Pharmacologically functional and other common genetic variants across the genome fail to explain the wide interindividual variability in serum etonogestrel concentrations among contraceptive implant users.
Abstract
OBJECTIVE:
To identify novel genetic loci associated with differences in serum etonogestrel concentrations among contraceptive implant users.
METHODS:
We conducted a cross-sectional analysis in which we enrolled healthy, reproductive-aged (age 18–45 years) participants who had been using etonogestrel implants for 12–48 months. Participants underwent a single-time blood draw for measurement of serum etonogestrel concentrations by liquid chromatography–tandem mass spectrometry and the extraction of DNA from whole blood. We genotyped participants using the Illumina Infinium Global Diversity Array with Enhanced PGx and imputed genotyping results using the TOPMed imputation server. We performed genome-wide complex trait analysis using a linear mixed model leave-one-chromosome-out association analysis to identify genetic variants associated with serum etonogestrel concentrations.
RESULTS:
We enrolled 900 etonogestrel implant users, with a median age of 22.3 years (range 18.0–41.5 years), median body mass index (BMI) 26.0 (range 18.5–52.0), and median duration of implant use 27 months (range 12–48 months). Most participants self-reported their race as White (49.3%) and ethnicity as Hispanic or Latina (52.9%). Participants had a median serum etonogestrel concentration of 126.9 pg/mL (range 39.4–695.1 pg/mL). Including BMI, duration of implant use, and three principal components as covariates in the genome-wide complex trait analysis, we identified no genetic variants with minor allele frequencies at or above 5% that were associated with serum etonogestrel concentrations at genome-wide significance (P<5.0×10−8). When including rare genetic variants (minor allele frequencies at or above 1%), we discovered 10 genetic loci of interest (RNF114; LINC02405; SYNE1; TSPAN14; CRYZL2P-SEC16B; CHRNA9; RIMS1; CCDC88C; and CBL), all containing genetic variants associated with increased serum etonogestrel concentrations. Among these novel genetic loci associated with serum etonogestrel concentrations, only one (CRYZL2P-SEC16B) has potential, albeit limited, physiologic plausibility.
CONCLUSION:
Despite enhanced coverage for known pharmacogenomic variants, we found no significant associations between interindividual variability in contraceptive implant pharmacokinetics and genetic loci directly involved in exogenous steroid hormone metabolism.
CLINICAL TRIAL REGISTRATION:
Users of hormonal contraception have consistently demonstrated wide interindividual variability in serum drug concentrations (ie, pharmacokinetics) across various modalities (eg, pills, implants, injections) and formulations.1–4 This pharmacokinetic variability can partially account for why no hormonal contraceptive method has 100% efficacy, even if used perfectly.5 Additionally, differences in serum concentrations of exogenous steroid hormones contained within hormonal contraception may partially account for the similarly wide interindividual differences in side effect profiles (ie, pharmacodynamics).6 Despite the real risks of contraceptive failure and early contraceptive discontinuation due to bothersome side effects that may be caused by pharmacokinetic variability, we have a paucity of data on what factors cause such wide differences in drug concentrations among users of the exact same contraceptive method.5,7,8
Pharmacogenomics, a component of personalized medicine, is the study of the relationship between genetic variation and interindividual variability in drug disposition, response, and toxicity. A pharmacogenomic study with etonogestrel implant users evaluated the influence of specific single-nucleotide variants (SNVs) in genes involved in exogenous steroid hormone metabolism and regulation on steady-state serum etonogestrel concentrations.9 Though this study identified three SNVs associated with serum etonogestrel concentrations, none remained statistically significant. Thus, to address this pharmacogenomic knowledge gap with hormonal contraception and exogenous steroid hormones in general, we conducted a genome-wide association study with etonogestrel contraceptive implant users. We hypothesized that novel genetic variants are associated with serum etonogestrel concentrations and can account for the pharmacokinetic variability for this contraceptive method.
METHODS
We conducted a cross-sectional genome-wide association study to identify associations between genetic variants and serum etonogestrel concentrations among contraceptive implant users. We recruited English- or Spanish-speaking, reproductive-aged participants (age 18–45 years) with etonogestrel contraceptive implants in place for at least 12 months but no more than 48 months. We chose this minimum duration of implant use because the etonogestrel implant has a pharmacokinetic burst after insertion that resolves to a relative steady-state at 12 months of use and remains at this relative steady-state during continued use.10 For this study, we initially used an upper cutoff of 36 months for duration of implant use based on the U.S. Food and Drug Administration–approved duration for contraceptive efficacy.11 However, during the recruitment period for this study, new research became available that the etonogestrel implant maintains equivalent contraceptive efficacy definitively out to 48 months of use and likely out to 60 months of use.12,13 Given a change in the clinical practice to recommend usage of the etonogestrel implant for at least 48 months, we extended our inclusion criteria to match in February 2021.13 We determined the duration of implant use by participant report, corroborated with medical records if available, and confirmed presence of the implant by physical exam (palpation). We recruited participants through broad community advertising and directly at contraceptive and adolescent care clinics at the University of Colorado Anschutz Medical Campus and Children's Hospital Colorado located in Aurora, Colorado.
