Clark, Wesley MPH; Bracken, Hillary PhD, MHS; Tanenhaus, Jini PA-C, MA; Schweikert, Suzanne MD; Lichtenberg, E Steve MD, MPH; Winikoff, Beverly MD, MPH
Guidelines for early medical abortion with mifepristone and misoprostol include a follow-up examination 7–14 days after mifepristone administration.1 At the follow-up visit, a clinician performs a pelvic examination and ultrasonography to confirm that the uterus is empty and no further treatment is required. However, sonography at the follow-up is not stipulated in the U.S. Food and Drug Administration–approved mifepristone label2 and is not used routinely in many countries.3 In some circumstances, routine use of ultrasonography at follow-up has been deemed to lead to unnecessary surgical completion procedures.4
Evidence suggests that clinicians and women can accurately determine expulsion of the gestational sac during medical abortion with mifepristone–misoprostol without the routine use of ultrasonography or a physical examination.5,6 Rossi et al6 found that when both the clinician and patient felt that the gestational sac had passed, expulsion was confirmed by sonography in 99.1% of the cases. Low-sensitivity qualitative urine pregnancy tests are also a promising technology that could be used at home by women themselves to verify a successful medical abortion.7,8
Reducing the routine use of ultrasonography would result in substantial time and cost savings for women and providers. In acceptability studies of mifepristone–misoprostol medical abortion, many women report the number of required office visits as a reason for not choosing the method.9,10 In fact, anecdotal evidence suggests that in routine clinical practice, a large percentage of women do not comply with standard protocols and do not return for their follow-up visit. Substantial staff time is often used in attempting to reach clients who have missed their follow-up appointments.
This analysis evaluated the success of algorithms created from data collected by the clinician and woman to predict the need for additional follow-up care after mifepristone–misoprostol medical abortion.
MATERIALS AND METHODS
The multicenter study was conducted at 10 clinics within three clinic networks: Planned Parenthood of San Diego and Riverside County (six clinics), Planned Parenthood of New York City (two clinics), and Family Planning Associates Medical Group of Chicago (two clinics). All women presenting at study sites seeking first-trimester termination of pregnancy were offered a chance to participate. Women who returned for and completed the follow-up visit received $25 as compensation for time and travel expenses. The Allendale Investigational Review Board reviewed and approved the study.
Women giving informed consent received the standard medical abortion care provided at each site up to 63 days from the last menstrual period according to on-site vaginal ultrasonography. Eligible women ingested mifepristone 200 mg in the clinic verified by direct observation. Each woman then self-administered 800 micrograms misoprostol. The route (oral, buccal, or vaginal) and timing (6–72 hours after mifepristone administration) of misoprostol varied according to local clinic protocol.
Depending on the clinic, women were advised to return 7–10 days (Chicago) or 7–14 days (Planned Parenthood clinics) after mifepristone administration. Before returning for their follow-up visits, women completed a symptom diary documenting their medical abortion experiences. Each woman was asked to record whether and when she saw expulsion of the pregnancy tissue and to indicate when she experienced heavy bleeding (ie, bleeding heavier than a normal period), pain, or cramps. On the self-administered questionnaire, each woman indicated whether she still “felt pregnant” or exhibited any pregnancy symptoms (breast tenderness, nausea or morning sickness, frequent urination, or exhaustion/tiredness) on the day of her follow-up appointment.
At follow-up, clinicians reviewed the women's symptom diary and, at the Chicago and San Diego sites, conducted a physical examination (a speculum or bimanual examination or both). At the Planned Parenthood of New York City clinics, a clinician met with the woman before sonography, but did not physically examine her unless indicated. At all sites, the clinicians documented their final assessments of the status of the abortion based on the patient interview or physical examination or both before viewing the results of the ultrasonography and urine pregnancy test. Clinicians then requested a urine sample from each participant. A laboratory technician performed a low-sensitivity urine pregnancy test, Immuno Pregnacol (Immunostics, Ocean, NJ). (A low-sensitivity test was used because a standard over-the-counter high-sensitivity test is likely to give a positive result at 7–14 days after treatment even in women who have had a successful abortion.) A clinician then performed vaginal ultrasonography to assess whether the abortion was complete. The result of the ultrasound examination was also documented on study forms.
Sample size calculations were based on the least common outcome of interest: ongoing pregnancy after medical abortion. Based on routine service statistics collected in two of the three clinic networks, we assumed ongoing pregnancy would occur in approximately 0.5% of early medical abortion cases. Given an assumption that sonography, as the gold standard, diagnoses 100% of ongoing pregnancies, the sample size was set to allow us to be 95% confident that screening methods that do not rely on ultrasound examination are able to identify 85% of ongoing pregnancies, ±15%. To do this, we required 22 cases of ongoing pregnancy, or a target enrollment of 4,400 study participants overall.
