Levonorgestrel intrauterine device use and incident idiopathic intracranial hypertension among commercially insured women.

The University of Michigan IRB deemed this study exempt from review. The Clinformatics Data Mart database (OptumInsight, Eden Prairie, MN) is a de-identified administrative claims dataset including all beneficiaries in a US, managed-care network. The dataset includes medical claims and sociodemographic data on beneficiaries with at least one International Classification of Diseases, Ninth Revision and Tenth Revision, Clinical Modification (ICD-9/10) code for eye-related diagnoses; at least one Current Procedural Terminology (CPT) code for eye-related visits or diagnostic or therapeutic procedures; or any other claim submitted by an ophthalmologist or optometrist from January 1, 2001 through December 31, 2015. Women were eligible if they were 18–45 years old at incident use of a form of highly effective contraception, participated in the medical plan from 2001–2015, and were continuously enrolled for at least one year. We defined highly effective contraception as a hormonal (LNG-IUD, etonogestrel subcutaneous implant) or non-hormonal device (copper IUD), or surgical procedure (hysterectomy or tubal device or surgery [e.g. tubal ligation, intratubal occlusion device, oophorectomy, etc]). We focused on long-acting procedural or surgical forms of contraception to maximize the likelihood that patients actually received the contraception, unlike short acting reversible methods (e.g., pills, rings) and injectables requiring periodic administration. Although hysterectomy is not used primarily for contraception, we included these women as an additional reference group as the operation results in permanent contraception. We identified incident use of highly effective contraception using a one-year lookback period from date of first contraception use with no other codes for the forms of contraception under study. LNG-IUD, copper IUD, etonogestrel implant, hysterectomy, and tubal device or surgery were identified by a combination of ICD-9/10, CPT, and Healthcare Common Procedure Coding System alphanumeric codes ( Supplemental Table 1 with full inclusion/exclusion criteria). We excluded women with unspecified contraception devices, failed insertions, or mechanical complications, because successful device placement could not be assumed. We excluded women using multiple study contraceptive methods or a study contraceptive with other specific hormonal contraceptives, those using a levonorgestrel implant (as these are no longer available in the US), and women who became pregnant while using one of the study contraceptives ( Supplemental methods ). We excluded women with one or more ICD-9/10 codes for cerebral dural venous sinus thrombosis, cerebral edema, and brain tumors diagnosed while on a form of study contraception because these diagnoses may be associated with elevated intracranial pressure and must be ruled out by the clinician to make a diagnosis of idiopathic intracranial hypertension. Last, we excluded women using medications associated with elevated intracranial pressure including minocycline, tetracycline, doxycycline, and vitamin A derivatives (e.g. tretinoin) in the 6 months prior to the idiopathic intracranial hypertension diagnosis ( Supplemental Table 2 ). We considered the first ICD/10 code for idiopathic intracranial hypertension (348.2, G93.2) following a one-year look back period without any idiopathic intracranial hypertension codes as an incident diagnosis. Subjects were required to have a CPT code for an MRI of the brain, CT scan of the head, or lumbar puncture in the one-year prior to idiopathic intracranial hypertension diagnosis, as these diagnostic tests are required to fulfill the Modified Dandy Criteria for idiopathic intracranial hypertension [ 5 ]. We abstracted the following data for all eligible subjects: age at initiation of one of the study methods, race, ethnicity, education, and income. Race and ethnicity data were defined within the database. We identified subjects with the following ICD-9/10, HCPCS, and CPT codes ( Supplemental Table 2 ) at any time during the study period: obesity, polycystic ovarian syndrome, bariatric surgery, abnormal uterine bleeding, leiomyoma, endometriosis, and conditions to construct the Charlson comorbidity index [ 6 ]. We used the Charlson comorbidity index as a surrogate marker of health conditions potentially related to obesity, because obesity is inconsistently recorded in administrative data [ 7 ]. We selected covariates based on possible associations with idiopathic intracranial hypertension and/or contraception selection [ 8 – 13 ]. Descriptive statistics summarized features of women by contraception type. Kaplan-Meier analysis was used to estimate the time-dependent probability of idiopathic intracranial hypertension at one and five years after incident contraception use, stratified by contraception type. We investigated the association between contraception type and time to incident idiopathic intracranial hypertension diagnosis (primary comparison: LNG-IUD versus copper IUD) using multivariable Cox proportional hazards to adjust for covariates. Time to idiopathic intracranial hypertension was calculated from the first date associated with a CPT code for contraception until either disenrollment from the health plan, end of the study, an incident idiopathic intracranial hypertension event, or device removal code. The proportional hazards model assumption was investigated with interactions between covariates and time, with no violations found. The Cox model was fit several times while alternating the reference category for contraception to present hazard ratios (HR) between all pairs of contraception methods. Results are reported with HRs and 95% confidence intervals (CIs). We performed a sensitivity analysis using the propensity score (continuous) of contraception type derived by multinomial logistic regression for the five-level contraception variable as an alternative method to adjust the Cox model for potential confounders and balance the groups. All analyses were performed using SAS version 9.4 (SAS Institute, Cary, NC).

