Text Mining of Adverse Events in Clinical Trials: Deep Learning Approach

Background Pharmacovigilance and safety reporting, which involve processes for monitoring the use of medicines in clinical trials, play a critical role in the identification of previously unrecognized adverse events or changes in the patterns of adverse events. Objective This study aims to demonstrate the feasibility of automating the coding of adverse events described in the narrative section of the serious adverse event report forms to enable statistical analysis of the aforementioned patterns. Methods We used the Unified Medical Language System (UMLS) as the coding scheme, which integrates 217 source vocabularies, thus enabling coding against other relevant terminologies such as the International Classification of Diseases–10th Revision, Medical Dictionary for Regulatory Activities, and Systematized Nomenclature of Medicine). We used MetaMap, a highly configurable dictionary lookup software, to identify the mentions of the UMLS concepts. We trained a binary classifier using Bidirectional Encoder Representations from Transformers (BERT), a transformer-based language model that captures contextual relationships, to differentiate between mentions of the UMLS concepts that represented adverse events and those that did not. Results The model achieved a high F1 score of 0.8080, despite the class imbalance. This is 10.15 percent points lower than human-like performance but also 17.45 percent points higher than that of the baseline approach. Conclusions These results confirmed that automated coding of adverse events described in the narrative section of serious adverse event reports is feasible. Once coded, adverse events can be statistically analyzed so that any correlations with the trialed medicines can be estimated in a timely fashion.

Safety of SGLT2 Inhibitors: A Pharmacovigilance Study from 2013 to 2021 Based on FAERS

Background: Sodium-glucose co-transporter-2 inhibitors (SGLT2is) are widely used in clinical practice for their demonstrated cardiorenal benefits, but multiple adverse events (AEs) have been reported. We aimed to describe the distribution of SGLT2i-related AEs in different systems and identify important medical event (IME) signals for SGLT2i.Methods: Data from the first quarter (Q1) of 2013–2021 Q2 in FAERS were selected to conduct disproportionality analysis. The definition of AEs and IMEs relied on the system organ classes (SOCs) and preferred terms (PTs) by the Medical Dictionary for Regulatory Activities (MedDRA-version 24.0). Two signal indicators, the reported odds ratio (ROR) and information component (IC), were used to estimate the association between SGLT2is and IMEs.Results: A total of 57,818 records related to SGLT2i, with 22,537 SGLT2i-IME pairs. Most SGLT2i-related IMEs occurred in monotherapy (N = 21,408, 94.99%). Significant signals emerged at the following SOCs: “metabolism and nutrition disorders” (N = 9,103; IC025 = 4.26), “renal and urinary disorders” (3886; 1.20), “infections and infestations” (3457; 0.85). The common strong signals were observed in diabetic ketoacidosis, ketoacidosis, euglycaemic diabetic ketoacidosis and Fournier’s gangrene. Unexpected safety signals such as cellulitis, osteomyelitis, cerebral infarction and nephrolithiasis were detected.Conclusion: Our pharmacovigilance analysis showed that a high frequency was reported for IMEs triggered by SGLT2i monotherapy. Different SGLT2is caused different types and the association strengths of IMEs, while they also shared some specific PTs. Most of the results are generally consistent with previous studies, and more pharmacoepidemiological studies are needed to validate for unexpected AEs. Based on risk-benefit considerations, clinicians should be well informed about important medical events that may be aggravated by SGLT2is.

Identifying health-related discussions of cannabis use on Twitter: a proof-of-concept study (Preprint)

BACKGROUND The cannabis product and regulatory landscape is changing in the United States. Against the backdrop of these changes, there have been increasing reports on health-related motives for cannabis use and of adverse events from its use. The use of social media data in monitoring cannabis-related health conversations may be useful to state and federal-level regulatory agencies as they grapple with identifying cannabis safety signals in a comprehensive and scalable fashion. OBJECTIVE This study attempted to determine the extent to which a medical dictionary, the Unified Medical Language System (UMLS) Consumer Health Vocabulary (CHV), could identify cannabis-related motivations of use and health consequences of its use as discussed on Twitter in 2020. METHODS Twitter posts containing cannabis-related terms were obtained from January 1 to August 31, 2020. Each post from the sample (n = 353,353) was classified into at least one of 17 a priori categories of commonly health-related topics, using a rule-based classifier with each category defined by the terms in the medical dictionary. A subsample of posts (n=1094) was then manually annotated to help validate the rule-based classifier and determine if each post pertained to health-related motivations for cannabis use or perceived adverse health effects from its use or neither. RESULTS The validation process suggested that the medical dictionary could identify health-related conversations in 31.2% of posts. Specifically, 20.4% of posts were accurately identified as relating to a health-related motivation for cannabis use, while 10.8% of posts were accurately identified as relating to a health-related consequence from cannabis use. Potential health-related conversations around cannabis use ranged from issues with the respiratory system and stress to the immune system and gastrointestinal problems, among other health topics. CONCLUSIONS The mining of social media data may prove helpful in improving surveillance of cannabis products and their adverse health effects. However, future research needs to develop and validate a dictionary and codebook that captures cannabis use-specific health conversations on Twitter.

