Comparison of levonorgestrel level and creamatocrit in milk following immediate vs. delayed postpartum insertion of levonorgestrel IUD: a pilot randomized controlled trial

Background: Both breastfeeding and the use of postpartum contraception critically inuence infant and maternal health outcomes. In this pilot study, we explore the effects of timing and duration of postpartum levonorgestrel exposure on breastfeeding through objective assessments of milk lipid and levonorgestrel content to establish baseline data for future research. Methods: This substudy includes thirty-two participants from multiple clinical sites in Salt Lake City, Utah. All participants planned to breastfeed, and self-selected for this sub-study from among 259 women enrolled in a parent randomized controlled trial comparing immediate to delayed (4-8 weeks) postpartum levonorgestrel IUD insertion. Participants provided two milk samples: rst at 4-8 weeks postpartum (before IUD placement for the delayed group) and second four weeks later. Participants were paired by group assignment, and we used the Wilcoxon rank sum (inter-group) and signed rank (intra-group) tests to compare milk lipid content (creamatocrit) and levonorgestrel levels between groups and time points. Results: Fifteen participants were allocated to the immediate group and 17 to the delayed group. Initially, median levonorgestrel concentration of the immediate group (n=10) (32.5 pg/mL, IQR: 24.8, 59.4) exceeded that of the delayed group (n=12) (17.5 pg/mL, IQR: 0.0, 25.8) (p=0.01). Four weeks later, the values aligned: 26.2 pg/mL (IQR: 20.3, 37.3) vs. 28.0 pg/mL (IQR: 25.2, 40.8). Creamatocrits were similar between both groups and timepoints. Conclusions: Immediate postpartum levonorgestrel IUD placement results in steady, low levels of levonorgestrel in milk without signicant effects on lipid content. These ndings support the safety of immediate postpartum levonorgestrel IUD initiation, and future research should assess time-based meaningful outcome changes to capture clinically signicant differences. Trial Registration: This controlled trial registered (registry

Background Both breastfeeding and the use of postpartum contraception critically in uence not only infant but also maternal health outcomes (1,2). Immediate postpartum initiation of highly effective, long-acting contraception eliminates the risk of early pregnancy and avoids the need to return to the clinic for (typically less comfortable) intrauterine device (IUD) insertion (3). Understanding the effect of immediate postpartum hormonal contraception on breastfeeding through objective assessments of milk production and exogenous progestogen content can inform counseling and decision-making.
Four decades ago, researchers used an accurate but logistically challenging radioimmunoassay to assess levonorgestrel levels in milk from women with high-dose levonorgestrel IUDs (4). Using a new commercially-available enzyme immunoassay for levonorgestrel as well as traditional creamatocrit measurements, we sought exploratory data to assess differences in milk levonorgestrel and lipid content over time to provide point estimates for future research on exogenous progestogen and creamatocrit lipid levels among women receiving an early versus delayed postpartum levonorgestrel IUD.

Design
This exploratory secondary analysis derives from the Breastfeeding Levonorgestrel IUD Study, a randomized controlled noninferiority trial comparing breastfeeding continuation at eight weeks postpartum between women who received immediate (within 30 minutes) post-placental levonorgestrel IUD placement and delayed (4-8 weeks postpartum) placement. Thus, criteria for eligibility for this substudy were prior consent and enrollment in the parent study. Parent study details, including randomization allocation and other relevant recruitment data, have been previously reported (5). All substudy participants enrolled in Salt Lake City, Utah. The Institutional Review Board at the University of Utah approved the study [IRB#_00062844]. This sub-study adheres to the CONSORT guidelines for randomized trials.
From January 2014 through November 2016, research personnel used medical records to approach potential participants at multiple clinical sites in Salt Lake City. All participants had plans to breastfeed and a stated interest in the levonorgestrel IUD. Study participants provided written consent, and opted in to the sub-study using a checkbox on the parent study's consent documents. Figure 1 details the study timeline by assignment group. We asked all sub-study participants to provide self-expressed milk samples at both the rst postpartum visit (at 4-8 weeks) and the second four weeks later. In this exploratory study, we compared both levels of levonorgestrel and creamatocrit-a simple measure of lipid content in milk that can provide insight into the fat and energy content of human milk (6)-in the milk samples of women who received IUDs at different timepoints.

Sample
Of the 259 women who enrolled in the parent study, a portion self-selected to participate in this sub-study. All participants were age 18-40 years and had delivered a healthy term infant (³ 37 weeks gestation). Substudy participants returned for a follow-up visit four weeks after IUD insertion. We planned to cease recruitment when we obtained 12 paired (initial and four-week follow up) samples from each group. We oversampled by 25% in case of loss to follow-up (7).

