacient treatment. The other seven studies enrolled women in six countries who presented for outpatient medical [4] 2014 abortion with and between 2010 – 14 days and 2014 (Table 1). – 400 mcg sublingual 800 mcg buccal 800 mcg buccal 400 mcg sublingual 70 2013 and/or exam Tunisia 10 Studies A–D were non-comparative studies designed to assess the diagnostic accuracy of the strategy. Each participant had an MLPT before ingesting mifepristone and was given a second test to perform at home 1 week (study A [2]) or 2 weeks (studies B [3], C (unpublished study), and D [4]) later. She reported the result at a scheduled in-person clinic visit. If a participant had not done the test at home, a test was performed at the clinic. The provider then assessed Study C (unpublished data) Study D 70 Not specified 2013 MLPT plus ultrasound MLPT plus ultrasound Mexico the abortion outcome using ultrasound and/or examination 2 weeks and determined whether the participant needed further treatment for ongoing pregnancy or another reason. Study E was a randomized trial designed to assess the effect of the timing of the post-treatment test on the diagnostic [3]

accuracy of the strategy [5]. In one group, each participant had 2010 an MLPT before the abortion and was then asked to perform – Study B 63 Not specified 2009 and/or exam Vietnam additional tests at home 3, 7 and 14 days later. She was told to 2 weeks return to the clinic for assessment with ultrasound if any follow-up MLPT showed either a rise or a decline in urine hCG concentration or at 14 days regardless of the result. In the comparison group, each participant had a highly sensitive ≥ [2] urine pregnancy test able to detect 25 mIU hCG/ml on Days 2011 3, 7 and 14, and was evaluated by the clinician on the day the – Study A United States result was negative or 14 days after enrollment, whichever MLPT plus ultrasound MLPT plus ultrasound 1 week was earlier. Studies F [6] and G [7] were randomized trials that compared the MLPT strategy to standard clinical follow-up after medical abortion. Participants assigned to the MLPT groups performed the post-treatment MLPTs at home 2 weeks after taking the mifepristone and were instructed to return to the clinic for ultrasound only if the urine hCG concentration for ongoing pregnancy after treatment N enrolledStudy designYears of enrollment 2010 490 CS 300 CS 400 CS 403 CS 300 RCT 1433 RCT 2400 RCT Table 1 Gynuity Health Projects studies evaluating urine MLPT included in this analysis Country Maximum GA (days) doseMisoprostol and routePosttreatment assessment Not specified 63 Timing of assessment did not decline or specified symptoms occurred; otherwise, Abbreviations: CS, case-series study; RCT,Note: randomized All clinical studies trial; included MLPT, women multilevel obtaining pregnancy outpatient test. medical abortion with mifepristone and misoprostol. E.G. Raymond et al. / Contraception xx (2017) xxx–xxx 3 they were considered to have had complete abortions and were 3. Results dismissed from the study. In the comparison groups, all participants returned to the clinic at two weeks for evaluation Across all seven studies, we identified no ectopic with exam and/or ultrasound. pregnancies among women who used the MLPT strategy for We used the primary data from the seven studies to perform medical abortion follow-up. Ongoing pregnancies diagnosed two main analyses. The purpose of the first analysis was to among study participants were treated with surgical abortion, estimate the diagnostic accuracy of the strategy for identifying additional abortifacient medication, or expectant management. ongoing pregnancy, focusing on its sensitivity (proportion of The first analysis included data from Studies A-E. These participants with known ongoing pregnancies who did not studies enrolled a total of 1893 participants assigned to use the have a decline in hCG concentration as measured by the MLPT for assessment of abortion outcome, of whom 1847 MLPT) and negative predictive value (proportion of partici- (97.6%) were ≤63 days of gestation (Table 2). Of these, 248 pants with a decline in hCG concentration who had no ongoing (13%) were excluded from the analysis because they did not pregnancy) [8]. This analysis included all studies designed to have both a follow-up MLPT test and definitive assessment of evaluate each participant for ongoing pregnancy using both the abortion outcome. Thus, the analysis population consisted of MLPT strategy and a clinical reference standard (ultrasound or 1599 participants. For 1442 (90%) of these participants, the exam), which were studies A-D and the MLPT group of study