Pharmacovigilance study of anti-infective-related acute kidney injury using the Japanese adverse drug event report database

Background Acute kidney injury (AKI) is associated with significant increases in short- and long-term morbidity and mortality. Drug-induced AKI is a major concern in the present healthcare system. Our spontaneous reporting system (SRS) analysis assessed links between AKI, along with patients’ age, as healthcare-associated risks and administered anti-infectives. We also generated anti-infective-related AKI-onset profiles. Method We calculated reporting odds ratios (RORs) for reports of anti-infective-related AKI (per Medical Dictionary for Regulatory Activities) in the Japanese Adverse Drug Event Report database and evaluated the effect of anti-infective combination therapy. The background factors of cases with anti-infective monotherapy and combination therapy (≥ 2 anti-infectives) were matched using propensity score. We evaluated time-to-onset data and hazard types using the Weibull parameter. Results Among 534,688 reports (submission period: April 2004–June 2018), there were 21,727 AKI events. The reported number of AKI associated with glycopeptide antibacterials, fluoroquinolones, third-generation cephalosporins, triazole derivatives, and carbapenems were 596, 494, 341, 315, and 313, respectively. Crude RORs of anti-infective-related AKI increased among older patients and were higher in anti-infective combination therapies [anti-infectives, ≥ 2; ROR, 1.94 (1.80–2.09)] than in monotherapies [ROR, 1.29 (1.22–1.36)]. After propensity score matching, the adjusted RORs of anti-infective monotherapy and combination therapy (≥ 2 anti-infectives) were 0.67 (0.58–0.77) and 1.49 (1.29–1.71), respectively. Moreover, 48.1% of AKI occurred within 5 days (median, 5.0 days) of anti-infective therapy initiation. Conclusion RORs derived from our new SRS analysis indicate potential AKI risks and number of administered anti-infectives.

A novel corpus developed to evaluate the impact of hospital noise on speech intelligibility

Hospital noise often exceeds recommended sound levels set by health organizations leading to reductions in speech intelligibility and communication breakdowns between doctors and patients. However, quantifying the impact of hospital noise on intelligibility has been limited by stimuli employed in prior studies, which did not include medically related terminology. To address this gap, a corpus of medically related sentences was developed. Word frequency, word familiarity, and sentence predictability, factors known to impact intelligibility of speech, were quantified. Nearly 700 words were selected from the Merriam-Webster Medical Dictionary. Word frequency was taken from Lexique, a 51-million-word corpus of American subtitles (Brysbaert & New, 2009). Word familiarity was rated by 41 monolingual listeners. The words were then used to construct 200 sentences. To determine sentence predictability, the sentences were presented to 48 participants with one word missing; their task was to fill in the missing word. Three 40 item sentence sets with different familiarity / frequency types (low/low, high/low, high/high) were selected, all with low predictability levels. These sentences and 40 standard speech perception sentences were recorded by two male and two female talkers. This corpus can be used to assess how hospital noise impacts intelligibility across listener populations. Brysbaert, M., & New, B. (2009). Moving beyond Kučera and Francis: A critical evaluation of current word frequency norms and the introduction of a new and improved word frequency measure for American English. Behavior Research Methods, 41, 977-990. doi:10.3758/BRM.41.4.977.

Text Mining of Adverse Events in Clinical Trials: Deep Learning Approach (Preprint)

BACKGROUND Pharmacovigilance and safety reporting, which involve processes for monitoring the use of medicines in clinical trials, play a critical role in the identification of previously unrecognized adverse events or changes in the patterns of adverse events. OBJECTIVE This study aims to demonstrate the feasibility of automating the coding of adverse events described in the narrative section of the serious adverse event report forms to enable statistical analysis of the aforementioned patterns. METHODS We used the Unified Medical Language System (UMLS) as the coding scheme, which integrates 217 source vocabularies, thus enabling coding against other relevant terminologies such as the International Classification of Diseases–10th Revision, Medical Dictionary for Regulatory Activities, and Systematized Nomenclature of Medicine). We used MetaMap, a highly configurable dictionary lookup software, to identify the mentions of the UMLS concepts. We trained a binary classifier using Bidirectional Encoder Representations from Transformers (BERT), a transformer-based language model that captures contextual relationships, to differentiate between mentions of the UMLS concepts that represented adverse events and those that did not. RESULTS The model achieved a high F1 score of 0.8080, despite the class imbalance. This is 10.15 percent points lower than human-like performance but also 17.45 percent points higher than that of the baseline approach. CONCLUSIONS These results confirmed that automated coding of adverse events described in the narrative section of serious adverse event reports is feasible. Once coded, adverse events can be statistically analyzed so that any correlations with the trialed medicines can be estimated in a timely fashion.