Creamatocrit Measurement
Participants provided two mid-feeding milk samples (10ml each) at a 4-8 week postpartum clinic visit and second samples at home approximately four weeks later. For those in the delayed group, initial sample collection occurred at the same clinic visit as IUD placement. All other samples were collected after IUD placement, as in Figure 1. Research personnel transported samples to an -80 °C freezer within one hour of expression. Later, a research team member (NGH) blinded to group assignment thawed each sample and measured lipid levels using the creamatocrit method twice and reported mean values for analysis (6). Samples were returned to the -80 °C freezer to await the subsequent levonorgestrel assay.

Levonorgestrel Assay
We shipped frozen milk samples to Arbor Assays in Ann Arbor, Michigan, for analysis. There, technicians processed each sample using Arbor Assay DetectX Levonorgestrel Enzyme Immunoassay Kit instructions (8). Samples were run in duplicate with a known levonorgestrel standard. SoftMax® 4 parameter logistic tting software (Molecular Devices, San Jose, CA) was used to calculate results. See Supplement A for assay details and commentary.

Data Analysis
Participants were paired by group assignment, and data analysis was blinded. Using Stata 15 statistical software (StataCorp LP, College Station, TX USA), we conducted the Wilcoxon rank sum (inter-group) and signed rank (intra-group) tests to compare levonorgestrel levels and creamatocrit between groups and times. We also conducted a sensitivity analysis to compare outcomes among those who experienced IUD expulsions (and subsequent reinsertion) (expulsion in delayed group=1 [5%], expulsion in immediate group=2 [9%]) and no expulsion (n=19, [86%]). Table 1 presents participant characteristics. In total, 32 participants provided an initial milk sample at the rst postpartum visit. Of these, 22 (immediate=12, delayed=10) provided a second sample, and we limit analyses to this group. Ten participants (immediate=5, delayed=5) did not provide a second sample for reasons including discontinuing breastfeeding and IUD expulsion without reinsertion.

Discussion
This exploratory study provides point estimates for milk levonorgestrel levels and lipid levels in postpartum individuals initiating levonorgestrel IUD contraception immediately and 4-8 weeks later. We found that women receiving delayed IUD placement eventually have similar levonorgestrel milk levels and similar milk lipid content compared to those receiving immediate postpartum IUD placement. Our ndings align with previous research, which suggests the temporal stability of levonorgestrel levels in milk (4).
It is unclear why some participants in the delayed group had trace amounts of levonorgestrel in their initial milk sample before the insertion of a levonorgestrel IUD. Notably, our results are consistent with those from the assay development, which also identi ed trace amounts of levonorgestrel in urine samples of individuals (including males) not using a levonorgestrel-containing product (8). While these ndings may represent contamination or cross-reactivity of the assay, further research investigating background levonorgestrel levels in human milk may help clarify the signi cance of this nding.
Study limitations include small sample size and limited power to detect small differences between groups, between primiparous and multiparous women, across BMI, across body fat distribution, and across time by the number of weeks postpartum. Furthermore, all participants identi ed as Latina. Thus, although ethnicity likely does not affect lipid or hormone levels in human milk, results from this study may not be generalizable. Initially, we planned to obtain infant plasma levonorgestrel levels; however, no participants agreed to an infant blood draw. Moreover, at follow-up, many participants reported changes in breastfeeding behavior such as bottle feeding, supplementation, or exclusive arti cial milk feeding. Breastfeeding behavior affects milk composition, and these changes may have in uenced lipid levels. Lastly, creamatocrit has traditionally been used as a proxy measure of milk energy content. However, we now know that a variety of factors, including the time of day collected, time since the last feeding, whether fresh or stored, and whether foremilk or hindmilk may in uence creamatocrit (9).

Conclusion
Overall, our study provides evidence that immediate postpartum levonorgestrel IUD placement results in steady, low levels of levonorgestrel in milk without signi cant effects on lipid content. These ndings add to the growing body of literature supporting the safety of immediate postpartum levonorgestrel IUD initiation. While not powered to detect noninferiority, our study presents baseline data to inform future research. Future studies examining differences between immediate and delayed contraceptive use should assess time-based meaningful outcome changes to capture clinically signi cant differences.

Consent for Publication
Not applicable.

Availability of Data and Materials
The datasets used and analyzed during the current study are available from the corresponding author on reasonable request.