JMIR Publications

ABSTRACT

Background:

There is limited data on the COVID-19 vector vaccine Gam-COVID-Vac (Sputnik V) safety profile. Previous infodemiology studies showed that social media discourse could be mined to assess the most concerning adverse events (AEs) of drugs.

Objective:

We aimed to investigate mild AEs of Sputnik V based on Telegram participatory trial in Russian language. We try to compare extracted AEs from Telegram with other comparable limited databases on Sputnik V and other COVID-19 vaccines. We explore symptom co-occurrence patterns and how count of administered doses, age, gender, and sequence of shots could confound reporting AEs.

Methods:

We collected a unique dataset, consisting of 11,515 text reports with self-reported Sputnik V vaccine AEs posted in Telegram Group using Telegram client, and utilized natural language processing methods for AE extraction. Specifically, we performed multi-label classification using deep neural language model BERT “DeepPavlov”, pre-trained on Russian language corpus and tuned on Telegram messages. Resulting AUC score was equal to 0.991. We have chosen symptom classes representing AE: fever, pain, chills, fatigue, nausea/vomiting, headache, insomnia, lymph nodes enlargement, erythema, pruritus, swelling, diarrhea.

Results:

By retrospective analysis, we found that female report more AEs than males (1.2 folds, P<.001), there are more AEs in first than second dose (1.13 folds, P<.001) and No. AEs decreases with age (β = .05 per year, P<.001). Sputnik V AEs are more similar to other vector(132 units) than mRNA vaccines (241 units) by average Euclidean distance between vectors of AE frequencies. Elderly Telegram users report significantly more (5.6 folds on average) systemic AEs than their peers according to III phase clinical trial results published in Lancet. However, reported in Telegram AEs are consistent (Pearson correlation r=.94, P=.02) with Argentinian post-registration AEs registry.

Conclusions:

Telegram users after Sputnik V vaccination complain mostly about Pain (47%), Fever (47%), Fatigue (34%) and Headache (25%) and the revealed adverse events profile of Sputnik V is comparable with other COVID-19 vaccines. Mining participatory trials of sentinel properties could provide meaningful information about pharmaceutics, especially if there is only limited information on AEs provided by producer with all limitation related to self-reporting biases (for instance local or gastric AEs could be under-reported as well as willingness to reporting satisfy typical product life-cycle temporal characteristics).