Endocrine Abstracts

Aim: Obesity can be addressed through lifestyle interventions like modifying diet and engaging in physical activity. In cases where these measures prove ineffective, there is the option of administering medication (RxOb) or of surgical intervention. Sentiment analysis (SA) is an automated procedure employed to extract emotional information prevalent in a given body of text. Specifically, SA enables the determination of the "sentiment polarity" –whether the overall tone is positive or negative. Given the feasibility of implementing SA in medical texts (Artif Intell Med 2015; 64: 17-27), our objective was to employ SA to assess the medical literature’s stance, distinct from that of the lay press, concerning RxOb.

Methods: We collected (using MeSH terms) the English language abstracts of randomized clinical trials (RCTs) on RxOb in PubMed, from 2013 to 2023. For SA, we used two, state-of-the-art artificial intelligence SA tools, to generate averaged sentiment scores for each publication. SA scores were categorized into positive (maximum +1.00), neutral (0), or negative (maximum -1.00). Statistical analysis for differences between each year’s publications’ SA was done with analysis of variance (ANOVA); Pearson’s correlation was used to assess time trends.

Results: Seventy abstracts of RCTs were retained from an initial collection of 217 papers. The average± SD SA was positive at 0.59± 0.35. The lowest SAs were noted in 2017 and the highest in 2022 (P<0.05, ANOVA). A positive correlation between SA and time, specifically within the period 2017-2022, was found (r: = + 0.38, p: 0.008).

Discussion: RxOb is a dynamic and evolving field, with an expanding body of literature. SA of this literature offers a unique perspective by helping to uncover prevailing sentiments, trends, and perceptions. The application of SA to medical literature allows for the swift assessment of extensive published data collections, offering a rapid synthesis of overall findings that align with the outcomes of a conventional meta-analysis (Reg Anesth Pain Med 2022; 47: 151-154). Nevertheless, our findings are based on SA of existing literature and may not capture all relevant factors influencing sentiments. Future research could delve deeper into the underlying causes of this positive trend, such as specific breakthroughs or changing perceptions among healthcare professionals.