Skip to main content

30-06-2011 | Article

Network analysis of VAERS improves vaccine, AE pattern visualization


Free abstract

MedWire News: Study findings suggest network analysis of Vaccine Adverse Event Reporting System (VAERS) helps aid the visualization of patterns among vaccines and adverse events (AEs), which may provide a new way to identify vaccines and AEs of most clinical interest.

Safety monitoring of medicinal products is currently performed using a combination of expert reviews of reported cases and statistical data mining algorithms (DMAs), which are mostly based on the disproportionality of reporting - a method with a variety of limitations, namely that single drug-AE combinations should not be considered in isolation.

"Current DMAs do not identify all known associations in VAERS, in part because not all known associations are reported disproportionally, and there is no consistency among methods in which associations are signalled," say Taxiarchis Botsis and R Ball, both from the US Food and Drug Administration in Rockville, Maryland, USA.

The team says that network analysis can solve this problem, as it allows for simultaneous representation of complex connections among the key elements of a system, which are represented as edges or links in the network, similar to the way the World Wide Web is constructed.

"If vaccines and AE coding terms are considered nodes in a network, AEs that are known to be caused by several different vaccines, or vaccines with many associated AEs, may behave like hubs in the network. Metrics that describe the hubs might be used to identify these AEs and vaccines in a way that is not possible using disproportionality DMAs," explain the researchers.

Botsis and Ball therefore applied network analysis to the VAERS database of spontaneous reports (of 6354 AEs for 74 vaccines) to evaluate whether it could detect known patterns among co-administered vaccines and co-occurring AEs. Using VAERS subset data from 2005 to 2007, stratification with network analysis was used to analyze syncope among patients aged 2 to <11 years, 11 to 18 years, and 19 to <65 years. The team then evaluated the ability of network analysis to identify the known safety signal of intussusception after Rotashield vaccine using VAERS reports from 1998 to 1999.

Analysis of the partitions of calendar time, age (syncope in adolescents), and vaccine (rotavirus), showed that VAERS had characteristics of a "scale-free" network, with the top 90% of the nodes obeying the power-law distribution and the hubs being highly connected to other nodes in the network.

"This finding is important because it provides a new way to describe VAERS data and, at a macroscopic level, demonstrates that certain vaccines and AEs act as hubs," say the researchers.

Indeed, the team says that potential advantage of the network analysis framework is its ability to identify nodes of importance, irrespective of whether they are disproportionally reported.

"Further evaluation of this approach as a complement to statistical DMAs in prospective application to VAERS and other safety databases is recommended," conclude the authors in the journal Clinical Pharmacology and Therapeutics.

MedWire ( is an independent clinical news service provided by Springer Healthcare Limited. © Springer Healthcare Ltd; 2011

By Ingrid Grasmo