# Active monitoring vs. spontaneous reporting of antineoplastic drug–related adverse drug reactions: evidence from the Chinese hospital pharmacovigilance system

## TL;DR

This study compares active monitoring and spontaneous reporting for detecting cancer drug side effects in Chinese hospitals, showing active monitoring is more effective and faster.

## Contribution

The study demonstrates the advantages of active monitoring over spontaneous reporting in detecting and predicting adverse drug reactions in cancer therapy.

## Key findings

- Active monitoring detected more ADRs (160 vs. 50) and identified them earlier (4.2 vs. 10.7 days) than spontaneous reporting.
- Severe ADRs led to hospitalization (34.6%), treatment discontinuation (23.1%), and death (9.6%).
- Predictive models using active monitoring data showed strong performance (AUC = 0.84) compared to spontaneous reporting (AUC = 0.72).

## Abstract

Adverse drug reactions (ADRs) remain a major barrier to safe and effective cancer therapy. Existing pharmacovigilance systems predominantly rely on spontaneous reporting, which suffers from underreporting and delays. The Chinese Hospital Pharmacovigilance System (CHPS) provides an opportunity for active monitoring using multidimensional hospital data.

We conducted a retrospective cohort study, including 500 patients who received chemotherapy, targeted therapy, or immunotherapy. ADRs were identified through CHPS, classified by the Common Terminology Criteria for Adverse Events (CTCAE), and assessed using both active monitoring and spontaneous reporting. Signal detection employed disproportionality analyses (PRR, ROR, IC). Risk factors were analyzed with logistic regression, and predictive models for severe ADRs were evaluated with ROC curve analysis.

The overall ADR incidence was 37.0% (185/500), with 28.1% classified as severe. Hematologic (29.7%), gastrointestinal (26.0%), and skin/mucosal (19.5%) events were most common. Severe ADRs led to hospitalization (34.6%), treatment discontinuation (23.1%), and death (9.6%). Independent risk factors included age ≥65 years, polypharmacy, hepatic/renal dysfunction, and prolonged drug exposure (≥14 days). Signal detection confirmed known associations and identified potential novel signals, including skin hyperpigmentation with PD-1/PD-L1 inhibitors and cardiotoxicity with tyrosine kinase inhibitors. Active monitoring detected more ADRs than spontaneous reporting (160 vs. 50, P < 0.001) and provided earlier detection (mean 4.2 vs. 10.7 days). Predictive modeling demonstrated strong performance of the multivariable model (AUC = 0.82), with active monitoring outperforming spontaneous reporting (AUC = 0.84 vs. 0.72).

CHPS-based active monitoring improves the detection, timeliness, and predictive assessment of ADRs compared with spontaneous reporting. These findings support the integration of active monitoring into hospital pharmacovigilance systems and highlight novel safety signals requiring further validation.