Data were collected on several post–medical abortion clinical outcomes, including ongoing pregnancy, receipt of curettage, empiric treatment for possible infection, administration of additional uterotonics, and other medical abortion–related care (eg, reassurance, expectant management, provision of ibuprofen or iron supplement). Final diagnosis of ongoing pregnancy was based on findings from the ultrasound examination. Data on additional follow-up care were collected up to 1 month after mifepristone administration. Only women who attended the follow-up visit and did not have a uterine suction evacuation or other additional treatment before this visit were included in this analysis. Women who returned to the clinic before the scheduled follow-up visit and received additional treatment were deemed to be “self-referred” cases who independently identified a need for additional care; their experience was thus not germane to our principal study question.
The clinic standard of care included data collected from the follow-up visit physician assessment, physical examination (if performed), and ultrasound examination. Women were identified as potentially needing additional treatment based on the results of the ultrasonography and the clinician assessment. Women who were missing data from the postprocedure ultrasound examination or in which the provider viewed the results of the ultrasound (n=6) were excluded from the analyses comparing the algorithms.
We constructed five model algorithms for evaluating women's postabortion status, each of which used a different assortment of data, but none of which used data from ultrasound examination. Table 1 indicates the conditions under which a woman would be referred for further evaluation under the five algorithms. Four of the algorithms (algorithms 1–4) rely on data collected by the woman and on the results of the low-sensitivity pregnancy test. Algorithm 5 relies on the woman's assessment, the results of the pregnancy test, and follow-up physician assessment or examination or both. In approximately one third of cases (n=1,012), women did not undergo a physical examination at the follow-up visit, because this was not the standard of care at the Planned Parenthood of New York City clinics. For these women, algorithm 5 includes physician assessment, without physical examination data. All cases in which the provider viewed the results of the pregnancy test before interviewing the woman or performing the physical examination (n=43) were excluded from analysis in algorithm 5.
The analyses compared the ability of each model to identify women who received additional post–medical abortion follow-up care. We classified post–follow-up care into three groups based on key medical abortion outcomes in descending urgency: treatment for ongoing pregnancy, receipt of curettage for other indications than ongoing pregnancy or empiric treatment for possible infection, and other medical abortion–related care (eg, additional uterotonics, provision of ibuprofen or iron tablets as part of expectant management). The principal analysis compared the results of each algorithm to the clinic standard of care using the McNemar χ2 test. Independent t tests were used for continuous variables. The χ2 or Fisher exact test, as appropriate, was used for independent nominal data. Confidence intervals were used where appropriate, and statistical significance was defined as a P value less than .05. All statistical analyses were performed using SPSS statistical software v15.0 (SPSS Inc., Chicago, IL).
From June 2005 through February 2007, 4,484 women were enrolled in the study at the three study sites (Fig. 1). Three women were excluded from analysis because of unanswered data queries. Of the remaining 4,481 women, 390 were determined on examination to be ineligible for medical abortion. The remaining 4,091 enrolled women received medical abortion treatment. Approximately 20% of women (n=795) who received mifepristone did not return for the postabortion follow-up visit. The sociodemographic characteristics of women who received medical abortion treatment, including those who returned and did not return for follow-up, are shown in Table 2.
Forty-nine women returned to the clinic for treatment before their scheduled follow-up visit (Fig. 1). As noted previously, these women were excluded from the evaluation of the algorithms because they received treatment for their condition based on a self-assessed problem (before any ultrasound examination or clinical evaluation). One hundred ninety-three women who returned for follow-up were missing laboratory, ultrasound, or patient assessment data required for the algorithm analyses. The cases with incomplete data included two women who required a surgical evacuation for an ongoing pregnancy, seven who received curettage for other indications or empiric treatment for possible infection, and one woman who received other medical abortion–related follow-up care. This left 3,054 women with complete data for our analyses.
Table 3 shows the postabortion treatment received by women who did not receive additional treatment before the follow-up visit, returned for follow-up, and had adequate data for evaluation. In total, of the 3,054 women with complete data, 96.6% did not receive any additional medical abortion–related treatment, besides expectant management, during the medical abortion process. A few women (n=12) received treatment not related to the medical abortion at the follow-up visit. These women, along with those who received only reassurance or expectant management (n=48), were not included in the category of patients who received additional postabortion treatments. Less than 1% (0.7%) of women were diagnosed with an ongoing pregnancy and consequently underwent a surgical evacuation. Less than 1% (0.9%) of women received a curettage for all other indications (for bleeding, debris, etc) or empiric treatment for possible infection. Slightly less than 2% (1.8%) of women received other medical abortion–related care.