We excluded people who were younger than 18 and older than 45 years of age, owing to potential altered drug metabolism from aging.14 We also excluded individuals using medications or supplements that could alter serum etonogestrel levels through inhibition or induction of cytochrome P450 (CYP) enzymes (specifically CYP3A4). We reviewed concomitant medications through both participant report and medical record review, when available, and screened for any medications included in the U.S. Food and Drug Administration’s list of known CYP3A4 inhibitors or inducers.15 We similarly excluded people who reported any medical conditions that could affect baseline liver function (eg, hepatitis, cirrhosis). Because low body mass index (BMI, calculated as weight in kilograms divided by height in meters squared) has been associated with abnormal drug metabolism, we measured height and weight at the screening visit and excluded individuals with a measured BMI less than 18.5.16 We did not set an upper limit for BMI, because currently published studies have shown inconclusive or only minimal associations between high BMI and serum etonogestrel concentrations.1,17 The protocol for this study was approved by our Colorado Multiple IRB, and all participants gave written informed consent before initiating any study procedures.
After participants provided informed consent, we collected two blood samples from each participant at a single time point: one for serum etonogestrel concentration analysis and one for DNA extraction. Of note, participants from a prior candidate gene study were included in this genome-wide association study if they previously consented for future genetic research using their specimens and data.9 We also collected self-reported race and ethnicity from participants, because these outcomes have strong correlations with genetic ancestry and are scientifically important variables for pharmacogenomic research.18
For the serum etonogestrel analysis, we collected one of the blood samples in a red top tube and allowed the blood to clot for at least 10 minutes at room temperature. We then centrifuged the samples and stored the serum in aliquots at −80°C. We shipped batches of deidentified frozen serum samples to the Biomarkers Core Laboratory of the Irving Institute of Clinical and Translational Research at Columbia University Medical Center. We used a previously validated liquid chromatography-mass-spectrometry assay protocol for quantification of serum etonogestrel concentrations.19 The batches consisted of 100–150 samples at a time to reduce inter-assay variability while ensuring safe delivery of the samples to the Biomarkers Core Laboratory. The lower limit of quantification, defined as the level at which the residual of the calibration line was less than 20% of the expected concentration combined with a signal-to-noise ratio greater than 10, was determined to be 20 pg/mL. For all analyses, we included three levels (low, medium, and high) of quality controls along with a study pooled sample to assess batch effects. The mean intra-assay precision for the assay was 3.2%. The inter-assay variability across the batches, assessed using the pooled sample, was 1.8%.
We extracted genomic DNA from whole blood samples using commercially available Qiagen kits. We then performed genotyping using the GDA (Illumina Infinium Global Diversity Array with Enhanced PGx), available through a collaboration with CARGO (Colorado Anschutz Genetics Organization). We selected the GDA because it includes enhanced pharmacogenomic coverage while encompassing genetic variation across multiple racial and ethnic populations.20 We performed automated DNA amplification and genotyping by using the standard GDA protocol that includes quality controls and by using the Genome Reference Consortium Human Build 38. HapMap and duplicate samples were included as quality control measures across all GDA plates. The Colorado Anschutz Genetics Organization performed the initial quality assessment of both samples and genetic variants passing the missing call rate threshold recommended by Illumina. The minimum sample call rate was 99.1% for the ratio of genetic variants that could be called among all genetic variants assessed in each sample. Samples with call rates below this threshold were removed from the analysis. The minimum genetic variant call rate was 97.0%, below which genetic variants were removed from the analysis. We then performed a secondary check for genetic variant missingness and excluded any genetic variants remaining that exceeded a 0.03 missingness threshold.
We evaluated for severe violations of Hardy-Weinberg equilibrium (HWE) and determined both genetic variant call rates and minor allele frequencies as quality control measures.21,22 If a diverse study population was obtained, we planned to stratify participants by self-reported race and ethnicity for HWE evaluation. We then used Q-Q plots of the HWE P-values to select an appropriate cutoff for exclusion of genetic variants that severely violated HWE. We also excluded all monomorphic alleles and rare variants (minor allele frequencies less than 0.01) from the data set. We then evaluated for kinship between participants using Plink 2.0.23 We selected a cutoff of 0.08838 for kinship based on the geometric mean between second- and third-degree relative relatedness. Participants exceeding this kinship factor were excluded from our principal components analysis but were included in our genome-wide association study because we used genome-wide complex trait analysis that could incorporate kinship factors into the analysis.24 Using Plink 2.0, we performed principal components analysis to identify the most informative principal components for our study population.23 We then selected the top three principal components to include as covariates for our genome-wide complex trait analysis.