Of the 268,280 women meeting the inclusion and exclusion criteria ( Figure 1 ), 78,175 (29%) had a LNG-IUD, 20,275 (8%) had a copper IUD, 8,715 (3%) used an etonogestrel implant, 108,216 (40%) had a hysterectomy, and 52,899 (20%) had a tubal device/surgery. Women using LNG-IUDs were more often White, educated (Bachelor’s degree or higher), and had higher income (≥$100k). Women who underwent hysterectomy were older, had higher Charlson comorbidity index scores, were more obese, had bariatric surgery, and carried diagnoses of abnormal uterine bleeding, leiomyoma, and endometriosis. Table 1 compares demographic and clinical characteristics among women using the different contraception types. During the study period, 208 (0.08%) women developed idiopathic intracranial hypertension during a mean follow-up period from initiation of study contraceptive of 2.4 years (SD=2.4 years; median=1.6). Kaplan-Meier probabilities of idiopathic intracranial hypertension among women at one year/five years following contraception initiation was 0.0004/0.0021 for LNG-IUD, 0.0002/0.0012 for etonogestrel implant, 0.0005/0.0006 for copper IUD, 0.0003/0.0016 for hysterectomy, and 0.0003/0.0015 for tubal device/surgery. No significant difference in hazard of idiopathic intracranial hypertension was found between women using different contraception methods after adjustment for potential confounders (age, race, education, Charlson comorbidity index, obesity, polycystic ovarian syndrome, bariatric surgery, abnormal uterine bleeding, fibroids, endometriosis). Specifically, women using LNG-IUD did not show a significantly different hazard of developing idiopathic intracranial hypertension compared to women using copper IUD (HR 1.84 [95% CI: 0.88–3.85]). Additionally, the hazard of idiopathic intracranial hypertension was not significantly different among women using LNG-IUD versus etonogestrel implant, hysterectomy, or tubal device/surgery. See Table 2 for all comparisons ( Supplemental Table 3 for full model). Sensitivity analysis using a propensity score adjusted Cox model showed similar results ( Supplemental Table 4 ).

We did not observe a significantly increased hazard of idiopathic intracranial hypertension among women in this large, national cohort using LNG-IUD compared to women using copper IUD, or other forms of highly effective contraception after adjusting for potential important confounders. Additionally, we did not observe an increased hazard of idiopathic intracranial hypertension among 8000 women using etonogestrel implants compared to LNG-IUD, copper IUD, hysterectomy, or tubal device/surgery. This finding is important, because systemic progestin exposure is approximately two to three-fold higher among women using etonogestrel implants compared to more locally acting hormone in LNG-IUD [ 14 ]. The observed absolute risk of idiopathic intracranial hypertension was very low regardless of contraception use type. Our findings can be used to reassure women considering LNG-IUD placement or continued use. It is important to acknowledge that the effect was in the direction of an increased hazard of idiopathic intracranial hypertension for LNG-IUD when compared to copper IUD. Unmeasured confounding, particularly obesity, may account for these observed trends as obesity is associated with both idiopathic intracranial hypertension and contraception selection. Only 14–26% of women had an ICD9/10 code for obesity ( Table 1 ), well below the national estimates (40%) of obesity among women ages 20–39 years old in the US. [ 16 ] A prior case control study found people (two men included) with higher body mass index (BMI) had significantly higher odds of idiopathic intracranial hypertension, increasing with BMI (BMI 30–35kg/m 2 : OR 19.5 [95% CI: 3.5,109.9]; BMI >35kg/m 2 OR 26.0 [95%CI; 4.9,135.9]) [ 13 ]. Although that study was small, these effect sizes are much larger than what was observed in our study ( Supplemental Table 3 ). Additionally, women with a BMI ≥ 35 kg/m 2 have greater odds IUD placement or surgical sterilization compared to women with lower BMIs [ 15 ]. Other unmeasured factors related to LNG-IUDs, progestins, or contraception selection may also account for the observed trend. Future studies of idiopathic intracranial hypertension risk with contraception will likely rely on observational data given the low idiopathic intracranial hypertension incidence rate, making clinical trials impractical. Our findings underscore the importance of controlling not only for confounders of idiopathic intracranial hypertension [ 17 ], but also contraceptive selection. Methodologic issues have been particularly problematic for prior studies examning the relationship between idiopathic intracranial hypertension and LNG-IUDs. Etminan et al . [ 1 ] reported higher odds (OR 1.78 [95% CI: 1.41,2.25]) for any cause of intracranial hypertension among women using LNG-IUD versus ethinyl estradiol-norgestimate by querying the United States Food and Drug Administration’s Adverse Events Reporting System database. They later revised this finding, demonstrating no evidence of increased rates of “benign” intracranial hypertension in age restricted analyses (15–40 