Figure 2 compares the percentages of women who received postabortion treatments of various kinds that were identified by each model algorithm. The first image represents the clinic standard of care and each subsequent image represents one of our proposed alternate algorithms. Each woman is represented by one light dot. In the clinic standard of care image, red dots represent women with ongoing pregnancies, yellow dots represent women who received postabortion curettage for some other reason or who were treated for a possible postabortion infection, and dark blue dots represent women who received other medical abortion–related care. These red, yellow, and dark blue dots appear in one of two columns, indicating whether each model would have identified the woman as needing some sort of postabortion treatment. Those in the “needs identified” column were, indeed, “flagged” by the particular algorithm in question; those in the “not identified” column eventually received treatment, but were not (or would not have been) flagged as possibly needing such treatment. Dots encircled in blue, at the right of the image, represent women who would have been given ultrasound examinations under each algorithm. The percentage of the total receiving ultrasound examinations is also indicated in bold blue numerals.
The results obtained under the clinic standard of care appear in the top image of Figure 2. As specified, the diagram indicates that routine post–medical abortion ultrasound examination (the current clinic standard of care) identified all women diagnosed with an ongoing pregnancy (n=20) and 17 of the 26 women (65%) who eventually received curettage or empiric treatment for a possible infection. This model also identified 52 of 55 women (94.5%) who eventually received some other medical abortion–related care.
A total of 256 women were missing data from the provider assessment and were excluded from the analysis of algorithm 5. Algorithm 5 (the image at the bottom of Fig. 2), the screening algorithm based on three patient-reported outcomes plus a urine test plus a nonsonographic clinical evaluation, performed as well as sonography in identifying all women with ongoing pregnancies at follow-up. Algorithm 5 identified 19 of 24 women (79%) who received curettage or empiric treatment for a possible infection, a somewhat higher percentage of the same sample than was identified using sonography (16 of 24, 65%). The algorithm was slightly less effective at identifying women who received other medical abortion–related care, identifying only 39 of the 48 women with complete data (79.6%) who received such treatments compared with the clinic standard of care that identified 45 of the 48 women (94%). There was no statistically significant difference in the number of women who received any care identified by algorithm 5 compared with the clinic standard of care, although the percentage of women who received other medical abortion–related care approached significance (P=.07).
Whereas algorithm 5 would require all women to return to the clinic for a follow-up examination, algorithms 1–4 rely on data gathered by the woman herself together with the results of the low-sensitivity pregnancy test—which, although administered at the clinic in our study, could conceivably be performed by a woman at home. A purpose of algorithms 1–4, among other things, is to help investigate whether data gathered by the woman at home might safely and effectively identify which women should return to the clinic for follow-up and which women could skip this extra visit.
As Figure 2 indicates, the different algorithms offer differing tradeoffs between sensitivity and convenience. For example, algorithm 1 would require only slightly less than one third (31.7%) of women to return to the clinic, but would also miss 2 of the 20 women with ongoing pregnancies. Algorithm 4, which includes more detailed questions, would be more sensitive and identify percentages of women in the various categories comparable to those identified by the current clinic standard of care (20 of 20 ongoing pregnancies, 21 of 26 women who received curettage or empiric treatment for possible infection, and 43 of 55 women who received other medical abortion–related care). However, algorithm 4 would also require almost two thirds (64.0%) of all women to return to the clinic. Algorithms 2 and 3 fall in between. There was no statistical difference in the number of ongoing pregnancies, surgical completions, or cases receiving empiric treatment for possible infection identified with algorithms 1–4 compared with the clinic standard of care. Algorithms 1–4 were significantly less effective at identifying women who received other medical abortion–related care (P≤.05).
There was intersite variation in the performance of the five algorithms. At the New York site, algorithms 1–5 identified all women who received any type of care. Almost all (data not shown) women (95% or 52 of 55 women) who required other medical abortion–related care were recruited at the Chicago and San Diego sites, perhaps reflecting intersite differences in standards for who should and should not receive additional minor treatments. In San Diego, algorithms 1 and 2 were less accurate in identifying women who required other medical abortion–related care compared with the current clinic standard of care. In Chicago, there was no difference in the ability of the five algorithms to identify women who required any care when compared with the clinic standard of care.
Our study found that using patient-controlled diagnostics along with postabortion examination to determine whether a woman requires further evaluation after medical abortion can have effectiveness similar to that of a current clinic standard of care that includes sonography used for the same purpose. Women and their providers could identify all women with an ongoing pregnancy and as high a percentage of those who eventually received curettage or empiric treatment for possible infection as did ultrasonography.