We used the TOPMed Imputation Server to perform imputation of our array results using the Genome Reference Consortium Human Build 38 reference genome and the most recent panel that best fit the racial and ethnic mix of our study population.25–27 We then filtered the imputed results to include only genetic variants that were genotyped on the GDA or had an r2≥0.7. From the filtered imputed genotypes, we excluded rare genetic variants based on a minor allele frequency of less than 0.05 for our primary analysis, because rare variants have a higher chance of creating false-positive associations in our genome-wide association study, but planned a secondary analysis including rare variants (minor allele frequency 0.01 or greater) for exploratory identification of additional novel genetic loci.28
We used SPSS 28 statistical software for all nongenetic statistical analyses. We performed descriptive frequencies and conducted both univariate and multivariate linear regression modeling to identify variables associated with serum etonogestrel concentrations. We chose the pertinent participant characteristics and demographics of age, BMI, duration of implant use, self-reported race, and self-reported ethnicity as potential variables for our model.1 For multivariable modeling, we used a backwards stepwise approach where all variables of interest are entered into the model initially. We then sequentially removed variables without significant associations (P>.05) until we obtained a model with the minimal Akaike information criterion value.29
For our genetic statistical analyses, we performed genome-wide complex trait analysis that included all participants using Plink 2.0.24 We used the linear mixed model based association analysis tool within genome-wide complex trait analysis combined with the leave-one-chromosome-out option.28 A linear mixed model–based association analysis can prevent false-positive associations by building a genetic relationship matrix modeling genome-wide sample structure (including relatedness), estimating its contribution to phenotypic variance using a random-effects model, and then computing association statistics that account for this component of the phenotypic variance.28 The linear mixed model–leave-one-chromosome-out analysis additionally excludes the chromosome on which the candidate genetic variant is located from calculating the genetic relationship matrix, thereby avoiding double-fitting the candidate genetic variant into the model and resulting in increased statistical power.28 For covariates, we included all variables significantly associated with serum etonogestrel concentrations using multivariable linear regression described above and the top three principal components. We plotted the genome-wide complex trait analysis results in Manhattan plots using the qqman package in R statistical software to identify genetic loci of interest and key genetic variants.30
For our sample size calculation, we used the median and range of serum etonogestrel concentrations from a prior candidate gene study of 350 contraceptive implant users to calculate a population mean and standard deviation.9 We then used Quanto to determine that a sample size of 900 participants would give us a power of 0.84 to identify a genetic variant (assuming 5% prevalence) with an R2 association of at least 0.045 and a significance cutoff of 5.0×10−8. Because study procedures entailed only a single visit with no potential for loss to follow-up, we planned to enroll 900 total participants for this study.
RESULTS
From the prior candidate gene study, 335 participants consented to future genetic research and so were included in this larger study.9 We then enrolled an additional 565 participants (Fig. 1) over 40 months (January 2019 to April 2022). Among all 900 participants, the median age was 22.3 years (range 18.0–41.5 years), median BMI was 26.0 (range 18.5–52.0), and median duration of implant use was 27 months (range 12–48 months) (Table 1). Participants predominantly self-reported their race as White (49.3%, 444/900) and ethnicity as Hispanic or Latina (52.9%, 476/900). Table 1 shows the breakdown of self-reported race and ethnicity across all categories.
Fig. 1. Flow diagram of participant recruitment. *Participants excluded based on duration of implant use before expansion of inclusion criteria from 12–36 months to 12–48 months.
Lazorwitz. Etonogestrel Implant, Genome-Wide Association Study. O&G Open 2025.
Table 1.
Characteristics and Demographics for the Etonogestrel Contraceptive Implant Users (N=900)
| Characteristic | Value |
| Age (y) | 22.3 (18.0–41.5) |
| BMI (kg/m2) | 26.0 (18.5–52.0) |
| Duration of implant use (mo) | 27.0 (12–48) |
| Race | |
| American Indian or Alaska Native | 21 (2.3) |
| Asian or Pacific Islander | 53 (5.9) |
| Black or African American | 99 (11.0) |
| More than 1 | 98 (10.9) |
| No response or unknown | 185 (20.6) |
| White | 444 (49.3) |
| Ethnicity | |
| Hispanic or Latina | 476 (52.9) |
| Non-Hispanic | 424 (47.1) |
Data are median (range) or n (%).
Two participants had serum etonogestrel concentrations below the lower limit of the assay (less than 20 pg/mL) and were excluded from all pharmacokinetic analyses. For the remaining 898 participants, the median serum etonogestrel concentration was 126.9 pg/mL (range 39.4–695.1 pg/mL). Figure 2 shows the distribution of serum etonogestrel concentrations using box plots. The majority of participants had serum etonogestrel concentrations above the threshold for consistent ovulatory suppression with the etonogestrel implant (90 pg/mL), with 19.5% (175/898) of participants having a serum etonogestrel concentration below this cutoff.10 Using univariate linear regression modeling, longer duration of implant use and higher BMI were both significantly associated with lower serum etonogestrel concentrations (β=−1.58, P=3.3×10−11 and β=−3.10, P=2.0×10−17, respectively). Age, self-reported race, and self-reported ethnicity were not significantly associated with serum etonogestrel concentrations (P>.05). Both duration of implant use (β=−1.61, P=2.0×10−12) and BMI (β=−3.14, P=1.3×10−18) remained significantly associated with serum etonogestrel concentrations in multivariable linear regression modeling.
Fig. 2. Box plot of serum etonogestrel concentrations for 898 participants. The box represents the first and third quartiles (interquartile range [IQR] 69.6 pg/mL), with the band inside the box representing the median (126.9 pg/mL). Whiskers represent the data within 1.5 IQR of the upper and lower quartile. Circles indicate outliers with values between 1.5 and 3 times the IQR, and asterisks indicate outliers with values greater than 3 times the IQR. The red dotted line indicates 90 pg/mL, the threshold for consistent ovulatory suppression.
Lazorwitz. Etonogestrel Implant, Genome-Wide Association Study. O&G Open 2025.