post-abortion evaluation included ultrasound; in the rest, the E. In this analysis, we included only participants treated at abortion outcome was assessed by examination. ≤63 days of gestation who provided MLPT results both before Of the 1599 participants in the analysis population, 21 and after the abortifacient treatment and had a definitive (1.3%) had ongoing pregnancies (Table 3A). The proportion reference post-treatment assessment of abortion outcome was substantially higher in Studies B and E than in Studies A, (defined as an ultrasound showing no ongoing pregnancy, an C, or D. A total of 1482 participants (93%) had a drop in examination at the study clinic showing evidence of resolving urinary hCG concentration between the two MLPTs; this pregnancy, a negative serum pregnancy test, or an ongoing proportion varied significantly among studies (pb.001, I2= pregnancy identified by any means). If a participant in studies 89.2%). None of the 21 participants with ongoing pregnancies A–D had multiple follow-up MLPTs, our analysis considered had a decline. Thus, in combined data, the sensitivity of the only the result of the earliest test performed; if that result was MLPT strategy for identifying ongoing pregnancy was 100% inconclusive or invalid, we considered that the value had not (95% CL 85.2%, 100%; heterogeneity p=.92, I2=0%) and the declined. In study E, we used the result of the last test negative predictive value was also 100% (95% CL 99.3%, performed. To assess the relationship between the diagnostic 100%, heterogeneity p=1, I2=0%). The positive predictive accuracy of the MLPT strategy and the interval between the value varied considerably across studies, from 0 in study C (in initial and follow-up MLPTs, we used data from Studies A–D which no ongoing pregnancies occurred) to 31.8% in Study E. to compare results in participants with an interval of ≤10 days The timing of the post-abortion MLPTs in Studies A–D to those in participants with an interval of ≥11 days. We was generally consistent with the different study protocols excluded Study E from this comparison because the timing of (Table 2). Across Studies A–D, the range was 1–32 days the final test for each participant was dependent on the results after the mifepristone, except for four outliers that had of her prior test. recorded intervals of 45–154 days. In combined data from Our second main analysis combined data from Studies F these four studies, a decline in hCG concentration occurred and G to compare ascertainment of ongoing pregnancy using in 285 of the 316 participants (90.2%) whose follow-up test the MLPT strategy with ascertainment using routine clinic was performed at ≤10 days and in 929 of the 992 (93.6%) follow-up. whose follow-up test was performed at≥11 days (p=.08; We used meta-analytic methods, conducted in R, version heterogeneity p=.7, I2=0%). 2.13.0, using the “metafor” package, version 1.6 [9,10],to The second analysis, which used data from Studies F and G, combine data across studies. When estimating percentages of included a combined total of 3833 participants, all of whom combined data, we used the Freeman–Tukey double arcsine were ≤63 days of gestation (Table 4). Abortion outcomes were transformation. When combining data across randomized ascertained in 3762 (98%) of these participants. Of participants trials, we used the log relative risk. We report p-values from the assigned to the MLPT groups, 15.3% (106/713) in Study F and chi-square test of heterogeneity and associated I2 statistic. We 3.4% (40/1200) in Study G were asked to return to the clinic calculated estimates and 95% confidence intervals using the because of either a lack of decline in hCG concentration or DerSimonian–Laird random-effects model. symptoms. Among participants with follow-up, the combined We used data as available from all seven included studies analysis did not find a significant difference between groups in to examine secondary outcomes related to the MLPT the proportion of participants with identified ongoing strategy, including the range of pre-treatment MLPT results, pregnancies (RR 0.88; 95% CI 0.50, 1.54). The proportion the proportions of participants with declines in urinary hCG of participants with ongoing pregnancies was substantially concentrations, and the incidence of invalid test results. higher in both randomized groups of Study F than in either No ethics approval was needed for this review of existing, group of Study G. The two studies were not statistically de-identified data. heterogeneous in this analysis (p=.70, I2=0%). One participant 4 E.G. Raymond et al. / Contraception xx (2017) xxx–xxx

