Exploration of CPRD and Yellow Card real-world data to support efficacy and safety targets for levonorgestrel contraceptives.
Andrew Butler, Preeti Datta-Nemdharry, Sophie Scanlon, Emily Cannell, Mary Malamatari, Katherine Donegan, Susan Cole
Medicines and Healthcare products Regulatory Agency
Introduction: Failure of contraception is uncommon but inconsistent use is a significant cause of unintended pregnancies. Short-acting contraceptives rely on user compliance and have higher failure rates during typical use (~9%) than during optimal use (~0.3%) [1]. Effectiveness of long-acting reversible contraceptives (LARC) does not depend on daily concordance. To improve compliance and outcomes, alternative products with varied durations of action are desired. Products developed with this aim often use active pharmaceutical ingredients (APIs) which are marketed under different forms. Bioequivalence for modified release products depends on comparative clinical efficacy and safety data plus pharmacokinetic (PK) data [2]. However, if the relationship between the API plasma concentration and clinical response is well-defined, clinical trials may be unnecessary. Thus, if the optimum exposure bracket is well-quantified, one can rely on PK studies only to understand clinical performance of products [3, 4]. Regarding PK properties, levonorgestrel (LNG) is one of the best characterized hormonal contraceptives.
Objectives: Review real-world data (RWD), in the form of the Yellow Card and Clinical Practice Research Datalink (CPRD) databases in the UK, to investigate their utility for defining the exposure bracket for LNG to aid design of new LARCs.
Methods: The Yellow Card database (a voluntary post-market adverse drug reaction [ADR] registry [5]) was queried for UK ADR reports containing LNG as a suspect drug. Data was categorised according to likeliness of ADR occurrence due to insufficient or excessive exposure to LNG. The CPRD database (a longitudinal anonymised database derived from UK primary care records [6, 7]) was used to ascertain records for women using oral LNG or an LNG implant (Norplant). Descriptive analysis of CPRD data was conducted to assess the suitability of this RWD to determine effectiveness (unintended pregnancies) and safety (adverse events; AEs) of LNG.
In order to strengthen the exposure-response understanding, the potential to use the RWD in conjunction with population PK and PBPK models (developed at the University of Liverpool) was assessed.[PD1]
Results: Several limitations associated with using these data sources for quantitative exposure-response analyses were identified. There were difficulties in determining the exposure at the time of AE/ADRs and in using predetermined clinical end points due to limitations around the collection and reporting of the data. Significant irregularities were noted throughout the data, particularly with regards to implants. These included missingness of treatment start and end dates, simultaneous prescription of multiple contraceptives[DNP2] , overlap of pregnancy dates and records of Norplant following its removal from the UK market. However, it was possible to make some observations from the data.[PD3]
The Yellow Card data revealed increased incidence of ADRs in the first few months of dosage, with this declining from ~550 reports in the first month to <20 reports after 12 months. The majority (~88%) of ADRs were those defined as likely to occur due to a high dose of LNG. When ADRs associated with a low LNG dose (pregnancies/abortions) were filtered by route of administration, the oral contraception exhibited a higher number of reports compared to LARC routes. A Pearl index (pregnancies per 100 patient years) of 0.005 was estimated for years 1-2 following implantation of an LNG-based device.
The CPRD data revealed 135,613 treatment episodes of oral LNG in 84,468 women (median treatment length = 105 days). There were 543 [PD4] unintended pregnancies per 10,000 women. 296 women with a Norplant insertion date were identified with 1,824 [PD5] pregnancies per 10,000 women. The non-availability of implant insertion/removal dates reduced the number of women included in the Norplant cohort. Pearl index appeared similar for oral (5.4) and implanted (6.4) LNG. [PD6] AEs decreased with increased time post-implantation. In the oral cohort, menstrual disorders (4.0%) and skin disorders (3.7%) were the most common AEs. Abdominal pain (12.5%) and menstrual disorders (10.8%) were most commonly reported in those with implants, up to 5 years after treatment start.
Conclusions: This study highlights limitations in the use of CPRD and Yellow Card databases to correlate safety or efficacy data with the pharmacokinetics of LNG.[PD7]
The authors are grateful to the Bill and Melinda Gates Foundation for funding this research.
References:
[1] Trussell, J., Contraceptive failure in the United States. Contraception, 2011. 83(5): p. 397-404.
[2] European Medicines Agency, Guideline on the pharmacokinetic and clinical evaluation of modified release dosage forms, 2014.
[3] Food and Drug Administration, Guidance for Industry: Providing Clinical Evidence of Effectiveness for Human Drug and Biological Products, 1998.
[4] Li, L., et al., Use of Model-Informed Drug Development to Streamline Development of Long-Acting Products: Can These Successes Be Translated to Long-Acting Hormonal Contraceptives? Annual Review of Pharmacology and Toxicology, 2021. 61(1): p. 745-756.
[5] Medicines and Healthcare products Regulatory Agency, The Yellow Card scheme: guidance for healthcare professionals, patients and the public. 2021; Available from: https://www.gov.uk/guidance/the-yellow-card-scheme-guidance-for-healthcare-professionals
[6] Medicines and Healthcare products Regulatory Agency, Clinical Practice Research Datalink. 2024; Available from: https://cprd.com/
[7] Ghosh, R.E., et al., How Clinical Practice Research Datalink data are used to support pharmacovigilance. Ther Adv Drug Saf, 2019. 10: p. 2042098619854010.