Women undergoing elective abortion are at high risk of having another unintended pregnancy within the next year.1,2 With proven fertility and, in most cases, a demonstrated need for better contraception, many women with unintended pregnancies have an acute need for effective contraception that is effective and easy to use. Intrauterine devices (IUDs) are among the most effective forms of reversible contraception available. Postabortal insertion of IUDs has been shown to be a safe3–16 and effective3–14,17 form of contraception. For women who have difficulty arranging transportation to a clinic or time off work, postabortal insertion also offers the convenience of a single visit.
However, postabortal IUD insertion is rarely offered in the United States. Instead, women are commonly asked to return 2–8 weeks after their abortion procedure to have an IUD inserted. This is due in part to concerns about increased rates of IUD expulsion and infection with postabortal insertion. Additionally, issues of reimbursement and restrictions on use of Title X funding in facilities that provide abortion services may also contribute to the structure of service provision. To compare the number of pregnancies and repeat abortions expected in the 12 months after an abortion followed by immediate insertion of an IUD with insertion of an IUD at a later follow up visit, we performed an evidence-based decision analysis. In addition, we examine those parameters that are most critical in determining rates of pregnancy in the 12 months after an abortion.
MATERIALS AND METHODS
We developed an evidence-based decision model (Fig. 1) to analyze the optimal timing of IUD placement for women who, after having an uncomplicated surgical abortion, want to use intrauterine contraception to avoid pregnancy for at least 1 year. The model evaluated two options for IUD insertion: 1) insertion of the IUD immediately after the abortion procedure (immediate insertion), and 2) insertion of the IUD at a follow-up visit (delayed insertion).
We compare IUD placement strategies by examining the proportion of women who become pregnant within 12 months of their abortion. The model was constructed by using TreeAge Pro 2006 (TreeAge Software, Williamstown, MA).
We reviewed the relevant literature to estimate the most likely value (as well as the range of lowest and highest reasonable values) for each parameter included in the model (Table 1) (Costales AC, Jensen JT, Nelson AL, Korner P, Uddin MA. A U.S. multicenter open-label trial with the levonorgestrel-releasing intrauterine system—clinical and device-related experience [abstract]. Contraception 2006;74:178).1–3,5,12,18–24 The variables are 1) failure to return for delayed IUD insertion, 2) IUD 12-month continuation, 3) IUD expulsion with immediate insertion, 4) pregnancy with IUD in place, and 5) pregnancy without an IUD. Pertinent data were identified by searching MEDLINE with the terms “intrauterine device” and “abortion, induced” or “abortion, therapeutic” and by reviewing the bibliographies of identified references.
One-way and two-way sensitivity analyses were performed for the reasonable range of values identified for all parameters.1 Monte Carlo simulation25 was also performed by using 10,000 samples with triangular distributions for all variables. Triangular distributions were used to allow the synthesis of data from multiple studies identified in the published literature. The lower limit of each triangular distribution was defined by the lowest reasonable value; the mode was defined by the most reasonable value; and the upper limit was defined by the maximum reasonable values. The output of the Monte Carlo simulation was analyzed with Stata 9 (Stata Corp, College Station, TX). The results of the Monte Carlo simulation were stratified and graphed by categorical ranges of variables found to be influential in the one-way analyses. The Monte Carlo results were also used to examine variable values with which immediate IUD insertion was anticipated to prevent more than 80 pregnancies. The expected value of perfect information, defined as the best option for a given scenario assuming one could know and choose the better choice in advance, was also calculated for the model by using the Monte Carlo simulation results.