Overall, 942 samples underwent genotyping on the GDA, including all 900 participant samples, 25 HapMap control samples, and 17 duplicate samples. Among all samples, the median call rate was 99.92%, with a minimum call rate of 33.96% and maximum of 99.97%. Based on CARGO's threshold, 880 participant samples passed quality assessment and were used for genetic analyses. In total, 1,883,463 genetic variants were analyzed, with 1,858,966 (98.70%) passing quality assessment based on CARGO's quality control measurements. As additional quality control measures, average HapMap concordance with 1,000 Genomes was found to be 99.94% and duplicate sample concordance was 99.06%. Of note, two samples had gender discrepancies between their reported female sex and genetic sex. These two participants were included in the final analyses as their samples passed all other quality control measures.
Among the 880 participants with genotyping data, 464 (52.7%) self-identified as Hispanic or Latina, 242 (27.5%) self-identified as White and non-Hispanic, and 86 (9.8%) self-identified as Black and non-Hispanic. Given this diversity in the study population, we stratified participants into those three racial and ethnic groups for HWE evaluation. Based on the Q-Q plots, we selected a HWE P-value cutoff of <10−3 for removal of genetic variants, which resulted in exclusion of 10,008 total genetic variants from the data set. When evaluating kinship among participants, we identified 22 (2.5%) participants with kinship factors above our selected cutoff of 0.08838. We therefore used data from the remaining 858 participants for our principal components analysis and selected the top three principal components as covariates for our genome-wide association study analyses (Appendix 1, available online at http://links.lww.com/AOG/E8).
After imputation and filtering, we had 40,684,396 genetic variants available for analysis. We then further filtered the genetic variants based on minor allele frequency (including only 0.05 or greater), which left a total of 6,720,064 genetic variants. As described above, we performed genome-wide complex trait analysis using the linear mixed model–leave-one-chromosome-out option to identify genetic variants associated with serum etonogestrel concentrations. Including the covariates of BMI, duration of implant use, and our top three principal components in the genome-wide complex trait analysis, we found no genetic variants associated with serum etonogestrel concentrations meeting the cutoff for genome-wide significance (P<5.0×10−8) (Fig. 3). Using an expanded minor allele frequency cutoff of 0.01 or greater for the genome-wide complex trait analysis (11,080,935 genetic variants), we identified 17 genetic variants with associations meeting the cutoff for genome-wide significance (P<5.0×10−8) (Fig. 4). Table 2 contains detailed information regarding these 17 genetic variants including their rsIDs, minor allele frequencies, β-coefficients, and P-values. For all 17 genetic variants, participants with the variant allele had higher serum etonogestrel concentrations compared with participants with the respective wild-type allele. Only two genetic variants were associated with gene coding consequences: a missense variant in the CRYZL2P-SEC16B gene (rs34246968), and an insertion and deletion (indel) variant in a noncoding region of chromosome 4 (rs145524315). The remaining 15 genetic variants were intronic variants, synonymous variants, or variants of unknown consequence.
Fig. 3. Manhattan plot of genome-wide complex trait analysis results for serum etonogestrel concentrations using minor allele frequencies cutoff of 0.05. The red line indicates the threshold for genome-wide significance (5.0×10−8), and the blue line indicates the threshold for suggestive significance (1.0×10−5).
Lazorwitz. Etonogestrel Implant, Genome-Wide Association Study. O&G Open 2025.
Fig. 4. Manhattan plot of genome-wide complex trait analysis results for serum etonogestrel concentrations using minor allele frequencies cutoff of 0.01. The red line indicates the threshold for genome-wide significance (5.0×10−8), and the blue line indicates the threshold for suggestive significance (1.0×10−5).
Lazorwitz. Etonogestrel Implant, Genome-Wide Association Study. O&G Open 2025.
Table 2.
Single-Nucleotide Variants With Minor Allele Frequency 0.01 or Higher Meeting Genome-Wide Significance Threshold for Association With Serum Etonogestrel Concentrations Among 880 Contraceptive Implant Users
| rsID | Chr | BP | A1 | A2 | Gene | MAF | β | P |
| rs192047010 | 20 | 50035247 | A | G | NA | 0.011 | 101.87 | 3.27×10−13 |
| rs142433765 | 20 | 49902251 | A | G | NA | 0.011 | 101.39 | 4.11×10−13 |
| rs7549228 | 20 | 49937799 | T | C | RNF114 | 0.012 | 93.55 | 7.41×10−12 |
| rs149649642 | 12 | 126918787 | G | A | LINC02405 | 0.010 | 85.57 | 6.01×10−9 |
| rs372175486 | 6 | 152468357 | G | GT | SYNE1 | 0.011 | 81.78 | 1.08×10−8 |
| rs58877385 | 10 | 80505096 | T | C | TSPAN14 | 0.020 | 65.15 | 1.32×10−8 |
| rs112170255 | 10 | 127597603 | A | G | NA | 0.011 | 79.29 | 1.45×10−8 |
| rs79329941 | 5 | 14089152 | C | T | NA | 0.021 | 56.46 | 1.59×10−8 |
| rs34246968 | 1 | 177960366 | G | C | CRYZL2P-SEC16B | 0.014 | 70.62 | 1.80×10−8 |
| rs145524315 | 4 | 54502499 | CGGTGGTG | C | NA | 0.016 | 66.73 | 1.84×10−8 |
| rs55974552 | 4 | 40335951 | A | G | CHRNA9 | 0.014 | 70.34 | 1.86×10−8 |
| rs137960049 | 3 | 71851406 | T | C | NA | 0.014 | 68.06 | 2.69×10−8 |
| rs564887776 | 6 | 72004251 | G | T | RIMS1 | 0.010 | 81.61 | 2.70×10−8 |
| rs111801179 | 6 | 72011033 | A | C | RIMS1 | 0.010 | 81.61 | 2.70×10−8 |
| rs112741166 | 6 | 72011730 | C | T | RIMS1 | 0.010 | 81.61 | 2.70×10−8 |
| rs61220159 | 14 | 91339460 | T | C | CCDC88C | 0.011 | 71.74 | 4.15×10−8 |
| rs147749042 | 11 | 119244102 | A | C | CBL | 0.013 | 71.71 | 4.30×10−8 |
rsID, reference single nucleotide polymorphism cluster ID; Chr, chromosome; BP, base position; A1, variant (minor) allele; A2, wild-type (major) allele; MAF, minor allele frequency; NA, not applicable.
