Reversible Effects of Oral Contraceptive Removal on Serotonergic Neurotransmission

• Background:

Large register-based studies have shown that initiation of combined oral contraceptives (COCs) is associated with an increased risk of developing depressive episodes. The biological mechanisms underlying this association remain unclear, but alterations in the serotonergic brain system may play a role. The investigators have demonstrated that COC use reduces global cerebral serotonin 4 receptor (5-HT4R) levels in healthy women. The magnitude of this difference is comparable to what has been observed in individuals with depression relative to healthy controls.

This study aims to determine if COC discontinuation (COCd) results in recovery of the 5-HT4R brain levels and over what time frame this may occur. Further it investigates other potential neurobiological effects of COCd and how these map onto relevant signatures of mental states, including mood, memory function and sexual desire. We anticipate that this work will substantially advance the understanding of whether COC effects on serotonergic brain biology is reversible and whether such insights could provide novel preventive and therapeutic opportunities in depression.

- Study design: The investigators will conduct a randomized controlled trial assessing brain changes after COCd over a timeframe of up till one year. The investigators will include up to 60 healthy women (or until 25 women have completed follow-up in each arm) at 18-39 years of age who use a second-generation COC (containing 150 ug levonorgestrel and 30 ug ethinylestradiol), and have done this for a minimum of three months. The participants will undergo the same investigational program at baseline and at follow-up, which consists of PET (only COC discontinuers) and MRI/fMRI scans of the brain, oral glucose tolerance test, neuropsychological testing, collection of blood, saliva, and stool samples, completion of questionnaires regarding various trait- and state-related measures including a month at baseline and follow-up of daily questionnaires measuring psychometrics of mood/affect, sexual desire and sleep quality.

The timing of the follow-up assessment is determined by an adopted group sequential design; the first 15 COC discontinuers and 15 COC continuers will be distributed across week 8-24 after discontinuation/continuation. Hereafter, a planned assessment of the PET outcome will determine whether the last 10 COC discontinuers and 10 continuers will have their follow-up distributed within week 1-7 or week 8-52 after discontinuation (see more details under "Statistical analysis plan"). This procedure will allow an informed timing of the latter scans to better capture an estimate of the recovery time for 5-HT4R levels. Women allocated to COC continuation (COCc) will have their follow-up timepoint approximately matched to a COC discontinuer. Additionally, the timing of the follow-up will be during the active pill phase for COC continuers and during the follicular phase for COC discontinuers if menstrual cycle has returned. This will be planned based on pill cycle (COC continuers) and the reported first day of bleeding, menstrual cycle length, and LH tests during the follow-up (COC discontinuers). For the latter case, adjustment of the follow-up date may happen due to irregular cycles. Since the timing of each participant's follow-up assessment also needs to match personal calendar and availability of scan slots, we allow follow-up to deviate from the planned cycle days.

- Randomization: Participants will be randomized in 1:1 in random block sizes (e.g., 2, 4, 6 and 8) to either COCd or COCc. The randomization will be carried out by an administrative staff member at Neurobiology Research Unit (Rigshospitalet, Denmark), who will not otherwise be involved in the study or enrollment.

- Evaluation of compliance: Pill count will be carried out at every pill cycle. The participants has to report to the investigators if they forget to take a pill during the study. Non-compliance is defined as >2 missed pill in a cycle. If more than two pills are missed during the baseline or follow-up pill cycle, the participant will be asked if they are willing to postpone their assessment until the next pill cycle.

- Statistical analysis plan: The researchers use a group sequential design where they plan to conduct two sequential analyses including one interim analysis after 60% of the PET scans have been acquired (i.e., after 15 participants have been PET scanned at baseline and follow-up week 8-24) to decide whether the last 10 should be rescanned within week 1-7 or week 8-52. If the null hypothesis is rejected (i.e., they observe a change from baseline) in the interim analysis then the last 10 will be scanned within week 1-7 after COCd, but if it is not rejected (i.e., they observe no change from baseline), they will be scanned within week 8-52 after COCd. Only in the latter case will all scans be included in the primary aim test, i.e., to determine if there is an increase in 5-HT4R binding in caudate, putamen, and hippocampus from baseline to week 8 weeks or longer after COCd (in the former case, the null hypothesis would already be rejected with 15 participants). To account for the inflated type I error due to two planned tests, they will correct the alpha level (significance level) for each test by use of the alpha spending function (α*(t) = α ln(1 + (e - 1)t), which is similar to the Pocock correction, but the timing of the analysis does not need to be specified in advance, i.e., it is more flexible in the case of unforeseen problems with data acquisition. From the decision to distribute the last 10 scans early or late, they are able to optimize the timing of scans to enable us to determine over what timeframe the 5-HT4R level is returning to the age-corrected level observed in approximately 18 premenopausal women who had never used hormonal contraception (secondary hypothesis).

To test the primary hypothesis, they will conduct a one-sided test by use of a latent variable model on log-transformed 5-HT4R binding potentials in caudate, putamen, and hippocampus to estimate the change from baseline after COCd adjusted for change in injected tracer mass per kg body weight. Adjusted p-values, confidence intervals, and median unbiased estimates will be computed according to stagewise ordering.