The results of algorithm 5 suggest that clinics can safely forego routine ultrasound examinations at the follow-up visit. Although all women would still be required to return for follow-up, only one third would be referred for ultrasound examination. Moreover, the study findings indicate that with the use of a self-administered questionnaire and a home-administered low-sensitivity pregnancy test, women themselves may be able to accurately and safely determine their need for a follow-up visit after early medical abortion. As demonstrated by the comparison of algorithms 1–4, adding questions of increasing specificity identifies a higher percentage of women who might need additional postabortion care, although at the cost of requiring a greater percentage of clients to return for follow-up. Algorithm 4, the most comprehensive model, would require only two thirds of women to return for a follow-up visit, with effectiveness comparable to that of the current standard of care in these clinics, including routine ultrasound examination.
It is also worth reconsidering the gold standard we used for comparisons and the related concept of need with regard to postabortion care. The gold standard used in this study represents what is done now. Although an ongoing pregnancy after an attempted abortion clearly represents a treatment failure and needs to be addressed, it is extremely likely that ongoing pregnancies missed by postabortion diagnostics will eventually be identified in time to offer surgical abortion. Therefore, identifying 100% of ongoing pregnancies immediately after abortion, although certainly desirable, is not absolutely necessary. This is even truer regarding other postabortion problems: Curettage is often offered “defensively” to women with any signs of debris in the uterus or moderate ongoing bleeding, when such women might in fact not need this additional care. Women who do eventually need it are likely to self-identify in time for effective treatment to be offered. Thus, it is not necessary for a postabortion evaluation protocol to be perfect in order for it to be considered medically acceptable. Indeed, ultrasound examination, as demonstrated here, is itself imperfect in this regard, because it failed to identify all women who eventually received follow-up treatment.
As a comparison, it is interesting to consider the National Emergency X-Radiography Utilization Study (NEXUS) criteria, which are routinely used in emergency departments across the world to minimize the provision of unnecessary cervical spine x-rays.11 In a recent large study, the NEXUS criteria failed to identify 8 of 818 patients with cervical spine injury and would have resulted in avoiding unnecessary x-rays in only 12.6% of study patients. In this context, even our least successful algorithm (algorithm 1)—which would have missed 2 of 20 ongoing pregnancies, but would have relieved more than two thirds of women of the burden of returning for a follow-up visit, with an even greater reduction in the provision of unnecessary ultrasound examinations—looks promising.12
Although this study includes an extremely large sample of correlations between sonographic and nonsonographic diagnostic findings and clinical outcomes in medical abortion clients, it is not without limitations. First, loss to follow-up is a problem in any study of medical abortion. The loss to follow-up rate demonstrated in the study is consistent with that experienced at most clinics offering medical abortion and suggests the desirability to many women of finding a way to reduce the number of required visits for the treatment. Second, this study was a modeling exercise and retrospectively compared the results of the algorithm with real-life experience. As designed, it provides ample evidence that it is ethical to test alternatives to routine ultrasound examinations and follow-up visits for post–medical abortion follow-up prospectively.
The low-sensitivity pregnancy test used in this study was not 100%, accurate, correctly identifying only 16 of the 20 women with complete data who had ongoing pregnancies. Use of the test, however, significantly improved the accuracy of the diagnostic algorithms we constructed. Our test was administered in the clinic rather than at home, but it would be possible to design a home-administered model, as well as one that gave more accurate results. Indeed, there already are a number of low-sensitivity tests that may be administered by women at home and promise to be more accurate than the test used in this study.13 There is reason to hope, then, that a low-sensitivity test, together with self-assessment, could eliminate the need for a routine follow-up visit for a large percentage of women.
This study retrospectively compared the results of the ultrasound examination and other tests. However, the identification of the desirability of care alone may not result in the receipt of that care. Future research must test the algorithms in routine clinic practice and how best to administer them—whether via self-administered paper questionnaire, a phone interview, or other media. Different approaches may prove more effective at ensuring women's cooperation and the transmission of information to the clinic.
This study suggests that nonsonographic means of evaluating women after medical abortion may be practicable. In future papers, we hope to be able to demonstrate that routine use of ultrasonography before abortion is also unnecessary. If so, these papers hold out the prospect that future protocols could be developed to allow practitioners without ultrasound equipment to offer medical abortion and to allow women to evaluate the success of their medical abortions safely and effectively at home. This service delivery model would allocate follow-up visits and the use of ultrasonography to the minority of women most likely to benefit from them. Reducing the requirement for sonography would allow more practitioners to offer medical abortion, reducing the number of visits required would allow more women to opt for the treatment, and simplifying and lightening the demands of medical abortion on healthcare resources would contribute to containing costs and inefficiency in our increasingly strained health care system. The next step is to test these models in clinical practice and compare their results in terms of safety, patient and staff acceptance, burden on clinics, and cost.
© 2010 by The American College of Obstetricians and Gynecologists.