Based on the Manhattan plot (Fig. 3), a genetic locus of potential interest was evident in chromosome 2, but did not meet the threshold for genome-wide significance. This signal was present around base positions 123201859–123253526, with rs1033292 having the most significant association (β=18.7, P=7.0×10−8), but no gene is known at this location.
DISCUSSION
In this genome-wide association study of etonogestrel contraceptive implant users, we did not identify genetic variants in metabolic pathways significantly associated with alterations in serum etonogestrel concentrations. We did find a novel association between a genetic locus in chromosome 2 and increased serum etonogestrel concentrations that approached genome-wide significance, but with unknown clinical significance given the lack of encoded genes in this locus. From our exploratory analyses, we identified 17 novel genetic variants associated with serum etonogestrel concentrations, with 11 found in known genes and one (rs34246968) with a functional coding consequence.
The variant rs34246968 is a missense variant found in the gene CRYZL2P-SEC16B on chromosome 1, which is a naturally occurring read-through transcription site between two neighboring genes: crystallin zeta-like 2, pseudogene (CRYZL2P) and SEC16 homolog B, endoplasmic reticulum export factor (SEC16B).31 The encoded protein (SEC16B) primarily functions as a protein transporter for the endoplasmic reticulum in the small intestine and liver, and several genome-wide association studies have identified associations between SEC16B and obesity.32–35 Though SEC16B is a protein transporter, no currently published studies have evaluated if SEC16B has functional involvement in the transport of metabolic enzymes such as CYP.33 Because CYP enzymes (primarily CYP3A) are responsible for the majority of exogenous steroid hormone metabolism, less available CYP enzymes could lead to higher serum etonogestrel concentrations.5 However, there are no currently published studies that have identified any other genetic variants in SEC16B associated with alterations in drug metabolism and data are needed regarding actual intracellular involvement of SEC16B in CYP enzyme transport.36,37
Based on our prior candidate gene study, we hypothesized that variants previously associated with serum etonogestrel concentrations would maintain associations in this larger study (Table 3).9 Unfortunately, we did not find this to be true. Of particular interest was CYP3A7*1C denoted by rs45446698, which was again associated with lower serum etonogestrel concentrations (β=−26.1, P=8.1×10−3).9 However, we were dramatically underpowered to find a statistically significant association between CYP3A7*1C and serum etonogestrel concentrations with this β-coefficient and a minor allele frequency of only 0.022 in this study. We would have needed 16,660 participants (including 367 CYP3A7*1C carriers) to achieve statistical significance (P<5.0×10−8).38 Thus, the CYP3A7*1C variant remains a physiologically plausible candidate, but will require much larger studies to demonstrate statistically significant effects.9 The other two variants (PXR rs2461817, PGR rs537681) flipped in directionality of their associations from the prior study (Table 3).9 As was found in the prior candidate gene study, clinically significant variants in major CYP3A enzymes (eg, CYP3A5 rs776746 [*3], CYP3A4 rs35599367 [*22]) maintained no significant associations with serum etonogestrel concentrations (β=−1.8, P=.61 and β=20.4, P=.02; respectively).9,39
Table 3.
Comparison of Associations Found Between Candidate Genetic Variants and Serum Etonogestrel Concentrations in the Prior Candidate Gene Study With Etonogestrel Contraceptive Implant Users (n=350)11 With Findings From This Genome-Wide Association Study (n=880, Including 335 From the Prior Study)
| Genetic Variant | Candidate Gene Study Findings | GWAS Study Findings | ||
| β-coefficient* | P | β-coefficient | P | |
| NR1I2 (PXR) rs2461817 | 13.36 | .005 | −0.60 | .84 |
| PGR rs537681 | −29.77 | .007 | 4.5 | .15 |
| CYP3A7*1C | −35.06 | .025 | −26.1 | 8.1×10−3 |
GWAS, genome wide association study.