Estimates of the effects of increased IUD use in the United States after abortion were calculated by using national data. Because there were an estimated 1.3 million abortions performed annually in the United States between 1995 and 2000,26,27 we describe the impact of changes in the timing of postabortion IUD insertion for a population of 1.3 million U.S. women. Two scenarios were considered: 1) all women undergoing abortion opt to have an IUD placed, and 2) 20% of women opt for an IUD after abortion, instead of non-IUD contraception. Predicted changes in the number of pregnancies and abortions expected in the following year with each of these scenarios were calculated by subtracting the number of annual pregnancies and abortions from available national data. We conservatively estimated that half of women who became pregnant in the year after their abortion procedures would terminate the subsequent unintended pregnancy because this is the overall proportion of unintended pregnancies ending in elective abortion.28 The scenario in which 20% of women opt for an IUD after abortion is felt to be a realistic goal. In populations in which IUD use is well accepted, approximately 20–35% of reproductive-age women (including those trying to conceive and those who are sterile or have been sterilized) use IUDs.29,30 Given that this analysis examines the subset of reproductive-age women who have definitively declared their desire not to be pregnant, 20% is a modest goal.
Only IUDs currently available in the United States were considered: the copper T380a (CuT380a, ParaGard, Duramed, Pomona, NY) and the levonorgestrel-releasing IUD (LNG-IUD, Mirena, Berlex, Montville, NJ). The first-year typical use effectiveness rates for the LNG-IUD and the CuT380a are 0.1% and 0.8%, respectively.23 Based on the available data, both the CuT380a5,7 and the LNG-IUD6 are expected to have similar contraceptive effectiveness, whether inserted immediately or 2–8 weeks after a first-trimester abortion. We averaged these two values and used 0.5% as the contraceptive efficacy in the base case, with a range of efficacy from 0.1% to 1.0%.
In most published studies of IUDs, the 12-month continuation rate ranges from 70 to 90 (Costales et al, 2006 [abstract]).19,20,31 In American women using the LNG-IUD, the 12-month continuation rate was 79% (Costales et al, 2006 [abstract]). The 12-month continuation used in the model was 80%, with a range of 65–95% in sensitivity analyses.
To obtain an estimate of the rate of continued IUD use in units of woman-years of use, we converted the continuation rate in the literature to an area under the curve (AUC), estimated by using a linear approximation of this exponential distribution. At 12 months, the linear approximation was found to be equivalent to the integration of the exponential distribution. This method allows women who discontinue use of the IUD during the year to be assigned the proper risk of pregnancy for the whole year.
Two studies compared immediate and delayed insertion of the CuT380a in a nonrandomized fashion.21,22 Both studies found that expulsion rates were not different between the groups. Expulsion of Mirena after first-trimester insertion12 does not appear to be higher than the rates reported after interval insertion at 12 months.32
Although it is most likely that the expulsion risk is not increased with immediate insertion after first-trimester abortion, the studies are small and cannot guarantee equivalence. Therefore, after first-trimester abortion, we conservatively estimated that expulsion may be up to 5% higher with immediate postabortal insertion. The expulsion rate may be as high as 30% after second-trimester abortion.5 In this analysis we assumed a 5% rate in the base case analysis. For sensitivity analyses, the expulsion risk was varied from 0% up to 30%.
Although described as expulsion, this variable can represent any differential discontinuation of IUDs after immediate insertion. These discontinuations could include infection, prolonged bleeding, unexpected pain, or other subjective complaints that could be attributed to the IUD. However, no studies were identified to show discontinuation rates for these events.
In prior studies, high percentages of women in the delayed group have not returned for IUD insertion. The only published randomized trial comparing immediate with delayed postabortal insertion of an IUD, conducted in Montreal, found that 42% of women did not return for delayed insertion.3 In a nonrandomized Egyptian study of the CuT380a,22 25% of women who chose delayed insertion did not return. Based on these studies, we used 35% for the base case, with a range of 5–55% in the sensitivity analyses.
Studies of postabortal women demonstrate that, for a given contraceptive method, recurrence of unintended pregnancy within the 12-months is more common than would be expected from contraceptive failure rates with typical use. One trial of postabortal injection of depot medroxyprogesterone acetate2 found that 22% of women studied were pregnant again within 12 months.23 Another study that followed women for 12 months after a medication abortion,1 found that 25% (95% confidence interval 16–35%) of women experienced another unplanned pregnancy despite attempting to use a variety of contraceptive methods. We, therefore, assumed that women using non-IUD contraception would face a baseline annual risk of pregnancy of 20% for the decision analysis, with a range of 8–30%.
