Braaten, Kari P. MD, MPH; Benson, Carol B. MD; Maurer, Rie MA; Goldberg, Alisa B. MD, MPH
Intrauterine contraceptive devices (IUDs) are becoming increasingly common, and are now the contraceptive method of choice for 5.5% of United States women.1 With frequent use of pelvic ultrasonography to evaluate gynecologic complaints, the discovery of malpositioned IUDs has also become a more common occurrence. Little is known about factors that predispose to IUD malpositioning or the ideal management of malpositioned devices. Anecdotally, there is concern that IUDs placed in the late postpartum period may be associated with increased rates of malpositioning, although no prior studies have examined this directly. Some studies have shown increased rates of perforation when IUDs are inserted in the postpartum period and in lactating women,2–4 whereas other studies have not shown significant differences in perforation rates according to postpartum or breastfeeding status.5,6
With regard to efficacy, studies have shown that copper IUDs have decreased contraceptive efficacy if not located in the fundal portion of the uterus.7–9 Therefore, copper IUDs noted to be malpositioned should be removed and replaced. The levonorgestrel-releasing intrauterine system has a different mechanism of action, and it seems likely that the local effect of the levonorgestrel is adequate for contraception even if the device is not at the fundus. This hypothesis is supported by a randomized clinical trial that compared an intracervical device that releases the same dose of levonorgestrel as the levonorgestrel-releasing intrauterine system with the levonorgestrel-releasing intrauterine system placed at the fundus. This study demonstrated no difference in pregnancy rates between the intracervically and the fundally positioned devices.10
The goal of our study was to estimate if IUD insertion 6–9 weeks postpartum increases the risk of malpositioning. We also sought to identify other possible risk factors for IUD malpositioning and to examine the management and outcomes of women with a malpositioned IUD noted on ultrasonography, including removal and subsequent pregnancy rates.
MATERIALS AND METHODS
We performed a case–control study to estimate the effects of postpartum IUD insertion and other factors on the risk of subsequent IUD malpositioning noted by ultrasonography. We searched the radiology clinical database that includes reports of ultrasound examinations done at Brigham and Women's Hospital from January 1, 2003, to June 30, 2008, using the terms “IUD” or “intrauterine device.” Those in the case group were those patients whose ultrasound examinations demonstrated an IUD that was described as being in the lower uterine segment, cervix, rotated, embedded, intraperitoneal, or expelled. The determination of the location of the IUD was made by experienced attending radiologists who interpreted the transvaginal ultrasound examinations. No specific criteria or measurements such as distance from the fundus were used. Control participants were defined as patients with normally positioned IUDs and were selected as the next consecutive patient in the radiology database with a normally located IUD after the given study case. Control participants were chosen in a 1:1 ratio with case participant. Only patients who had a formal ultrasonograph done in the radiology department were included in this study; no ultrasound examinations done in clinics or offices were included. Once participants for the case and control groups were identified, the hospital electronic medical record system was searched using medical record numbers, and the primary investigator abstracted all data onto a standardized coded abstraction form. The electronic medical record system used for our institution includes hospital records for all patients and office records for some physician groups. Paper records were not searched and patient interviews were not conducted. Data were abstracted between January and August 2009 and entered into a secure database created for the purpose of this study. The study was approved by the Partners Healthcare institutional review board.
The primary exposure variable was IUD insertion 6–9 weeks postpartum. Other potential exposure variables that were hypothesized to be potential risk factors for IUD malpositioning included immediate postabortal insertion, breastfeeding at the time of insertion, type of IUD (levonorgestrel-releasing intrauterine system or copper IUD), pregnancy history including parity and mode of delivery, presence of uterine leiomyomas or suspected adenomyosis, prior loop electrosurgical excision procedure, and whether the IUD was placed for contraception or the treatment of gynecologic symptoms. Although a definitive diagnosis of adenomyosis requires surgical pathology, adenomyosis is often suspected on clinical or radiologic grounds. Women with a clinical diagnosis of suspected adenomyosis in the electronic medical record or an ultrasonography report that described a possible diagnosis of adenomyosis were defined as having “suspected adenomyosis.” All radiologic diagnoses of suspected adenomyosis were made by attending radiologists who subspecialize in ultrasonography. The diagnosis was made by transvaginal ultrasonography when a region of the myometrium appeared thickened and heterogeneous, had irregular shadowing, and contained small cysts and interspersed hyperechoic tissue. We also collected data on the indication for the ultrasound examination to examine if there was a relationship between presenting symptoms and malpositioned devices and to control for this factor in the event of confounding. Demographic data including age, ethnicity, marital status, and type of insurance coverage were also collected.