To test the secondary hypothesis, they will first model two latent variable models, one determining the back-transformed change in the log of the global 5-HT4R binding between follow-up and baseline and the other determining the back-transformed difference in log of the global 5-HT4R binding between premenopausal women who never used hormonal contraception (i.e., another cohort) and the baseline (i.e., from the current cohort). Both latent variables will be estimated across caudate, putamen and hippocampus, which in the first model will be adjusted for change in injected tracer mass per kg bodyweight, and in the second model for injected tracer mass per kg bodyweight and age mean-centered around the mean age at follow-up (i.e., for the current cohort). Hereafter, they will evaluate the change in global 5-HT4R binding over time by using the back-transformed estimates of the change (denoted Y) as a function of time since discontinuation (denoted x) using the following logarithmic model: Y(x)=(Ymax*Y0)/((Ymax-Y0 )*e^((-k*x) )+Y0 ) , where only the rate constant k is fitted using the 'change in global binding' and non-linear least squares. The other two parameters Y0 and Ymax are set, respectively, to 1 and to the back-transformed mean global 5-HT4R difference from the second latent variable model. Based on the fitted model, we will estimate the time x such that the modeled mean change, Y(x), is equal to 90% of the mean global difference observed in the second latent variable model (Y0+0.9∙(Ymax-Y0)/Y0 ) to give an estimate of when the 5-HT4R binding has "recovered". If the model fit is poor, alternative models will be evaluated, including the logarithmic model without using a pre-defined Ymax (i.e., it will be estimated by the model) or a linear model with time on or not on the log-scale. If a linear model is used, time to recovery will be at the timepoint where the linear function is equal to Ymax.

For domain-specific hypotheses (see hypotheses below), we will compare change scores adjusted for baseline values (except for the memory encoding fmri domain where only follow-up data will be collected). If it turns out relevant to deviate from the original statistical plan, we will report this in detail in all publications of this study.

- Sample size calculation: Calculations based on intra-subject variability of the 5-HT4R binding from 40 premenopausal women with scan-rescan show that with two planned sequential analyses, including 60% and 100% of the planned scans, respectively, and with a power of 80%, an overall desired Type I error rate of 0.05 (one-sided), the alpha levels would be 0.035 and 0.028 for the interim and final analysis, respectively. With these, we are able to detect a difference of 8.2-9.5% (interim analysis) and 6.5-7.5% (final analysis) in binding potential in caudatus, putamen, and hippocampus for the interim and the final analysis, respectively. We will include up to 60 participants (or until at least 25 have completed baseline and follow-up in both arms) to allow a dropout rate of 20%. The primary analyses will be handled as complete case analysis. Missing data as well as dropouts will be reported in the relevant publications.

• Hypotheses:
• Primary hypothesis:

5-HT4R binding in caudate, putamen, and hippocampus increases from baseline to follow-up ≥ week 8 after COCd.

• Secondary hypotheses:

5-HT4R binding in caudate, putamen, and hippocampus has 'restored' within 12 months after COCd to the level seen in premenopausal women who never used hormonal contraception.

• Additional domain-specific hypotheses
• Memory encoding I. COCd compared to COCc improves verbal memory performance. II. COCd compared to COCc results in larger hippocampal brain activation during memory-encoding task.
• Insulin sensitivity III. COCd compared to COCc improves peripheral and brain insulin sensitivity. IV. COCd compared to COCC results in larger hypothalamic BOLD signal reduction after oral glucose load.
• Hedonic health V. COCd compared to COCc increases striatal reward response VI. COCd compared to COCc increases sexual desire
• Anxiety VII. COCd compared to COCc decreases anxiety levels.
• Hypothalamic-pituitary-adrenal axis dynamics VIII. COCd compared to COCc increases cortisol awakening response.
• Inflammatory system IX. COCd compared to COCc decreases extra-neurite mean diffusivity. XI: COCd compared to COCc increases, kynurenic acid, and kynurenic acid/quinolinic acid ratio and decreases pro-inflammatory markers.
• Brain structure XII. COCd compared to COCc results in increase in gray matter volumes of the (para)hippocampus and prefrontal cortex.
• Non-goal-oriented brain network organization XIII. COCd compared to COCc changes resting state functional connectivity brain networks.
• Gut microbiome XIV. COCd compared to COCc changes gut microbiome and gut microbiota-derived metabolites.
• Sleep:

XV: COCd compared to COCc improves sleep.

• Ethical considerations:

Exposure to radioactivity during the PET scans: The total exposure will not exceed 10 mSv equal to 3 years of natural background radiation in Denmark, which should be seen in the light of the new knowledge this study can generate.

Risk of unintended pregnancies in the COCd group. Those discontinuing COC are able to take other precautions and they will be thoroughly informed to use protection during intercourse. For this, we will offer condoms for free during the study period.

The study will be conducted in agreement with the Declaration of Helsinki.