Because women who are able to keep appointments and fill prescriptions are more likely to succeed in effectively using a contraceptive method, we assumed a correlation between observed effectiveness of non-IUD contraception and the probability that women succeeded in returning for a scheduled follow-up visit for IUD insertion. A correlation factor was therefore used to link the failure-to-return proportion to non-IUD contraceptive effectiveness. The correlation factor in the base case is 0.571, with a range of 0.228–0.857 for sensitivity analyses. When these values are multiplied by the “failure to return” values, the risk of pregnancy without an IUD s 20%, with a range of 8–30%.
No data could be identified on the risk of pregnancy in the first 6 weeks after the abortion. Consequently, we conservatively assumed that no pregnancies occurred between abortion and delayed IUD insertion, biasing toward delayed insertion.
Placing a foreign body immediately after a surgical abortion could theoretically increase the infection risk of the abortion procedure. We found no evidence that women who have an IUD placed immediately after an abortion procedure have increased rates of infection compared with women who do not have an IUD placed.25 In a large Scandinavian trial that followed women for 12 months after postabortal insertion, the LNG-IUD was discontinued because of infection in only 0.7% of women,6 compared with 0.3% after interval insertion in a large multicenter Scandinavian trial.32 An infection rate of 0.4% attributable to the abortion can be inferred. This compares favorably with published rates of postabortal infection in Scandinavia of over 5%.18 In two additional studies that compared immediate and delayed insertion of the CuT380a,21,22 infection rates were low and did not differ between the groups. Consequently, the risk of infection does not need to be considered because good evidence exists showing that the infection risk is not increased with immediate insertion.
One perforation of the uterus has been reported among the approximately 3,000 IUDs inserted immediately after abortion using T-shaped IUDs.3,5–17 As such, the perforation rate appears to be relatively negligible for immediate postabortal insertion. Therefore, the risk of perforation was not included in the model.
With immediate insertion, 34 pregnancies would be expected per 1,000 women in the 12 months after abortion, whereas 86 pregnancies would be expected with delayed insertion. In this model, 75% of women who had an IUD immediately inserted would be expected to have an IUD 12 months after abortion, compared with 52% of women offered delayed insertion.
Sensitivity analyses reveal that the model is strongly influenced by three variables: the expulsion risk, the proportion not returning for delayed insertion, and the effectiveness of non-IUD contraception. Two variables were found to have limited influence on the model: the effectiveness of the IUD and the IUD continuation rate. The differential effect of individual parameter variation between immediate and delayed insertion is shown in Figure 2.
The effectiveness of non-IUD contraception was found to influence immediate and delayed insertion proportionately and, thus, does not change the overall conclusions of the model, just the magnitude of the difference between arms. As might be predicted, the effectiveness of non-IUD contraception was much more influential than the effectiveness of the IUD itself because few pregnancies result with IUD use. Two-way sensitivity analyses confirmed these findings (Fig. 3).
The results of the Monte Carlo simulations support those of the base case and one-way sensitivity analyses (Fig. 4). In 90.9% of scenarios, immediate insertion was favored. On average, 27.9 pregnancies per 1,000 women were prevented by immediate insertion compared with delayed insertion in the first year after abortion. If the results are limited to scenarios where excess expulsion in the immediate group was less than 10%, immediate insertion is preferred in 99.3% of scenarios and prevents an average of 37.2 pregnancies compared with delayed insertion in the first year. Panels B, C, and D of Figure 4 show how the distribution changes based on the expulsion risk with immediate insertion, the proportion who fail to return for delayed insertion, and the risk of pregnancy using non-IUD contraception.
The expected value of perfect information is the best option for a given scenario, assuming one could know and choose the better option in advance. In this model, the expected value of perfect information is 0.4 pregnancies per 1,000 woman. Because immediate insertion is preferred overall, this is the additional number of pregnancies that could be prevented beyond immediate insertion alone, if one had knowledge of the best time to insert the IUD for each scenario.