We examined what response was taken by providers to the malpositioning. Management outcomes were defined as immediate removal (during the same day or encounter), delayed (removed as a result of malpositioning within 1 year), IUD left in situ, expelled, or unknown. For those women whose IUDs were removed or expelled, we noted whether they were provided with alternative contraception and what type of contraception. Information on pregnancy rates within 2 years for participants in both the study case and control groups was recorded, including whether the pregnancy was planned. For patients whose initial IUD ultrasound examination was done less than 2 years from data collection, pregnancies were recorded up until the time of data collection.
The final sample size was a convenience sample of all cases of malpositioned IUDs that were available in the ultrasonography database from January 1, 2003, to June 30, 2008. A sample size calculation was performed before the study using the two independent proportion methods.11 The control group was not matched on any variables, but rather chosen using a 1:1 ratio; therefore, matching was not accounted for in our sample size calculation or analysis. We assumed that the probability of postpartum insertion in the control group was 15%–40%. A sample size of 180 participants in the case group and 180 participants in the control group would allow us to detect an odds ratio (OR) ranging from 2.0 to 3.0 with a two-sided α of 0.05 and 70%–99% power. Given these numbers, and the exploratory nature of this study, we were aware of the possibility that we might be underpowered to detect a significant difference for our primary as well as secondary outcomes.
Statistical analysis of baseline characteristics and univariable analysis were performed using t tests for numerical variable comparisons and chi-squared or Fisher's exact tests for categorical variable comparisons. Several exposure variables had more than 10% missing data: postpartum insertion (15%), postabortion insertion (14%), breastfeeding (25%), type of IUD (13%), and IUD placed for treatment of gynecologic symptoms (14%). Breastfeeding status was most commonly not available; however, the pattern of missing data appeared completely random. Therefore, we applied the multiple imputation method for analysis of missing data. Exposure variables with a univariable significance of P<.10 were selected for logistic regression model building. Both forward and stepwise selection methods were used and those maintaining a P<.05 were kept. The univariable predictors not included in the model were manually re-entered for an assessment of confounding effects. Interaction effects among the significant variables were also assessed. Assessment of model fit was tested using Hosmer and Lemeshow chi-squared tests. Odds ratios with 95% confidence intervals (CIs) and two-sided P values were reported for the final model.
We identified 1,748 pelvic ultrasound reports containing the terms “IUD” or “intrauterine device” in the period from January 2003 to June 2008. From these, 182 ultrasound examinations revealed malpositioned IUDs, and the women with these ultrasound examinations were our study case group. This represented a rate of malpositioning of 10.4% (95% CI 9%–12%) among IUD users having pelvic ultrasonography for any indication. Baseline demographic factors of those in the study case and control groups are shown in Table 1. There were no differences between those in the study case and control groups by age, parity, or marital status. There were statistically significant differences in race and ethnicity and insurance status between groups. There was a higher proportion of African American and a lower proportion of white patients among those in the case group. Those in the case group also were less likely to have private insurance than were those in the control group.
The most common type of malpositioning was the presence of the IUD in the lower uterine segment or cervix, which was found in 133 (73.1%) of the 182 women in the case group. Of these, 29 were also described as being embedded, rotated, or both. There were 21 IUDs that were embedded only or embedded and rotated (11.5%), 13 were expelled (7.1%), and 13 were intraperitoneal (7.1%). Thirty-four IUDs (18.7%) were described as being malpositioned in more than one way.