Table 2 Findings of studies designed to evaluate the diagnostic accuracy of the MLPT for diagnosing ongoing pregnancy among participants ≤63 days Study A [2] Study B [3] Study C Study D [4] Study E [5] Total N=490, N=299, (unpublished data) N=368, N=300, N=1847, n (%) n (%) N=390, n (%) n (%) n (%) n (%) Lost to follow-up 96 (19.6) 7 (2.3) 28 (7.2) 25 (6.8) 7 (2.3) 163 (8.8) Abortion outcome determined without MLPT 11 (2.2) 1 (0.3) 9 (2.3) 5 (1.4) 1 (0.3) 27‡ (1.5) MLPT performed but no definitive 0 (0.0) 0 (0.0) 14 (3.6) 36 (9.8) 8 (2.7) 58§ (3.1) assessment of abortion outcome In analysis population 383 (78.2) 291 (97.3) 339 (86.9) 302 (82.1) 284 (94.7) 1599 (86.6) Of analysis population Follow-up evaluation included ultrasound 378 (98.7) 167 (57.4) 336 (99.1) 277 (91.7) 284 (100.0) 1442 (90.2) Follow-up MLPT timing⁎ ≤3 days 13 (3.4) 1 (0.3) 0 (0.0) 0 (0.0) 14 (1.1) 4–10 days 284 (74.2) 0 (0.0) 8 (2.4) 10 (3.3) 302 (23.0) 11–17 days 80 (20.9) 283 (97.3) 303 (89.4) 272 (90.1) 938 (71.3) ≥18 days 3 (0.8) 7 (2.4) 28 (8.3) 16 (5.3) 54 (4.1) not available 3 (0.8) 0 (0.0) 0 (0.0) 4 (1.3) 7 (0.5) hCG concentration decline Ongoing pregnancy 0 (0.0) 0 (0.0) 0 (0.0) 0 (0.0) 0 (0.0) 0 (0.0) No ongoing pregnancy 348 (90.9) 251 (86.3) 329 (97.1) 292 (96.7) 262 (92.3) 1482 (92.7) No hCG concentration decline† Ongoing pregnancy 1 (0.3) 11 (3.8) 0 (0.0) 2 (0.7) 7 (2.5) 21 (1.3) No ongoing pregnancy 34 (8.9) 29 (10.0) 10 (2.9) 8 (2.6) 15 (5.3) 96 (6.0) Abbreviations: hCG, human chorionic gonadotropin; MLPT, multilevel pregnancy test. ⁎ In days after mifepristone ingestion; total n and % excludes Study E. † Includes participants with stable, increased, or uninterpretable follow-up MLPT results. ‡ None had ongoing pregnancies. § Fifty-six had a decline in urinary hCG concentration according to the MLPTs.

Table 3 Diagnostic accuracy of MLPT strategy for identification for ongoing pregnancy in five studies in which each participant was evaluated using both the MLPT strategy and ultrasound or exam A. Studies A, B, D, E [2–5] and C (unpublished data) hCG concentration per MLPT Ongoing pregnancy, N No ongoing pregnancy, N Total, N No decline (pos test) 21 96 117 Decline (neg test) 0 1482 1482 Total 21 1578 1599

% (95% confidence interval) heterogeneity Sensitivity 100% (95% CL 85.2%, 100%) p=.92, I2=0% Specificity 94.2% (90.8%, 96.8%) pb.01, I2=85.2% Positive predictive value 17.4% (6.3%, 32.5%) p=.01, I2=71.4% Negative predictive value 100% (95% CL 99.3%, 100%) p=1.0, I2=0%

B. Studies C (unpublished data), D, and E [4,5] hCG concentration per MLPT Ongoing pregnancy, N No ongoing pregnancy, N Total, N No decline (pos test) 9 33 42 Decline (neg test) 0 883 883 Total 9 916 925

% (95% confidence interval) Heterogeneity Sensitivity 100% (95% CL 75.1%, 100%) p=.73, I2=0% Specificity 96.3% (94.6%, 97.7%) p=.19, I2=40.4% Positive predictive value 21.5% (8.2%, 38.9%) p=.20, I2=37.5% Negative predictive value 100% (95% CL 99.2%, 100%) p=1.0, I2=0% Note: Part B of this table excludes studies A and B because of the anomalous distribution of the pretreatment MLPT results in those studies. Abbreviations: hCG, human chorionic gonadotropin; MLPT, multilevel pregnancy test. E.G. Raymond et al. / Contraception xx (2017) xxx–xxx 5

Table 4 Randomized trials of follow-up using MLPT strategy vs. in-person clinical evaluation. Study F [6], N (%) Study G [7], N (%) Total, N (%) Assigned to MLPT strategy Enrolled 713 1200 1913 No or incomplete follow-up 4 (0.6) 9 (0.8) 13 (0.7) Ongoing pregnancy identified 18 (2.5) 5 (0.4) 23 (1.2) No ongoing pregnancy identified 691 (96.9) 1186 (98.8) 1877 (98.1)