If all women in the United States opted for immediate insertion of an IUD after abortion, an estimated 216,000 unintended pregnancies and 102,000 abortions would be prevented in the next 12 months, compared with 148,000 pregnancies and 70,000 abortion prevented if all women chose delayed insertion (Table 2). More realistically, if 20% of women opted for immediate insertion of an IUD after abortion, an estimated 43,000 pregnancies and 20,000 abortions would be prevented in the subsequent 12 months, compared with 30,000 pregnancies and 14,000 abortions prevented if 20% of women opted for delayed insertion.
The results of this decision analysis support the hypothesis that immediate postabortal IUD insertion would increase IUD use and decrease rates of subsequent unintended pregnancy. Immediate postabortal insertion prevents 52 pregnancies per 1,000 women over the initial 12 months compared with delayed insertion. This projection only examines the effect on the first year after the abortion. However, with multiple years of continued use, the cumulative effect would be substantially larger. The expected value of perfect information, a measure of decision uncertainty, was low, implying that attempts to stratify women or populations into groups best suited for immediate or delayed insertion are unnecessary. In aggregate, all women receive near maximal benefit from immediate insertion.
The primary reason that fewer pregnancies occur following immediate postabortal IUD insertion is that IUD use is greatly increased compared with rates of IUD use when women must return to have the IUD inserted at a later date. Under our base case assumptions, immediate postabortal IUD insertion would allow an additional 18% of women to use an IUD 12 months after abortion. Because IUDs are more effective than most other contraceptive options, many fewer unintended pregnancies result. Only in a few situations would a general policy of delayed insertion be preferred over immediate insertion. Such situations may occur when expulsion rates with immediate insertion exceed 20% or when more than 90% of women return for delayed IUD insertion.
The model has several limitations. First, the model did not consider the possibility of pregnancy between the time of abortion and the time of delayed IUD insertion. Commonly, delayed insertion is not performed until 4–6 weeks after the abortion, by which time the majority of women have ovulated33,34 and resumed sexual activity.12 Because the current literature does not provide any estimates of risk of pregnancy during this time period, this parameter was omitted from the model. However, any pregnancies that occur during this time period would strengthen the argument in favor of immediate postabortal insertion. Second, the model assumes that women who fail to return for delayed insertion have the same contraceptive failure risk as women who returned but later discontinued use of the IUD. However, it is likely that women who fail to return for follow-up appointments to have an IUD inserted are more likely to also fail to fill prescriptions or make other necessary appointments and, thus, be more likely to experience a contraceptive failure with a non-IUD contraceptive method. As such, including this parameter in the model would strengthen the argument in favor of inserting IUDs immediately postabortion.
The decision model has several implications for future research. From sensitivity analyses of the five model variables, it is apparent that two variables have little impact on the results of the model. The risk of pregnancy with an IUD affects the model the least. Thus, the model can be applied to both the CuT380a (ParaGard) and the LNG-IUD (Mirena) without modification. The IUD continuation rate also has little impact on the model. Continuation rates may vary somewhat by region or population served, but the expected results based upon this model would not be substantively changed.
Expulsion rate, however, is influential on model results. However, to be clinically relevant for current practice, the probability of expulsion needs to be much higher than presently observed after suction aspiration procedures. Because the IUD is placed into a dilated cervical os in immediate postabortal insertion, there is a logical basis for concern that the IUD might be more frequently expelled. However, multiple studies have shown that expulsion risk is not substantially increased by immediate compared with delayed insertion after first-trimester abortion. The upper limit of expulsion in this model extends to 30% to allow the results to be applied to second-trimester terminations, based on the limited data currently available. The decision analysis supports the conclusion that, if expulsion is less than 10%, immediate insertion results in fewer pregnancies. Our review of the current literature generally shows immediate postabortal insertion of IUDs to be safe and effective, although further investigation of the rate of IUD expulsion after immediate postabortal insertion is warranted.
