Text mining with automation has emerged as a powerful tool for pharmaceutical industry helping them report adverse drug reactions (ADRs) to regulatory authority in a qualitative, efficient and cost effective manner, according to Kailash Chanduka, Director- Testing Shared Services at Aris Global Software Pvt Ltd.

Text mining is the process of retrieving relevant information from large amounts of both structured and unstructured text in electronic health record as well as unstructured text in social media, internet search history, biomedical literature and product information documents with the help of automated pattern learning and analyzing it with the help of machine learning and marking temporal relation between drug and event.

Regulatory authorities and pharma companies' interest in these secondary sources of adverse drug reactions are growing due to extensive under reporting of adverse drug reactions by medical professionals, inadequate clinical data etc.

Social media platforms are increasingly being used by patients and their caregivers to discuss health issues including adverse drug reactions. Hence these platforms are potentially an interesting source of health data. Global regulatory authorities such as US FDA, EMA and pharma companies are looking for social media platforms to get safety profile of drugs.

The Web-RADR project started in September 2014 to use social media for drug safety data and is funded by the Innovative Medicines Initiative, which in turn is funded by the European Commission and the European Federation of Pharmaceutical Industries and Associations. Drug regulators, pharmaceutical companies the World Health Organization and universities are part of the project.

European Medical Agency (EMA) along with drug makers will analyze the drug safety data from social media forums captured by the Web-RADR data mining tools to take a call on its use.

All data from social media is added to the EU’s pharmacoviligance ADR database, EudraVigilance, which is accessible to patients and researchers, to help signal detection.

Funded by the US FDA, Epidemico, a US based company conducted a study in 2014 which looked into 6.9 million Twitter posts and found 4,401 tweets resembling an ADR. The study found a high consistency between the data across organ classes while comparing with data held by the FDA. Epidemico uses its Med-WatcherSocial platform to detect ADRs from social media posts for around 1,400 drugs.

Indian Pharmacopoeia Commission (IPC) has also started mobile app in vernacular languages and helpline for patients to report ADRs. Companies like Novartis, Treato, IMS Health started looking into posts from social media to find potential drug safety signals.

Monitoring social media will not replace the traditional ADR reporting systems such as clinical trials, post-marketing spontaneous reporting and patient registries that can monitor safety, he said.

Besides social media, electronic health records are also used for collecting drug safety data.

The use of text mining tools and techniques to leverage secondary data resources will pave way for proactive holistic pharmacovigilance to meet the goal of identifying the adverse events as soon (speed) and accurately (precision) as possible leading to improved patient safety, said Chanduka.

However there are text mining challenges in secondary data source which can be categorised into data source challenges – acquisition, normalisation, cleansing, filtering of data; Operational challenges-- heterogenecity of medical text, colloquial terms of patients, vocabulary quality & completeness; Context of medical terms including negation; scalable and secure systems for electronic health record.

Despite the challenges, in recent years it has become clear that additional sources and data-mining techniques have been proven useful to obtain drug safety information which could contribute to decreasing the impact of ADRs on patient well-being, he concluded.