Assigned to in-person evaluation Enrolled 720 1200 1920 No or incomplete follow-up 58 (8.1) 0 (0.0) 58 (3.0) Ongoing pregnancy identified 18 (2.5) 7 (0.6) 25 (1.3) No ongoing pregnancy identified 644 (89.4) 1193 (99.4) 1837 (95.7) Abbreviation: MLPT, multilevel pregnancy test. in these studies had an identified ongoing pregnancy and concentrations of 2000–9999 mIU/ml wasb3%, much lower reported a decline in urinary hCG concentration. All identified than the proportion in the other studies (Fig. 1). The pregnancies in both studies were ultimately aborted. gestational ages in these two studies as well as in the other Across all seven studies, of the 3535 participants treated at four studies that collected gestational age data were all ≤63 days who had MLPT tests both before and after the approximately symmetrically distributed (Fig. 2). Excluding abortion, the hCG concentration declined in 3355 (94.2%, 95% Studies A and B from the first analysis did not appreciably CL 90.0%, 97.3%; heterogeneity: pb.001, I2=95.3%) Nearly change the results (Table 3B). all of the follow-up MLPTs (3484/3535, 99%) were performed The seven studies included a total of 37 participants (in by the participants at home. Invalid tests were reported by 33 Studies B, C, and D) who were treated at 64–70 days of women (0.9%, 95% CL 0.2%, 2.0%; heterogeneity: pb.001, gestation and had a post-abortion clinical evaluation. Three of I2=87.5%). these participants had ongoing pregnancies, only one of which In examining the data, we noted that the distributions of was identified by the MLPT strategy. Of the other two, one pretreatment MLPT results appeared anomalous in Studies A (Study C) was treated at 68 days of gestation; her hCG reading and B: in particular, the proportion of tests showing hCG declined from 2000–9999 mIU/ml to 25–99 mIU/ml 13 days

25-99 mIU/ml 100-499 mIU/ml 80% 500-1,999 mIU/ml 2000-9,999 mIU/ml 70% 10,000 mIU/ml

60%

50%

40%

30%

20%

10%

0% A B C D E F G N=383 N=291 N=339 N=302 N=284 N=703 N=1196 Study

Fig. 1. Distribution of pretreatment MLPT results among participants treated at ≤63 days of gestation, Studies A–G. 6 E.G. Raymond et al. / Contraception xx (2017) xxx–xxx 35 days 50% 36-42 days 43-49 days 45% 50-56 days 57-63 days 40%

