A Data Mining Approach to Comprehensive Signal Detection of Later Generation Contraceptive Drugs

Abstract

Contraceptive drugs are used by more than millions of women worldwide for the last many decades to prevent pregnancies and decrease menstrual pain, premenstrual syndrome, and also for the treatment of acne. The information available at the time of drug approval is incomplete, and drugs safety should be considered provisional. More information on safety can be obtained as the drug ages into the market and a broad category of people use that product. So, how the drug performs in real-world conditions is a test of its effectiveness. During post marketing surveillance studies, marketing authorization holder is obliged to report all the adverse events related to the drug promptly and this spontaneous report becomes main asset for early discovery of unknown ADR to drugs. Due to withdrawals of certain drug and number of cases of adverse drug reactions have reduced the confidence among people using certain drugs. The only way to instill confidence in patients is to promise safety and adherence to quality standards in available medicines. Signal Detection in Pharmacovigilance is one of the best ways to detect and prevent adverse events. A signal is a working hypothesis; one must assume that the suspected drug has caused the observed effect. One has to confirm or refute this hypothesis in signal assessment. The main aim of the thesis is to contribute to the existing knowledge regarding safety profile of contraceptive drugs using spontaneous reporting system. The objective of the study was to identify possible significant signal associated with drospirenone, dienogest, norgestimate, desogestrel, nomegestrol, etonogestrel, norelgestromin and gestodene by searching database of Federal Adverse Event Reporting System (FAERS), Eudravigilance (EV) and Canada Vigilance Adverse Reaction Online Database (CVARD). A total of 142823, 72982 and 2952 individual case safety reports (ICSRs) between the availability of first reports to Nov 2020 were downloaded from Federal Adverse Event Reporting System (FAERS), Eudravigilance and Canada Vigilance Adverse Reaction Online Database (CVARD) database respectively. These reports contained information of adverse events associated with drospirenone, dienogest, norgestimate, desogestrel, nomegestrol, etonogestrel, norelgestromin and gestodene. Disproportionality analysis (DPA), a data mining approach was applied for signal detection. Proportional reporting ratio (PRR), reporting odds ratio (ROR), PRR calculated by chi-square statistics, 95% confidence interval of PRR and ROR, and Bayesian confidence propagation neural network is the part of it. As per regulatory criteria, PRR ≥ 2, ROR ≥ 1, Chi-square statistics calculated PRR ≥ 4 and lower bound limit of 95% CI of PRR and ROR ≥ 1 to consider particular adverse drug reaction as a signal. Further by BCPNN method, if IC − 2SD ≤ 0 then that drug-ADR pair considered as no signal; if 0 < IC − 2SD ≤1.5, then that drug-ADR pair considered as weak signal; if 1.5 < IC−2SD ≤ 3.0, then that drug-ADR pair considered as middle signal; if IC − 2SD > 3.0, then that drug-ADR pair considered as strong signal. For breast cancer and norgestimate, the lower limit of 95 % CI of PRR is 16.40 and 23.04, PRR was 22.67 and 27.30, ROR is 22.87 (95%CI: 16.51, 31.67) and 27.69 (95% CI: 23.34, 32.85), Chi-square value is 723.34 and 3190.62, IC-2SD is 3.59 and 3.48 respectively for data obtained from EV and FAERS. So it is clear that there is a strong association between norgestimate and breast cancer as per EV and FAERS database. For urticaria and desogestrel, the lower limit of 95 % CI of PRR is 3.46 and 1.71, PRR was 5.05 and 2.34, ROR is 5.07 (95%CI: 3.47, 7.40) and 2.34 (95% CI: 1.71, 3.20), Chi-square value is 81.98 and 30.01, and IC-2SD is 1.80 and 0.86 respectively for data obtained from EV and FAERS. While for urticaria and dienogest, the lower limit of 95 % CI of PRR is 4.11 and 0.40, PRR was 5.96 and 0.97, ROR is 5.98 (95%CI: 4.12, 8.69), and 0.97 (95% CI: 0.40, 2.35), Chi-square value is 107.42 and 1.28 (p-value: .952418, not significant at p < .05), and IC-2SD is 2.02 and -0.93 respectively for data obtained from EV and FAERS. When compared to other contraceptive drugs of third and fourth generation for their association with urticaria, dienogest has higher ROR value i.e. 5.98 (gestodene: 2.53, nomegestrol: 3.59, desogestrel: 5.07) and higher IC-2SD value i.e. 2.03 (gestodene: 0.75, nomegestrol: 1.05, desogestrel: 1.80) as per EV database. The association between urticaria and dienogest for the results obtained from statistical analysis for the FAERS database is not significant. A signal cannot be concluded due to the negative value of IC-2SD. Also for asthma and etonogestrel, the lower limit of 95 % CI of PRR is 7.05 and 2.57, PRR was 8.95 and 3.15, ROR is 8.98 (95%CI: 7.08, 11.40) and 3.15 (95% CI: 2.57, 3.86), Chi-square value is 392.49 and 214.57, and IC-2SD is 1.58 and 0.60 respectively for data obtained from EV and FAERS. This indicates that there is a significant and a weak association between etonogestrel and asthma as per EV and FAERS database. For insomnia and nomegestrol, the lower limit of 95 % CI of PRR is 1.41, PRR was 2.56, ROR is 2.57 (95%CI: 1.41, 4.68), Chi-square value is 9.27, and IC-2SD is 0.69 for data obtained from EV database. There is a weak association between insomnia and nomegestrol as per the EV database, but the overall risk remains low. While for severe psychiatric disorders and desogestrel, the lower limit of 95 % CI of PRR is 3.77 and 0.59, PRR was 4.79 and 0.87, ROR is 4.83 (95%CI: 3.79, 6.16), and 0.87 (95% CI: 0.59, 1.28), Chi-square value is 190.34 and 1.10, and IC-2SD is 1.87 and -0.59 respectively for data obtained from EV and FAERS. There is significant but lower risk associated between desogestrel and severe psychiatric disorders as per EV database and no significant association for FAERS database. Gallbladder and drospirenone association was also studied. For this association, the lower limit of 95 % CI of PRR is 21.10 and 39.01, PRR was 23.13 and 42.20, ROR is 25.73 (95%CI: 23.45, 28.24) and 46.60 (95% CI: 43.06, 50.44), Chi-square value is 20973.10 and 20676.33, and IC-2SD is 0.65 and 1.29 respectively for data obtained from EV and FAERS. There is significant but lower risk associated between drospirenone and gallbladder diseases, whereas risk of occurring cholecystitis is more compared to other disorder of gallbladder diseases. Association between desogestrel and severe psychiatric disorders found in the Eudravigilance and FAERS database was also published in WHO international database, despite several confounders, suggests a potential harm for the desogestrel. Risk of having breast cancer is associated with the use of various contraceptive drugs (levonorgestrel, norgestimate, norethindrone, norgestrel). But for later generation contraceptive drug, norgestimate has the highest risk for having breast cancer and none observational study has verified this association, even though there were published case reports of contraceptive drug associated breast cancer in medical literature. Urticaria associated with desogestrel is a relevant safety concern as autoimmune progesterone dermatitis is a rare autoimmune response to exogenous progesterone whereas gallbladder disorder especially cholecystitis is associated with drospirenone. Asthma associated with etonogestrel and insomnia associated with nomegestrol lacks an evidence to be further considered for signal evaluation. But for this association, further studies should be designed. This information will assist the healthcare professionals in being careful about the prospect of encountering these adverse events when a contraceptive drug is being prescribed to women. Pharmacovigilance was the first method for post marketing surveillance and despite its inherent limitations such as lack of information or underreporting, it still contribute to the main objectives of post marketing surveillance to increase patient safety and to decrease the prescriber‘s euphoria.