years, OR 0.90 [95%CI: 0.55,1.47]) [ 1 , 3 , 18 , 19 ]. Valenzuela et al. [ 2 ] compared LNG-IUD use among newly diagnosed idiopathic intracranial hypertension patients at University of Utah and the Glostrup Hospital in Denmark. They observed increased odds of idiopathic intracranial hypertension in women with LNG-IUDs (Utah: OR 7.70 [95% CI: 3.7, 16.0]; Denmark: OR 3.91 [95% CI: 1.89, 8.06]); but, could not adjust for important counfounders such as age and BMI. Further analyses using exposure estimates that more closely reflected population estimates of LNG-IUD use resulted in no significant association between LNG-IUDs and idiopathic intracranial hypertension (Utah: OR 1.41 [95%CI: 0.67,2.98]; Denmark: OR 1.29 (0.61,2.70) (authors employed by Bayer) [ 2 , 20 ]. The primary strength of our study is our attempt to address prior methodologic deficiencies. We used conservative definitions of contraceptive use, employed an idiopathic intracranial hypertension definition that required multiple diagnostic and procedure codes, and excluded subjects with confounding conditions. We also tried to control for obesity, conditions associated with obesity (e.g. bariatric surgery), and contraception selection factors (e.g. abnormal uterine bleeding is often managed with LNG-IUD), as well as age and socio-demographic factors associated with both obesity and contraception selection. Our study has important limitations. Despite a large cohort the incidence of IIH was low and therefore, we do not have statistical power to identify small-moderate associations. The study population was confined to individuals who had an eye examination because it is required to make an idiopathic intracranial hypertension diagnosis. Also, racial/ethnic and low-income groups are underrepresented in this privately-insured sample. We cannot account for contraception that originated outside the plan or lifetime use of all forms of contraception. We also did not exclude women that concurrently used oral forms of contraception. However, concurrent use of long-acting forms of contraception, such as IUDs, along with short-acting hormonal contraceptives is low at an estimated 1.1% [ 21 ]. We did not adjust for the number of pregnancies in our cohort and a woman’s greatest exposure to endogenous progesterone occurs during pregnancy. However, a case control study of 109 patients did not demonstrate an increased risk of idiopathic intracranial hypertension recurrence in pregnancy [ 22 ]. Last, we cannot assure that those patients identified as having idiopathic intracranial hypertension actually had idiopathic intracranial hypertension. Although our definition of idiopathic intracranial hypertension was based on the modified Dandy criteria, we did not require lumbar puncture, as minor procedures are often poorly captured within claims data [ 23 ]. Furthermore, advancements in neuroimaging that allow clinicians to detect features of elevated intracranial pressure have led some experts to forego lumbar puncture among patients that meet the typical idiopathic intracranial hypertension phenotype [ 24 ]. LNG-IUD utilization has steadily increased from 2% to more than 10% of reproductive age women from 2002 to 2017 [ 25 ]. Women should be reassured that there is not a significantly increased risk of idiopathic intracranial hypertension among women with LNG-IUDs when compared to copper IUDs. Our findings can support shared decision making for women considering LNG-IUD placement or continued use.

Idiopathic intracranial hypertension predominately affects young, obese women of reproductive age and can cause headache and irreversible vision loss. The reports of an association between idiopathic intracranial hypertension and levonorgestrel intrauterine devices (IUD), containing a synthetic progestogen (i.e. progestin), are conflicting. Two retrospective studies documented increased odds of idiopathic intracranial hypertension in women using levonorgestrel intrauterine devices (LNG-IUDs) [ 1 , 2 ]. Both studies have serious methodologic limitations, including poorly defined inclusion/exclusion criteria, [ 1 , 3 ] use of inappropriate controls, [ 1 , 3 ] and lack of adjustment for confounders associated with both idiopathic intracranial hypertension and LNG-IUD use (e.g. age, obesity) [ 1 , 2 ]. A more recent case control study from a single center showed no association between LNG-IUD use and idiopathic intracranial hypertension, but only six women had LNG-IUD [ 4 ]. To address the limitations of prior studies, we used a large, national sample to evaluate our primary research question: do women exposed to LNG-IUDs have an increased hazard of idiopathic intracranial hypertension compared to those using copper IUDs, which have no hormone. We also explored the risk of incident idiopathic intracranial hypertension in women exposed to other highly effective forms of contraception, such as etonogestrel subcutaneous implants, as the systemic level of progestin is higher in comparison to LNG-IUD. We hypothesized that the hazard of idiopathic intracranial hypertension would not be increased among women using LNG-IUD compared to copper IUD after controlling for appropriate confounders.