There was no increased risk of IUD malpositioning with insertion 6–9 weeks postpartum (OR 1.46, 95% CI 0.81–2.63) or postabortion insertion (OR 0.78, 95% CI 0.12–5.18). The only factor which was found to be associated with IUD malpositioning in the univariable analysis was suspected adenomyosis (OR 3.48, 95% CI 1.36–8.91), whereas prior vaginal delivery (with or without another delivery by cesarean delivery) was protective (OR 0.68, 95% CI 0.43–1.09) (Table 2). There were no differences according to type of IUD, breastfeeding status, indication for IUD placement, known leiomyomas, history of loop electrosurgical excision procedure, or endometriosis. Not surprisingly, we found that more women in the case group had their initial ultrasonography performed for symptoms suggestive of IUD malpositioning, including pain, bleeding, and pregnancy. We also found that having symptoms suggestive of IUD malpositioning was associated with a diagnosis of “suspected adenomyosis.” Controlling for this confounding relationship in a multivariable model, we found that suspected adenomyosis remained a statistically significant risk factor for malpositioning (OR 3.04, 95% CI 1.08–8.52) as did having an ultrasound examination for symptoms suggestive of malpositioning (OR 1.74, 95% CI 1.11–2.72). Prior vaginal delivery (OR 0.53, 95% CI 0.32–0.87) and private insurance status (OR 0.38, 95% CI 0.24–0.59) remained protective (Table 3). The Hosmer-Lemeshow test from the multivariable model did not indicate a lack of model fit (P<.61). When taken together, all types of IUD malpositioning did not appear to be associated with postpartum placement or breastfeeding; however, when examining specifically the 13 patients with intraperitoneal (perforated) IUDs, an increased risk was observed with insertion 6–9 weeks postpartum (OR 8.77, 95% CI 2.41–31.88, P=.001) and breastfeeding (OR 11.81, 95% CI 2.03–68.79, P<.008) (data not shown).
Among the 182 women with malpositioned IUDs, 121 (66.5%) had their IUDs removed. Forty-one of these women (33.9% of 121) had the device removed immediately and 80 (66.1% of 121) had the device removed between 1 and 250 days after diagnosis of malpositioning. Only 28 (15.4%) IUDs were kept in place once they were discovered to be malpositioned. Of those whose IUDs were removed or expelled, 93 women (77%) planned to initiate an alternative method of birth control, but only 37 (30.6%) received another highly effective form of birth control (another IUD, implant, or sterilization) (Fig. 1).
In the 2 years after the index pelvic ultrasound examination, there were more pregnancies among those in the study case group than those in the control group (19.2% compared with 10.5%, P=.046) (Table 4). Of the 35 pregnancies that occurred among those in the study case group, 32 (91.4%) were among women whose IUD was known to have ultimately been removed or expelled. Two of the malpositioned IUDs that were ultimately removed were initially left in despite malpositioning and later removed for other reasons, one for pain and one for desired pregnancy. The remaining three pregnancies among those in the study case group were women for whom the outcome of their IUD was unknown. There were no pregnancies that occurred with a malpositioned IUD known to be in situ. Among the 19 control participants with a normally positioned IUD who became pregnant during the follow-up period, 18 had their IUD removed before becoming pregnant and one patient had an ectopic pregnancy with a normally positioned copper IUD in place. Despite an overall higher pregnancy rate among those in the study case group than those in the control group, the proportion of unplanned pregnancies was not statistically different between groups (40.0% compared with 57.9%, P=.45).
Our study did not demonstrate an association between IUD malpositioning and insertion 6–9 weeks postpartum. We did, however, find that malpositioned IUDs were more likely to be found among women with suspected adenomyosis and less likely to be found among women with a prior vaginal delivery and private insurance. Importantly, we also found a significantly higher rate of subsequent pregnancy among women with malpositioned IUDs, none of which occurred with the malpositioned device in place.