Calculated using multivariable generalized linear modeling.11
The major strength of this study was our metabolic phenotypic assessment, because we effectively and reliably captured each individual's pharmacokinetic profile using a single-time blood draw and directly measured the active progestin (etonogestrel) released by the contraceptive implant.10,40 The racial and ethnic heterogeneity of our study population was also a study strength; this admixture increases the generalizability of our findings to more diverse and varied populations. Finally, the use of a genotyping array with enhanced pharmacogenomic coverage ensured that we assessed a broad range of clinically significant genetic variants in key metabolic enzymes (eg, CYP enzymes).20
The most important limitation of our study was the sample size, because we were not powered to detect minor effects or effects of rare variants (less than 1% prevalence). By using a genome-wide association study approach, we also found many exploratory associations in noncoding genetic loci of unknown significance. Though noncoding genetic variants can have potential functional consequences on nearby genes, discerning the exact mechanisms behind noncoding genetic variants often requires creating model cells or organisms and complex computational tools.41 Further, larger independent studies are needed to verify our findings given the lack of clear physiologic plausibility for these genetic variants. We also only evaluated a small portion of the entire human genome in this study, and use of sequencing technology, such as whole exome sequencing, could evaluate millions more genetic variants, but would require very large sample sizes or complex analytical processes. Finally, though we had good heterogeneity in our study population, the cohort was not representative of all minority groups (eg, self-reported Asian or Pacific Islander, self-reported American Indian or Alaska Native), and thus may not be generalizable to these populations.
Ultimately, we did not find any genetic loci directly involved in steroid hormone metabolism or regulation that were associated with the primary outcome of this study, serum etonogestrel concentrations. More importantly, we found that clinically significant variants in key metabolic genes (eg, CYP3A5, CYP3A4) do not appear to influence the metabolism of etonogestrel among contraceptive implant users. Though we did not identify any clinically actionable metabolic-related drug-gene pairs, we have now established one of the largest biobanks of reproductive-aged, hormonal contraceptive users with pertinent pharmacokinetic phenotypes that cannot be found in existing biobanks. As researchers work to advance the historically lagging field of pharmacogenomics in women's health care, these kinds of biobanks will eventually create large enough study populations that can lead to a far better understanding of how individual genetic differences influence the wide pharmacokinetic variability experienced by millions of individuals using hormonal contraception.
Authors' Data Sharing Statement
Will individual participant data be available (including data dictionaries)? Yes, individual participant data will be uploaded to dbGaP.
What data in particular will be shared? Genotypes and the phenotypes of serum etonogestrel concentrations.
What other documents will be available? Study protocol and consent will be available on CT.gov: https://clinicaltrials.gov/study/NCT03092037.
When will data be available (start and end dates)? Deidentified data will be uploaded to dbGaP by December 2024 and remain available indefinitely.
By what access criteria will data be shared (including with whom, for what types of analyses, and by what mechanism)? Data will be shared according to dbGaP data use criteria.
Footnotes
This study was primarily supported by the Society of Family Planning Research Fund (grant number SFPRFSS19-01) and the Eunice Kennedy Shriver National Institute of Child Health and Human Development (R03 HD101551). This work was also supported by NIH/NCATS Colorado CTSA Grant Number UL1 TR001082. Dr. Lazorwitz's time and effort was supported by the NICHD K12 Women's Reproductive Health Research Scholar program (grant number 5K12HD001271-18). Contents are the authors' sole responsibility and do not necessarily represent official NIH views. All funding sources had no involvement in the study design, collection, analysis, interpretation of data, writing of this report, or decision to submit this article for publication.
Financial Disclosure Aaron Lazorwitz serves as Chief Medical Advisor for Dama Health and serves on a scientific advisory board for 3Daughters. Stephanie Teal serves on a Data Monitoring Board for a study funded by Merck and Co and has served as a consultant for Bayer Healthcare. The University of Colorado Department of Obstetrics and Gynecology has received research funding from Bayer, Organon, Agile Therapeutics, and Medicines360. The other authors did not report any potential conflicts of interest.
The authors thank Dr. Renu Nandakumar at the Biomarkers Core Laboratory at Columbia University for leading the etonogestrel analysis for this study. The authors also thank the Colorado Anschutz Research Genetics Organization lab and particularly Meher Boorgula and Monica Campbell for leading the Illumina Infinium Global Diversity Array analysis for this study.
Each author has confirmed compliance with the journal's requirements for authorship.
Peer reviews and author correspondence are available at http://links.lww.com/AOG/E9.
