[Retracted] Social Media Analytics for Pharmacovigilance of Antiepileptic Drugs

Table 1

Summary of previous research on utilizing social media for ADRs detection.

Reference	Data source	Target drug set	No. of drugs	ADR extraction approach	ADR signaling method
[44]	DailyStrength	NA	6	Lexicon-based	Association rule mining
[45]	Various OHC forums	Breast cancer	4	Lexicon-based	Association rule mining
[46]	Various parenting forums	Pediatric drugs (fever, pain, influenza, viruses)	9	Lexicon-based	Disproportionality analysis
[29]	Twitter	Cancer	5	ML-based (SVM)	NA
[47]	American Diabetes Association	Diabetes	NA	Lexicon-based	Shortest dependency path-based ML algorithm
[28]	Askapatient, Drugs.com, DrugRatingZ	Breast cancer	5	Pattern-based	NA
[48]	Twitter	NA	23	Lexicon-based	Aggregated frequencies
[33]	MedHelp	Heart disease	NA	Lexicon-based	Rule-based approach for relation classification
[49]	DailyStrength	NA	38	ML-based (SVM)	Probabilities of all comments associated with each drug combined to predict if drug should be categorized as normal or blackbox
[50]	Twitter, DailyStrength	NA	81	Supervised learning via conditional random fields (CRF)	NA
[24]	Five popular and open French forums	Methylphenidate	4	Lexicon-based	Disproportionality analysis
[27]	Askapatient.com; Medications.com; WebMD.com	Cholesterol-lowering drugs		Lexicon-based	Log-likelihood ratio
[51]	Twitter	Attention deficit hyperactivity. Disorder drugs	44	ML-based (RNN)	NA
[14]	PatientsLikeMe; MedHelp	NA	20	Lexicon-based	Association rule mining; Disproportionality analysis
[52]	Five forums in France	Oral antineoplastic drugs	8 (ATC subgroup)	Lexicon-based manual	Frequency
[30]	Askapatient.com	Diabetes drug Glucophage/metformin	1	ML-based	NA
[31]	Askapatient.com	Psychiatric drugs	4	ML-based	NA
[32]	Twitter	Malaria drugs	19	ML-based and rule-based (cTake)	Disproportionality analysis
[34]	Twitter and PubMed	Opioid drugs	3	ML-based (convolutional recurrent neural network (CRNN))	NA