It is encouraging that we did not demonstrate an overall increased rate of IUD malpositioning with insertion 6–9 weeks postpartum, because this is an important time for access to effective contraception for many women. However, we did see an elevated risk of intraperitoneal (perforated) IUDs in women who were 6–9 weeks postpartum or breastfeeding, which is consistent with prior findings.2–4 The absolute number of intraperitoneal IUDs observed was small and the CI was wide for this subgroup analysis. Therefore, we feel that concern for increased risk of perforation should not be a reason to defer IUD insertion during this time period; however, prospective research is needed to accurately estimate the risk of perforation with IUD insertion during the postpartum period and with breastfeeding.
The other notable finding of our study is that women with malpositioned IUDs were more likely to become pregnant within the 2 years that followed their ultrasound examination. Among the women for whom follow-up data were available, pregnancies among women with malpositioned IUDs occurred only when the IUDs were removed; there were no pregnancies seen with a malpositioned IUD in place. In women with normally positioned IUDs, pregnancies also predominantly occurred when IUDs were removed, although there was one ectopic pregnancy with a normally positioned copper IUD in place. The higher pregnancy rate among those in the study case group, therefore, indicates that women with malpositioned IUDs have high discontinuation rates and removed devices are often not immediately replaced with highly effective contraception, resulting in a higher pregnancy rate. Although health care providers may remove malpositioned IUDs due to fear of decreased efficacy, our study demonstrates that the typical use failure rates of other contraceptive methods appear far greater than the theoretical risk of decreased efficacy with a malpositioned IUD. Therefore, patients who are asymptomatic with a malpositioned IUD in the uterus may be best served by either leaving the device in place or removing the device only when it can be immediately replaced with another highly effective method of contraception. Women who have symptoms such as pain or bleeding and who are found to have a malpositioned device may require IUD removal to relieve their symptoms; however, they too would be best served by initiating another highly effective method at the time of IUD removal. Having had one malpositioned IUD removed or expelled is not a contraindication to reinsertion of a new device, and although malpositioning may recur, it is likely that women who were symptomatic from a malpositioned device might not have recurrence of symptoms with a device that is properly placed.
The association between malpositioning and suspected adenomyosis was an unexpected finding of our study. Previous studies have found higher rates of expulsion in women with heavy menses for whom the etiology of menorrhagia was unknown and in women with uterine leiomyomas.12,13 The reason for the association between adenomyosis and IUD malpositioning could be abnormal uterine contractility in women with adenomyosis, which has previously been demonstrated.14–16 These patients may also have more difficult IUD placements from distortion of the uterine cavity.17 It is also possible that this finding is the result of unaddressed bias. Although we found that the indication for ultrasonography (ie, referral for ultrasonography for pain or bleeding) confounded the relationship between suspected adenomyosis and malpositioning, when we controlled for this in our multivariable model, the association between suspected adenomyosis and malpositioning remained significant. We acknowledge that given the retrospective nature of this study, there were no predetermined criteria used by the radiologists for the definition of malpositioning or adenomyosis. It seems unlikely that radiologists characterize adenomyosis differently in the presence of a malpositioned IUD; however, it is possible that an IUD was more likely to be interpreted as being located in the lower uterine segment in an enlarged adenomyotic uterus. Interestingly, the same association was not observed for women with leiomyomas. Thus, this finding should be interpreted with caution.
Our study must be interpreted in the context of its study design. This was a small retrospective study, which began as exploratory in nature and was limited by the number of cases that were observed during the time period in question. As such, we only had power to detect an OR of 2–3 for our primary outcome and it is possible that there exists an increased risk of a smaller magnitude that we were not able to detect. Our sample size and power calculation were based on examining the relationship between IUD malpositioning and postpartum insertion, and the analysis of secondary outcomes such as pregnancy rate or subgroup analysis were likely underpowered. Despite this, our findings of increased malpositioning with suspected adenomyosis, the protective effect of vaginal delivery, and the increased rate of intraperitoneal IUDs with postpartum insertion or breastfeeding were statistically significant. Other limitations include the possibility of selection bias because we selected control participants from a large pool of possible control participants. However, we do not have any reason to believe that sequential selection of control participants from the ultrasonography database would result in systematic bias that would alter the results of our study. Finally, it is possible that there were pregnancies that were diagnosed and managed outside of our hospital system that we were not able to capture. Indeed, there were 21 individuals in the case group and 31 in the control group for whom there was no available information on future pregnancies in our medical records. However, there is no reason to suspect that the rate of pregnancies managed outside our hospital system would differ between these two groups.