REFERENCES
- 1.Lazorwitz A, Aquilante CL, Sheeder J, Guiahi M, Teal S. Relationship between patient characteristics and serum etonogestrel concentrations in contraceptive implant users. Contraception 2019;100:37–41. doi: 10.1016/j.contraception.2019.03.045 [DOI] [PMC free article] [PubMed] [Google Scholar]
- 2.Fotherby K. Variability of pharmacokinetic parameters for contraceptive steroids. J Steroid Biochem 1983;19:817–20. doi: 10.1016/0022-4731(83)90017-1 [DOI] [PubMed] [Google Scholar]
- 3.McNicholas C, Swor E, Wan L, Peipert JF. Prolonged use of the etonogestrel implant and levonorgestrel intrauterine device: 2 years beyond Food and Drug Administration-approved duration. Am J Obstet Gynecol 2017;216:586.e1–6. doi: 10.1016/j.ajog.2017.01.036 [DOI] [PMC free article] [PubMed] [Google Scholar]
- 4.Westhoff CL, Torgal AH, Mayeda ER, Pike MC, Stanczyk FZ. Pharmacokinetics of a combined oral contraceptive in obese and normal-weight women. Contraception 2010;81:474–80. doi: 10.1016/j.contraception.2010.01.016 [DOI] [PMC free article] [PubMed] [Google Scholar]
- 5.Hatcher RA, Nelson AL, Trussell J, Cwiak C, Cason P, Policar MS, et al. Contraceptive technology. 21st ed. Managing Contraception; 2018. [Google Scholar]
- 6.Lazorwitz A, Aquilante CL, Dindinger E, Harrison M, Sheeder J, Teal S. Relationship between etonogestrel concentrations and bleeding patterns in contraceptive implant users. Obstet Gynecol 2019;134:807–13. doi: 10.1097/AOG.0000000000003452 [DOI] [PMC free article] [PubMed] [Google Scholar]
- 7.Grunloh DS, Casner T, Secura GM, Peipert JF, Madden T. Characteristics associated with discontinuation of long-acting reversible contraception within the first 6 months of use. Obstet Gynecol 2013;122:1214–21. doi: 10.1097/01.AOG.0000435452.86108.59 [DOI] [PMC free article] [PubMed] [Google Scholar]
- 8.Cohen R, Sheeder J, Teal SB. Predictors of discontinuation of long-acting reversible contraception before 30 Months of use by adolescents and young women. J Adolesc Health 2019;65:295–302. doi: 10.1016/j.jadohealth.2019.02.020 [DOI] [PubMed] [Google Scholar]
- 9.Lazorwitz A, Aquilante CL, Oreschak K, Sheeder J, Guiahi M, Teal S. Influence of genetic variants on steady-state etonogestrel concentrations among contraceptive implant users. Obstet Gynecol 2019;133:783–94. doi: 10.1097/AOG.0000000000003189 [DOI] [PMC free article] [PubMed] [Google Scholar]
- 10.Le J, Tsourounis C. Implanon: a critical review. Ann Pharmacother 2001;35:329–36. doi: 10.1345/aph.10149 [DOI] [PubMed] [Google Scholar]
- 11.U.S. Food and Drug Administration . Nexplanon (etonogestrel implant). Reference ID: 3808594. Accessed June 1, 2023. https://www.accessdata.fda.gov/drugsatfda_docs/label/2015/021529s011lbl.pdf [Google Scholar]
- 12.Ali M, Bahamondes L, Bent Landoulsi S. Extended effectiveness of the etonogestrel-releasing contraceptive implant and the 20 µg levonorgestrel-releasing intrauterine system for 2 years beyond U.S. Food and Drug Administration product labeling. Glob Health Sci Pract 2017;5:534–9. doi: 10.9745/GHSP-D-17-00296 [DOI] [PMC free article] [PubMed] [Google Scholar]
- 13.Dethier D, Qasba N, Kaneshiro B. Society of Family Planning clinical recommendation: extended use of long-acting reversible contraception. Contraception 2022;113:13–8. doi: 10.1016/j.contraception.2022.06.003 [DOI] [PubMed] [Google Scholar]
- 14.Wilson K, Hanson J. The effects of extremes of age on drug action. Methods Find Exp Clin Pharmacol 1980;2:303–12. [PubMed] [Google Scholar]
- 15.U.S. Food and Drug Administration. Drug development and drug interactions: table of substrates, inhibitors, and inducers. Accessed January 17, 2017. https://www.fda.gov/drugs/developmentapprovalprocess/developmentresources/druginteractionslabeling/ucm093664.htm [Google Scholar]
- 16.Yang CS. Influences of dietary and other factors on xenobiotic metabolism and carcinogenesis-A review article in memory of Dr. Allan H. Conney (1930-2013). Nutr Cancer 2015;67:1207–13. doi: 10.1080/01635581.2015.1081010 [DOI] [PubMed] [Google Scholar]
- 17.Morrell KM, Cremers S, Westhoff CL, Davis AR. Relationship between etonogestrel level and BMI in women using the contraceptive implant for more than 1 year. Contraception 2016;93:263–5. doi: 10.1016/j.contraception.2015.11.005 [DOI] [PubMed] [Google Scholar]
- 18.Lazorwitz A, Aquilante CL, Shortt JA, Sheeder J, Teal S, Gignoux CR. Applicability of ancestral genotyping in pharmacogenomic research with hormonal contraception. Clin Transl Sci 2021;14:1713–8. doi: 10.1111/cts.13014 [DOI] [PMC free article] [PubMed] [Google Scholar]
- 19.Thomas T, Petrie K, Shim J, Abildskov KM, Westhoff CL, Cremers S. A UPLC-MS/MS method for therapeutic drug monitoring of etonogestrel. Ther Drug Monit 2013;35:844–8. doi: 10.1097/FTD.0b013e31829a10fa [DOI] [PMC free article] [PubMed] [Google Scholar]
