Does social media complement traditional data sources in determining the comparative effectiveness and safety of medications for inflammatory arthritis?
A recent study by Jeffrey R. Curtis, MD, MS, MPH, Professor of Medicine, Division of Clinical Immunology and Rheumatology, University of Alabama at Birmingham, and colleagues attempted to answer that question by sifting through all publicly available social media data (eg, Facebook, Twitter, blogs, and discussion boards) for posts mentioning arthritis. Over 2.5 billion posts were analyzed.
The limited information available about the use of social media specific to arthritis led Dr Curtis and his team to set 3 main objectives for their study, which were to demographically characterize people using social media to discuss arthritis, evaluate how valuable social media data were in researching drug safety, and note the content and timing of social media posts with regard to the launch dates of newer biologic drugs.
“It is possible that patients could be using social media to discuss and even report novel benefits or risks of newly licensed medications. This type of use would be particularly salient shortly after a product becomes available in the marketplace where real-world data are typically scant,” said Dr Curtis and colleagues.
The team used the Treato software platform to collect openly available, user-generated content on health topics from >10,000 sources. Natural language processing algorithms were used to analyze this content to identify mentioned medical concepts and extract patients’ self-reported descriptions of their health conditions and drugs. Treato also processes and resolves synonyms of medical terms (eg, myocardial infarction and heart attack). The specific medical concepts targeted in Dr Curtis and colleagues’ study were inflammatory arthritis (ie, rheumatoid arthritis, psoriatic arthritis, and psoriasis), as well as 4 medications used for rheumatoid arthritis (ie, tofacitinib, tocilizumab, abatacept, and etanercept).
As of October 2015, there were 785,656 arthritis-related posts, primarily from patients (87%) aged <40 years (61%) based in the United States (75%). The age distribution of post contributors was <19 years (18%), 19 years to 29 years (23%), 30 years to 39 years (20%), 40 years to 49 years (17%), 50 years to 64 years (17%), and ≥65 years (6%). Facebook was the most common platform for social media discussions.
Based on data accessed in May 2016 from Symplur, a platform that maintains a controlled database of vetted healthcare hashtags grouped with other relevant hashtags, the most common sources for mentions of rheumatoid arthritis on Twitter included a mix of professional patient advocacy organizations and individual patients.
To determine whether direct-to-consumer advertisements impacted social media posts, Dr Curtis and colleagues obtained direct-to-consumer launch dates of new US television advertisements for etanercept, tofacitinib, and abatacept from their manufacturers. There were 6 product-specific direct-to-consumer advertising campaigns for these 3 medications in 2014 and 2015. Using the paired t-test to evaluate the hypothesis that there was an increased volume of social media posts following each new direct-to-consumer advertising launch date, the researchers compared tweet volume in the 2 weeks before and 2 weeks following the direct-to-consumer ad launch.
There was no significant relationship between direct-to-consumer ad launch timings and the volume of social media posts about etanercept, tofacitinib, and abatacept. However, after data analysis was restricted to posts specifically mentioning direct-to-consumer advertisements (n = 481), such tweets were approximately 3 times greater in the 2 weeks following the direct-to-consumer launch than in the 2 weeks before it (P = .04).
Among the 481 posts related to direct-to-consumer advertisements, 12.3% expressed concerns about drug safety. The most frequent safety concerns included cancer, death, fatal infections, and gastrointestinal perforation. Posts expressing medication safety concerns were significantly more frequent than those that described favorable aspects of the medication (<5%) or expressed hope that the medication may provide an effective treatment option.
The study results confirmed known associations between herpes zoster and tofacitinib, and between gastrointestinal perforation and tocilizumab. “We also found social media content related to hair regrowth and skin re-pigmentation, a novel observation seen with patients receiving tofacitinib,” said Dr Curtis and colleagues.
By using social media data to identify drug-associated medical events, the investigators determined that the positive predictive value of an automated natural language processing classifier to find true cases of herpes zoster reported by patients was high (91%). When manual review is not feasible for a large-scale project, the system may be used in an automated fashion to detect early safety signals, they suggested.
“However, if time and resources permit, manual review of social media posts is helpful to identify more true cases and improve the specificity of classifying safety events,” Dr Curtis and colleagues stated.
Although the demographics of people using social media are still evolving, results of the study showed that 60% of patients were aged ˂40 years. In addition, mining demographic information depends on people being willing to use relaxed privacy settings, and to share medical information. Regardless, the investigators assert that social media is a promising data source that can complement the traditional methods of studying newly approved medications, particularly when paired with the filtering capabilities of the Treato software.
“The strengths of our study include a novel look at a global information source that has been infrequently studied in healthcare, using a robust platform that extracted medical terms, resolved synonyms, and enabled quantitative analysis,” concluded Dr Curtis and colleagues.