Our study also raises questions about the commonly held belief that all malpositioned IUDs should be removed, because we demonstrated that removed devices are often not immediately replaced with highly effective contraceptive methods and removal may result in higher pregnancy rates. Although women who have expelled one device have an increased likelihood of repeat expulsion, the risk of recurrence of IUD malpositioning is unknown.18 More research is needed to understand the risks and benefits of removal of malpositioned devices and whether removal and reinsertion, possibly under ultrasonographic guidance, offers improved outcomes.
1. Mosher WD, Jones J. Use of contraception in the United States: 1982–2008. Vital Health Stat 23 2010;29:1–44.
2. Caliskan E, Ozturk N, Dilbaz BO, Dilbaz S. Analysis of risk factors associated with uterine perforation by intrauterine devices. Eur J Contracept Reprod Health Care 2003;8:150–5.
3. Andersson K, Ryde-Blomqvist E, Lindell K, Odlind V, Milsom I. Perforations with intrauterine devices. Report from a Swedish survey. Contraception 1998;57:251–5.
4. Heartwell SF, Schlesselman S. Risk of uterine perforation among users of intrauterine devices. Obstet Gynecol 1983;61:31–6.
5. Kapp N, Curtis KM. Intrauterine device insertion during the postpartum period: a systematic review. Contraception 2009;80:327–36.
6. Chi I, Potts M, Wilkens LR, Champion MS. Performance of the copper T-380A intrauterine device in breastfeeding women. Contraception 1989;39:603–18.
7. Anteby E, Revel A, Ben-Chetrit A, Rosen B, Tadmor O, Yagel S. Intrauterine device failure: relation to its location within the uterine cavity. Obstet Gynecol 1993;81:112–4.
8. Bernaschek G, Spernol R, Beck A. The position of the IUD and intrauterine pregnancy [in German]. Geburtshilfe Frauenheilkd 1981;41:645–7.
9. Schmidt EH, Wagner H, Quakernack K, Beller FK. Results of the follow-up of the position of intra-uterine device by ultra-sonography [in German]. Geburtshilfe Frauenheilkd 1979;39:138–43.
10. Pakarinen P, Luukainen T. Five years' experience with a small intracervical/intrauterine levonorgestrel-releasing device. Contraception 2005;72:342–5.
11. Fleiss JL, Levin B, Pail MC. Statistical methods for rates and proportions. 3rd ed. New York (NY): John Wiley & Sons; 2003.
12. Hidalgo M, Bahamondes L, Perrotti M, Diaz J, Dantas-Monteiro C, Petta C. Bleeding patterns and clinical performance of the levonorgestrel-releasing intrauterine system (Mirena) up to two years. Contraception 2002;65:129–32.
13. Kaunitz AM. Progestin-releasing intrauterine systems and leiomyoma. Contraception 2007;75:S130–3.
14. Aguilar HN, Mitchell BF. Physiological pathways and molecular mechanisms regulating uterine contractility. Hum Reprod Update 2010;16:725–44.
15. Kissler S, Zangos S, Wiegratz I, Kohl J, Rody A, Gaetje R, et al.. Utero-tubal sperm transport and its impairment in endometriosis and adenomyosis. Ann N Y Acad Sci 2007;1101:38–48.
16. Mehasseb MK, Bell SC, Pringle JH, Habiba MA. Uterine adenomyosis is associated with ultrastructural features of altered contractility in the inner myometrium. Fertil Steril 2010;93:2130–6.
17. Peng FS, Wu MY, Yang JH, Chen SU, Ho HN, Yang YS. Insertion of the Mirena intrauterine system for treatment of adenomyosis-associated menorrhagia: a novel method. Taiwan J Obstet Gynecol 2010;49:160–4.
18. Bahamondes L, Diaz J, Marchi NM, Petta CA, Cristofoletti ML, Gomez G. Performance of copper intrauterine devices when inserted after an expulsion. Hum Reprod 1995;10:2917–8.