- 20.Illumina. Infinium global diversity array-8 v1.0: Data Sheet. Accessed January 4, 2022. https://www.illumina.com/content/dam/illumina/gcs/assembled-assets/marketing-literature/infinium-global-diversity-array-data-sheet-m-gl-00153/infinium-global-diversity-array-data-sheet-m-gl-00153.pdf [Google Scholar]
- 21.Cox DG, Kraft P. Quantification of the power of Hardy-Weinberg equilibrium testing to detect genotyping error. Hum Hered 2006;61:10–4. doi: 10.1159/000091787 [DOI] [PubMed] [Google Scholar]
- 22.Pearson TA, Manolio TA. How to interpret a genome-wide association study. JAMA 2008;299:1335–44. doi: 10.1001/jama.299.11.1335 [DOI] [PubMed] [Google Scholar]
- 23.Chang CC, Chow CC, Tellier LC, Vattikuti S, Purcell SM, Lee JJ. Second-generation PLINK: rising to the challenge of larger and richer datasets. Gigascience 2015;4:7. doi: 10.1186/s13742-015-0047-8 [DOI] [PMC free article] [PubMed] [Google Scholar]
- 24.Yang J, Lee SH, Goddard ME, Visscher PM. GCTA: a tool for genome-wide complex trait analysis. Am J Hum Genet 2011;88:76–82. doi: 10.1016/j.ajhg.2010.11.011 [DOI] [PMC free article] [PubMed] [Google Scholar]
- 25.National Heart, Lung, and Blood Institute. TOPMed Whole Genome Sequencing Methods: Freeze 8. NHLBI Trans-Omics for Precision Medicine. Accessed May 19, 2020. https://www.nhlbiwgs.org/topmed-whole-genome-sequencing-methods-freeze-8
- 26.Taliun D, Harris DN, Kessler MD, Carlson J, Szpiech ZA, Torres R, et al. Sequencing of 53,831 diverse genomes from the NHLBI TOPMed program. Nature 2021;590:290–9. doi: 10.1038/s41586-021-03205-y [DOI] [PMC free article] [PubMed] [Google Scholar]
- 27.Das S, Forer L, Schonherr S, Sidore C, Locke AE, Kwong A, et al. Next-generation genotype imputation service and methods. Nat Genet 2016;48:1284–7. doi: 10.1038/ng.3656 [DOI] [PMC free article] [PubMed] [Google Scholar]
- 28.Yang J, Zaitlen NA, Goddard ME, Visscher PM, Price AL. Advantages and pitfalls in the application of mixed-model association methods. Nat Genet 2014;46:100–6. doi: 10.1038/ng.2876 [DOI] [PMC free article] [PubMed] [Google Scholar]
- 29.Burnham KP, Anderson DR. Multimodel inference:understanding AIC and BIC in model selection. Sociological Methods Res 2004;33:261–304. doi: 10.1177/0049124104268644 [DOI] [Google Scholar]
- 30.Turner S. qqman: an R package for visualizing GWAS results using Q-Q and Manhattan plots. J Open Source Softw 2018;3:1–2. [Google Scholar]
- 31.Kitts A, Phan L, Ward M, Holmes JB. The database of short genetic variation (dbSNP). Accessed April 3, 2014. https://www.ncbi.nlm.nih.gov/books/NBK174586/
- 32.Huang J, Huffman JE, Huang Y, Do Valle Í, Assimes TL, Raghavan S, et al. Genomics and phenomics of body mass index reveals a complex disease network. Nat Commun 2022;13:7973. doi: 10.1038/s41467-022-35553-2 [DOI] [PMC free article] [PubMed] [Google Scholar]
- 33.Budnik A, Heesom KJ, Stephens DJ. Characterization of human Sec16B: indications of specialized, non-redundant functions. Scientific Rep 2011;1:77. doi: 10.1038/srep00077 [DOI] [PMC free article] [PubMed] [Google Scholar]
- 34.Yılmaz B, Gezmen Karadağ M. The current review of adolescent obesity: the role of genetic factors. J Pediatr Endocrinol Metab 2021;34:151–62. doi: 10.1515/jpem-2020-0480 [DOI] [PubMed] [Google Scholar]
- 35.Shi R, Lu W, Tian Y, Wang B. Intestinal SEC16B modulates obesity by regulating chylomicron metabolism. Mol Metab 2023;70:101693. doi: 10.1016/j.molmet.2023.101693 [DOI] [PMC free article] [PubMed] [Google Scholar]
- 36.Neve EP, Ingelman-Sundberg M. Intracellular transport and localization of microsomal cytochrome P450. Anal Bioanal Chem 2008;392:1075–84. doi: 10.1007/s00216-008-2200-z [DOI] [PubMed] [Google Scholar]
- 37.Whirl-Carrillo M, Huddart R, Gong L, Sangkuhl K, Thorn CF, Whaley R, et al. An evidence-based framework for evaluating pharmacogenomics knowledge for personalized medicine. Clin Pharmacol Ther 2021;110:563–72. doi: 10.1002/cpt.2350 [DOI] [PMC free article] [PubMed] [Google Scholar]
- 38.QUANTO 1.1 : a computer program for power and sample size calculations for genetic-epidemiology studies. Accessed September 1, 2018. https://keck.usc.edu/biostatistics/software/ [Google Scholar]
- 39.Deininger KM, Vu A, Page RL, II, Ambardekar AV, Lindenfeld J, Aquilante CL. CYP3A pharmacogenetics and tacrolimus disposition in adult heart transplant recipients. Clin Transplant 2016;30:1074–81. doi: 10.1111/ctr.12790 [DOI] [PubMed] [Google Scholar]
- 40.Lazorwitz A, Sheeder J, Teal S. Variability in repeat serum etonogestrel concentrations among contraceptive implant users during the steady-release pharmacokinetic period. Contraception 2022;108:65–8. doi: 10.1016/j.contraception.2021.12.008 [DOI] [PMC free article] [PubMed] [Google Scholar]
- 41.Zhang F, Lupski JR. Non-coding genetic variants in human disease. Hum Mol Genet 2015;24:R102–10. doi: 10.1093/hmg/ddv259 [DOI] [PMC free article] [PubMed] [Google Scholar]