# Assessing safety trends of withdrawn medications: A data-driven pharmacovigilance approach using growth models

## TL;DR

This study uses data-driven methods to analyze how adverse drug reactions (ADRs) change over time for withdrawn medications and cancer drugs, introducing a new metric called the Detriment Index to compare drug safety.

## Contribution

The novel Detriment Index metric enables comparative safety ranking of drugs based on ADR accumulation trends, regardless of reporting patterns.

## Key findings

- Withdrawn medications showed diverse ADR reporting patterns, including saturation, linear, exponential, and sigmoidal trends.
- The Detriment Index integrates ADR growth rate and cumulative burden, allowing for comparative safety assessments across drugs.
- Tamoxifen demonstrated a favorable safety profile compared to Pembrolizumab, which showed a higher risk of ADR accumulation.

## Abstract

Adverse drug reactions (ADRs) are critical in evaluating a medicine's safety profile during development and post-marketing surveillance. This study focuses on medications linked to major adverse drug reactions (ADRs) that were later taken off the market or removed from usage. Adverse drug reaction (ADR) reports were gathered from VigiAccess and the U.S. FDA (United States Food and Drug Administration) Adverse Event Reporting System (FAERS). Data visualization tools were then used to evaluate how reporting patterns changed over time. Although a decrease in adverse drug reaction (ADR) reports after a withdrawal may be expected, we observed various patterns over time. These included sigmoidal, exponential, and linear trends, which indicated that safety signals persisted differently. Assessment was performed using cumulative annual ADR reporting trends, supplemented by published evidence on typical latency intervals between drug exposure and event onset for the medicines of interest. Additionally, we developed a novel statistical metric termed the Detriment Index, based on curve-fitting and growth-rate modeling, to quantitatively compare the relative safety of drugs. This approach allows ranking of medicines with respect to ADR accumulation, supports identification of safer therapeutic alternatives, and provides practical decision support for clinicians and regulators.

A total of 39 withdrawn medications were included based on regulatory withdrawal records documented by WHO and FDA sources. For comparative safety assessment, 15 commonly used cancer medications (including Tamoxifen, Avastin, Bleomycin, Paclitaxel, Vincristine, Methotrexate, Cisplatin, Doxorubicin, Imatinib, Docetaxel, Rituximab, Trastuzumab, Revlimid, Lenalidomide, and Pembrolizumab) were analyzed. Four distinct growth patterns were identified based on model fit: Saturation (n = 17), Linear (n = 8), Exponential (n = 9), and Sigmoidal (n = 5), with overall model performance demonstrating strong goodness-of-fit (R2 = 0.83–0.97). To enable comparative interpretation of safety profiles, we introduce composite, model-based metric derived from the fitted cumulative ADR curve called ‘Detriment Index,’ which integrates the ADR growth-rate parameter (β) with the cumulative burden (area-under-curve) of ADR reports, providing a normalized metric that captures both the speed and magnitude of safety-signal accumulation. This allows comparative safety ranking across drugs regardless of differences in reporting patterns or curve shapes. Detriment Index integrates both the speed of ADR accumulation and the total accumulated harm, enabling comparative safety assessment across drugs.

To understand the context of these withdrew drug trends compared to currently used pharmaceuticals, we investigated group of active oncology drugs. In contrast, the occurrence of adverse drug reactions (ADRs) with tamoxifen increased more slowly, but pembrolizumab showed a quicker and more consistent rise in these events. These differences show that combining trend categorization with the Detriment Index improves our understanding of safety and boosts post-marketing surveillance systems. The findings show that reporting adverse drug reactions (ADRs) continues, even after a medicine is removed from the market, also show how using quantitative methods can improve safety assessments and support clinical decisions.

For analysis publicly available datasets from VigiAccess, the WHO's global database of reported potential side effects of medicinal products, and the FDA's Adverse Event Reporting System (FAERS) Public Dashboard have been used. A comprehensive analysis was performed on cumulative count of ADR reports for selected medications that have been withdrawn from the market. The investigation focused on finding patterns in the number of adverse drug reaction (ADR) reports related to the termination of these drugs. To visualize these trends, different curve-fitting methods have been used. These included both linear and non-linear statistical models, such as sigmoidal, exponential, and linear forms. A ranking was also made using an exponential growth rate model to compare the safety ratings of different cancer drugs. This model was used to look at and compare the drug's safety features.

When a medication is withdrawn or banned, its utilization is anticipated to cease, resulting in the absence of further reports of adverse drug reactions (ADRs). This is the predicted result for all pharmaceuticals that have been banned. The analysis of 39 medications indicated diverse linear and nonlinear patterns. Specifically, 17 drugs followed a saturation pattern, 10 showed a linear pattern, 7 exhibited an exponential pattern, and 5 displayed a sigmoidal pattern. The examples presented in this investigation demonstrated that the drugs Benoxaprofen, Rosiglitazone, Temazepam, and Rofecoxib exhibited a robust correspondence with diverse modeling approaches. Specifically, Benoxaprofen conformed to a saturating hyperbola model, achieving a R2 value of 0.98. Rosiglitazone, conversely, was best represented by a linear model, which yielded a R2 value of 0.96. Furthermore, Temazepam displayed an exceptional fit with the exponential model, resulting in a R2 value of 0.99. The data for rofecoxib showed a sigmoidal pattern, with a R2 value of 0.92, indicating a strong fit to the sigmoidal model. In a safety comparison of fifteen oncological agents, Tamoxifen demonstrates a more favorable safety profile, attributable to its reduced rate of ADR accumulation and a lower growth rate of 0.0972. Conversely, Pembrolizumab exhibits a higher exponential growth rate of 0.8277, which suggests an increased risk profile.

This study shows that adverse drug reaction (ADR) patterns, particularly those for banded/withdrawn drugs and some cancer therapies, demonstrate how safety issues alter over time. Adverse drug response (ADR) reports normally diminish after a drug is discontinued, however the rate varies. This pattern shows historical risk estimates, not those medications are safer.

Analyzing the overall patterns in adverse drug reactions (ADRs) provides an additional viewpoint to established ways of monitoring drug safety. This approach allows for a more nuanced understanding of long-term safety data. The Detriment Index, a new metric based on trends, allows for the comparison of the long-term burden of adverse drug reactions (ADRs) across different medications, regardless of their regulatory history or how they are used in treatment. This method helped separate medications with lower risks, like Tamoxifen, from others that build up in the body more quickly, such as Pembrolizumab. This shows its potential for comparing safety profiles.

This exploratory approach, in essence, enriches our understanding of how long post-marketing safety information lasts. It also provides a foundation for future research. This research could help support safer clinical decisions and regulatory evaluations for both pharmaceuticals that are no longer on the market and those that are still available.