35%

30%

25%

20%

15%

10%

5%

0% A B C D E G N=381 N=289 N=339 N=302 N=284 N=1200 Study

Fig. 2. Distribution of gestational ages at time of mifepristone ingestion, Studies A–E and G. after treatment. The other (study D) was treated at 67 days of Yee, Ameritek, personal communication 21 September 2016), gestation; her hCG reading declined from 2000–9999 mIU/ml no independent confirmation of our findings is available. Our to 100–499 mIU/ml 12 days after treatment. first analysis excluded 13% of the potentially eligible participants. Criteria for identifying ongoing pregnancy were not standardized, and 10% of women in the analysis were 4. Discussion evaluated only by clinical examination, which may be less sensitive than ultrasound for identifying ongoing pregnancies. Our research indicates that serial measurement of urinary In the two randomized trials, some ongoing pregnancies may hCG concentration using the five-level dBest MLPT is a highly have been missed by the MLPT strategy because, by design, reliable and efficient approach for the evaluation of outcome of most women using this strategy did not return to the clinic. medical abortion treatment at ≤63 days of gestation. A woman Moreover, in all studies, providers performing the post- whose pregnancy is no longer viable is likely to have a decline treatment assessments were aware of the MLPT results, which in urinary hCG concentration within a week, and in the absence could have affected their clinical judgments regarding abortion of clinical complications, she should have no need for a outcome. Notably, though, even if the MLPT strategy had post-treatment ultrasound, examination, or other clinical tests. missed some ongoing pregnancies in these studies, its diagnostic Indeed, since women can perform the MLPT themselves accuracy may still be considered acceptable. Study C data have at home, the MLPT strategy can allow most women having not undergone external peer review. medical abortion to avoid an in-person follow-up visit to the In the two earliest studies [2,3], the distribution of abortion facility. pretreatment MLPT results appeared anomalous: in particular, The large number of participants, the rigorous study designs, the proportion of tests showing an hCG concentration of and the consistency of findings from a wide variety of settings 2000–9999 was remarkably low. This pattern did not parallel suggest that the conclusions of our analyses are robust and the distribution of GAs in the two studies. Whether it reflects a generalizable. However, the data have limitations. All included malfunction of the tests themselves, difficulty in reading the studies were managed by a single organization; to our tests, errors in recording the data, or some other problem is knowledge, based on our ongoing surveillance of the medical entirely unclear to us. Regardless, the test strategy did not miss literature and consultation with the dBest test manufacturer (KC any continuing pregnancies in either study. E.G. Raymond et al. / Contraception xx (2017) xxx–xxx 7 Our first analysis found significant heterogeneity among References studies in the proportion of participants with an hCG decline, which ranged from 86% to 97% across the five studies. The [1] Chen MJ, Creinin MD. Mifepristone with buccal misoprostol for medical explanation for this heterogeneity is unclear. It was not abortion: a systematic review. Obstet Gynecol 2015;126:12–21. associated with either the planned or actual timing of the [2] Blum J, Shochet T, Lynd K, et al. Can at-home semi-quantitative pregnancy tests serve as a replacement for clinical follow-up of post-treatment MLPT; indeed, we found no evidence that the medical abortion? A US study. Contraception 2012;86:757–62. inter-test interval affected the probability of observing a [3] Lynd K, Blum J, Ngoc NT, Shochet T, Blumenthal PD, Winikoff B. decline in hCG concentration. However, a detailed analysis Simplified medical abortion using a semi-quantitative pregnancy test of data from Study E suggested that performing the second for home-based follow-up. Gynaecol Obstet 2013;121:144–8. test as early as 3 days after mifepristone ingestion decreases [4] Dabash R, Shochet T, Hajri S, Chelli H, Hassairic AE, Haleb D, et al. Self-administered multi-level pregnancy tests in simplified follow-up the chance that a woman with a complete abortion will have of medical abortion in Tunisia. BMC Womens Health 2016;16:49. an observable decline in hCG concentration and thus [5] Blum J, Sheldon WR, Ngoc NT, Winikoff B, Nga NT, Martin R, et al. increases the chance that she will be unnecessarily referred Randomized trial assessing home use of two pregnancy tests for for further evaluation [5]. determining early medical abortion outcomes at 3, 7 and 14 days after – The seven studies included only 37 evaluable participants mifepristone. Contraception 2016;94:115 21. [6] Ngoc NT, Bracken H, Blum J, Nga NT, Minh NH, van Nhang N, et al. who took the mifepristone after 63 days of gestation, and thus Acceptability and feasibility of phone follow-up after early medical abortion we were unable to estimate the diagnostic accuracy of the in Vietnam: a randomized controlled trial. Obstet Gynecol 2014;123:88–95. MLPT strategy in women treated beyond that limit. In a healthy [7] Platais I, Tsereteli T, Comendant R, Kurbanbekova D, Winikoff B. pregnancy, production of hCG begins to fall at about that time Acceptability and feasibility of phone follow-up with a semiquantita- [11], which potentially could reduce the sensitivity of the MLPT tive urine pregnancy test after medical abortion in Moldova and Uzbekistan. Contraception 2015;91:178–83. strategy. Two of the three ongoing pregnancies that occurred [8] Khan KS, Bakour SH, Bossuyt PM. Evidence-based obstetric and among the 37 participants were missed by the strategy. gynaecologic diagnosis: the STARD checklist for authors, peer- Given that the incidence of ongoing pregnancy after reviewers and readers of test accuracy studies. BJOG 2004;111:638–40. medical abortion treatment through 63 days of gestation is [9] R Development Core Team. R: a language and environment for very low [1] and that continuing pregnancies can often be statistical computing. Vienna, Austria: R Foundation for Statistical Computing; 2011. recognized based on symptoms alone [12], routine follow-up [10] Viechtbauer W. Conducting meta-analyses in R with the metafor may not be needed for all women. The World Health package. J Stat Softw 2010;36:1–8. Organization endorsed this conclusion in its 2012 abortion [11] Braunstein GD, Rasor J, Danzer H, Adler D, Wade ME. Serum human guidance document [13]. However, when confirmation of chorionic gonadotropin levels throughout normal pregnancy. Obstet – complete abortion is desired, the dBest MLPT can be a Gynecol 1976;126:678 81. [12] Grossman D, Grindlay K. Alternatives to ultrasound for follow-up after useful assessment tool. Currently, the test is not available in medication abortion: a systematic review. Contraception 2011;83:504–10. most countries, but several entities have expressed interest in [13] World Health Organization. Safe abortion: technical and policy marketing it or similar tests in the near future. guidance for health systems, 2nd ed. ; 2012.