For women living in rural areas of the developing world, access to effective contraception may be limited. Abortion services may also be limited, but dilation and curettage (D&C) is often performed for spontaneous abortion. For women who have traveled from rural areas, this may be their only point of contact with the medical system. In such settings, the proportion not returning for delayed insertion may approach 100%. The decision model is equally applicable to the developing world setting, without changing any parameters. Immediate placement of an IUD after a D&C would allow women to receive highly effective contraception at the time of their contact with the medical system. Based on extension of the sensitivity analysis for failure to return, immediate insertion would be much preferred in such scenarios. Furthermore, immediate insertion can be safely performed after suction curettage for incomplete abortion, many of which are the results of illegal abortion.11,35 Because the vast majority of providers of surgical abortion and postabortal care are skilled providers of IUDs, no additional training would be necessary.
Unfortunately, for many women living in the United States, access to immediate IUD insertion may be as limited as in the developing world. Many states prohibit the use of Title X funds for any contraceptive method within the facility where abortions are provided, a restriction that effectively prevents women from obtaining the contraception needed to prevent another abortion. The findings of this decision analysis indicate that policies designed to promote insertion of IUDs immediately after an abortion may be highly successful in reducing subsequent unintended pregnancy and repeat abortions.
1. Creinin MD. Conception rates after abortion with methotrexate and misoprostol. Int J Gynaecol Obstet 1999;65:183–8.
2. Goldberg AB, Cardenas LH, Hubbard AE, Darney PD. Post-abortion depot medroxyprogesterone acetate continuation rates: a randomized trial of cyclic estradiol. Contraception 2002;66:215–20.
3. Gillett PG, Lee NH, Yuzpe AA, Cerskus I. A comparison of the efficacy and acceptability of the Copper-7 intrauterine device following immediate or delayed insertion after first-trimester therapeutic abortion. Fertil Steril 1980;34:121–4.
4. Grimes D, Schulz K, Stanwood N. Immediate postabortal insertion of intrauterine devices (Cochrane Review). In: The Cochrane Library, Issue 3, 2002. Oxford: Update Software.
5. IUD insertion following termination of pregnancy: a clinical trial of the TCu 220C, Lippes loop D, and copper 7. Stud Fam Plann 1983;14:99–108.
6. Pakarinen P, Toivonen J, Luukkainen T. Randomized comparison of levonorgestrel- and copper-releasing intrauterine systems immediately after abortion, with 5 years' follow-up. Contraception 2003;68:31–4.
7. Timonen H, Luukkainen T. Immediate postabortion insertion of the copper-T (TCu-200) with eighteen months follow-up. Contraception 1974;9:153–60.
8. Nielsen NC, Nygren KG, Allonen H. Three years of experience after post-abortal insertion of Nova-T and Copper-T-200. Acta Obstet Gynecol Scand 1984;63:261–4.
9. Goldman JA, Dekel A, Reichman J. Immediate postabortion intrauterine contraception in nulliparous adolescents. Isr J Med Sci 1979;15:522–5.
10. Batar I, Wildemeersch D, Vrijens M, Delbarge W, Temmerman M, Gbolade BA. Preventing abortion and repeat abortion with the Gynefix intrauterine implant system–preliminary results. Adv Contracept 1998;14:91–6.
11. IUD insertion following spontaneous abortion: a clinical trial of the TCu 220C, Lippes loop D, and copper 7. Stud Fam Plann 1983;14:109–14.
12. Ortayli N, Bulut A, Sahin T, Sivin I. Immediate postabortal contraception with the levonorgestrel intrauterine device, Norplant, and traditional methods. Contraception 2001;63:309–14.
13. Bitsch M, Jakobsen AB, Prien-Larsen JC, Frolund C, Sederberg-Olsen J. IUD (Nova-T) insertion following induced abortion. Contraception 1990;42:315-22.
14. Chowdhury NN, Mandal GS, Das M. Comparative study of Lippes Loop and CuT inserted in immediate post-abortal period. J Obstet Gynaecol India 1979;29:234–44.
15. Nygren KG, Johansson ED. Insertion of the endouterine Copper-T (TCu 200) immediately after first trimester legal abortion. Contraception 1973;7:299–306.