THE TERMS-VERBS IN CONTEMPORARY MEDICAL TEXT

The article presents a fact-based study of the verbalization of medical knowledge, verbal nomination as one of the ways to create a Russian medical dictionary. The linguistic materials collected during the research indicate the ability of the verb to terminate concepts. Verb-terms, in contrast to noun-terms, nominate specific processes, phenomena. Verb terms are included in word-formation nests along with noun terms. Verb terms fall into two groups: 1) branch verbs and 2) common verbs. The first group unites verbs characteristic of the medical field of knowledge, the second group includes verbs, the terminological nature of which is manifested in the composition of a phrase with a dependent noun-term. In such verb-nominal phrases, the verb either expands the meaning, or concretizes the existing one. Verb terms are used mainly in those branches of medicine that are associated with a specif- ic action (for example, surgery). Verb terms have the same grammatical categories as verbs of the general literary language. The results obtained can be used for further research on the cognitive properties of verbs-terms based on new sources.

Medicīnas terminu vārdnīca: mobilās lietotnes izstrādes metodoloģija un izmantotie digitālie resursi

The research provides a snapshot of the development of a methodology for the „English-Latvian-English glossary of medical terms” (bilingual dictionary and phrase book) that was implemented as a cooperation project between two higher education institutions. The study describes the resources used in the drafting process and the structure and functionality of the dictionary. The research was conducted at the end of 2019 as a project Mobile application English-Latvian-English Bilingual Dictionary and Phrase Book of Medical Terms within the framework of internal research project competition Development of Scientific Research Activity at Ventspils University of Applied Sciences. The research concept originates in the up to now insufficiently deployed characteristic of the terminology of various fields (including the medical domain) in different respects: the selection and grouping of terms, their categorisation and translation. Given the experience of the authors of the available and so far published medical dictionaries and the results of empirical work and the survey of potential users about their expectations for a medical dictionary, possible problems of and solutions for the development of a new dictionary of medical terms were deployed.

To the 100th anniversary of the Russian medical dictionary M. Galin

Martyr Halyn is a military surgeon, from 1888 is a doctor of medicine in the Emperor’s university of Saint Volodymyr (now the Kyiv National University named after T.Shevchenko), from 1901 is a knight of the order of Saint Stanislav and other battle rewards of the Rossian empire. In 1903 appointed by the performer of duties of the Main doctor of the Kyiv combat hospital. Halyn, together with М.Hrushevskyj et al, participated in founding of 1907 of Ukrainian Scientific Society in Kyiv, from 1908 chairman of its naturally-medical section, and from 1911 – chairman of medical section, chairman of terminological commission on medicine. In 1917-1921 Halyn participated in organization of sanitary corps of UNR Army, cornet general, at hetmanate in 1918 managed the Terminological commission of Ministry of People’s Health and Guardianship, compiler of the first Ukrainian medical dictionaries. In 1920 at soviet power went out separate edition him the «Rosijsko-ukrajinskyj medychnyj slovnyk». To editing of dictionary of Halyn the known Ukrainian figures were attracted, that gave to its high scientific level, as professional so language, and confession of all Ukraine. One hundred years ago the Ukrainian linguists were oriented in term formation to national character of term system: all scientists agreed, that it is necessary to involve internal language resources, terminate new concepts with the use of both existent facilities and by scientific creation from own sources. Terminological principles of М. Halyn there were base on most lexicographic medical labours of 1920 years, among that there is the «Shkilnyj medychnyj slovnyk (za Halynym)» of B. Aleksandrovskyj (Poltava, 1924); «Nomenklatura khorob (latynsko-ukrajinski nazvy khorob ta rosijskyj pokazhchyk do nykh)» of O. Korchak-Chepurkivskyj (Kyiv, 1927) and most ukrainian medical dictionary of time of the shot up renaissance, final edition that time – «Medychnyj rosijsko-ukrajinskyj slovnyk» of Dr. V.Kysiljov (Odesa, 1928), that was accepted to printing of the medical section of the natural department of the Institute of scientific ukrainian language of the Allukrainian academy of sciences. Most given out then medical dictionaries often are not even in the large libraries of Ukraine. They were destroyed or hidden in the special storehouses, and about their existence presently it knows only to the very narrow circle. Only the Ukrainian diaspore did not forget Martyr Halyn. In 1969 Ukrainian Medical Association of North America reprinted in Detroit the Prague «Medychnyj latynsko-ukrajinskyj slovnyk» of М. Halyn in 1926. On occasion of 100 years of edition of him the «Rosijsko-ukrajinskyj medychnyj slovnyk» we must honour Martyr Halyn and to rehabilitate his approach in relation to creation of the ukrainian medical terms, his vision of «ukrainian medical terminology of the future» that was reflected in his dictionaries. М. Halyn had pride enough of place in the Rossian empire, but he chose the Ukrainian state. He was far unindifferent to the fate of his nation and native word, whatever will say about majority of modern ruler of Ukraine.