16. Moussa A. Evaluation of postabortion IUD insertion in Egyptian women. Contraception 2001;63:315–7.
17. Newton J, Elias J, Johnson A. Immediate post-termination insertion of Copper 7 and Dalkon Shield intrauterine contraceptive devices. J Obstet Gynaecol Br Commonw 1974;81:389–92.
18. Sawaya GF, Grady D, Kerlikowske K, Grimes DA. Antibiotics at the time of induced abortion: the case for universal prophylaxis based on a meta-analysis. Obstet Gynecol 1996;87:884–90.
19. Sivin I, Stern J, Diaz J, Diaz MM, Faundes A, el Mahgoub S, et al. Two years of intrauterine contraception with levonorgestrel and with copper: a randomized comparison of the TCu 380Ag and levonorgestrel 20 mcg/day devices. Contraception 1987;35:245–55.
20. Nilsson CG, Luukkainen T, Diaz J, Allonen H. Intrauterine contraception with levonorgestrel: a comparative randomised clinical performance study. Lancet 1981;1:577-80.
21. Gocmen A, Demirpolat N, Aysin H. The efficacy of immediate post-abortion intrauterine device insertion. Clin Exp Obstet Gynecol 2002;29:274–6.
22. El-Tagy A, Sakr E, Sokal DC, Issa AH. Safety and acceptability of post-abortal IUD insertion and the importance of counseling. Contraception 2003;67:229–34.
23. Trussell J. Contraceptive failure in the United States. Contraception 2004;70:89–96.
24. Sivin I, el Mahgoub S, McCarthy T, Mishell Jr, DR Shoupe D, Alvarez F, et al. Long-term contraception with the levonorgestrel 20 mcg/day (LNg 20) and the copper T 380Ag intrauterine devices: a five-year randomized study. Contraception 1990;42:361–78.
25. Briggs AH, Goeree R, Blackhouse G, O'Brien BJ. Probabilistic analysis of cost-effectiveness models: choosing between treatment strategies for gastroesophageal reflux disease. Med Decis Making 2002;22:290–308.
26. Strauss LT, Herndon J, Chang J, Parker WY, Bowens SV, Berg CJ. Abortion surveillance—United States, 2002. MMWR Surveill Summ 2005;54 (SS-7):1–36.
27. Finer LB, Henshaw SK. Abortion incidence and services in the United States in 2000. Perspect Sex Reprod Health 2003;35:6–15.
28. Finer LB, Henshaw SK. Disparities in Rates of Unintended Pregnancy in the United States, 1994 and 2001. Perspect Sex Reprod Health 2006;38:90–6.
29. United Nations Population Division. UN Department of Economic and Social Affairs. World contraceptive use 2005. New York (NY): United Nations; 2006.
30. ACOG unveils survey of women ob-gyns at media briefing. ACOG Today 2004;48:1, 6–7.
31. Arowojolu AO, Otolorin EO, Ladipo OA. Performances of copper T 380A and multiload copper 375/250 intrauterine contraceptive devices in a comparative clinical trial. Afr J Med Med Sci 1995;24:59–65.
32. Luukkainen T, Allonen H, Haukkamaa M, Holma P, Pyorala T, Terho J, et al. Effective contraception with the levonorgestrel-releasing intrauterine device: 12-month report of a European multicenter study. Contraception 1987;36:169–79.
33. Marrs RP, Kletzky OA, Howard WF, Mishell DR, Jr. Disappearance of human chorionic gonadotropin and resumption of ovulation following abortion. Am J Obstet Gynecol 1979;135:731–6.
34. Lahteenmaki P, Ylostalo P, Sipinen S, Toivonen J, Ruusuvaara L, Pikkola P, et al. Return of ovulation after abortion and after discontinuation of oral contraceptives. Fertil Steril 1980;34:246–9.
35. Goldsmith A, Goldberg R, Eyzaguirre H, Lizana L. Immediate postabortal intrauterine contraceptive device insertion: a double-blind study. Am J Obstet Gynecol 1972;112:957–62.