Embedding of semantic predications

This paper concerns the generation of distributed vector representations of biomedical concepts from structured knowledge, in the form of subject-relation-object triplets known as semantic predications. Specifically, we evaluate the extent to which a representational approach we have developed for this purpose previously, known as Predication-based Semantic Indexing (PSI), might benefit from insights gleaned from neural-probabilistic language models, which have enjoyed a surge in popularity in recent years as a means to generate distributed vector representations of terms from free text. To do so, we develop a novel neural-probabilistic approach to encoding predications, called Embedding of Semantic Predications (ESP), by adapting aspects of the Skipgram with Negative Sampling (SGNS) algorithm to this purpose. We compare ESP and PSI across a number of tasks including recovery of encoded information, estimation of semantic similarity and relatedness, and identification of potentially therapeutic and harmful relationships using both analogical retrieval and supervised learning. We find advantages for ESP in some, but not all of these tasks, revealing the contexts in which the additional computational work of neural-probabilistic modeling is justified.     

Monitoring adverse events following immunization with a new conjugate vaccine against group A meningococcus in Niger, September 2010

Introduction

MenAfriVac is a new conjugate vaccine against Neisseria meningitidis serogroup A, the major cause of meningitis outbreaks in sub-Saharan Africa. In Niger, the MenAfriVac introduction campaign was conducted in the District of Filingue, during September 2010, targeting 392,211 individuals aged 1–29 years. We set up an enhanced spontaneous surveillance system to monitor adverse events following immunization (AEFI) during the campaign period and 42 days thereafter.

Methods

All the 33 health centres of the district have been designated as surveillance units, which reported AEFIs on a daily basis to the health district headquarters. Health care workers were instructed to screen patients presenting with predefined conditions of interest and patients spontaneously presenting at units or at vaccination posts with complaints after vaccination. Cases were classified as serious (resulting in death, hospitalization or long-term disability) or minor. A National Expert Committee was established to determine if serious cases were causally associated with the vaccine.

Results

In total, 356,532 vaccine doses were administered. During 61 days of monitoring, 82 suspected AEFIs were reported: 16 severe and 66 minor. The cumulative incidence was of 23.0 per 100,000 doses. Among severe cases, 14 were classified as coincidences, one urticaria complicated by respiratory distress was classified as a probable vaccine reaction, and one death was unclassifiable because post-mortem information was unavailable. The number of units that reported at least one case was 19/33 (57.6%).

Conclusions

Although these results are limited by underreporting of cases, we did not identify safety concerns with MenAfriVac. The lessons learned from this experience should be used to reinforce the national pharmacovigilance system in Niger to make it complaint with international standards. In order to do so, we recommend using a lighter system for routine; and conducting regular training and supervisory activities to increase its acceptance among local health workers.     

Pharmacovigilance in Europe and North America: divergent approaches

Although international medicines regulators adopt a common system to assess the safety and efficacy of new drugs, pre-market evaluation is recognized as incomplete given the much larger post-market experience to follow. Adverse drug reactions contribute to more than 100,000 deaths in the United States annually and are among the top 10 leading causes of death. Regulators are developing active surveillance approaches to assess the risks of medicines in the post-market phase to enhance passive adverse drug reaction reporting systems that capture only one to ten percent of ADRs. The objective of this study is to compare international approaches to active surveillance and the manner in which regulatory agencies access and use post-market evidence in their decisions. A conceptual framework is used to guide the comparative analysis of pharmacovigilance governance and policy in the United Kingdom, France, the European Union, the United States and Canada using data gathered from key informant interviews and document review. While research networks are emerging internationally, we found a greater reliance on industry funding and oversight of post-market research in Europe compared to an emphasis on publicly funded programs in North America.     

Comparison of pharmacovigilance algorithms in drug hypersensitivity reactions

Background A firm diagnosis of drug hypersensitivity, because it may re-induce the reaction, is seldom confirmed. Causality assessment algorithms are therefore of interest.Aims The objective of this work was to compare three algorithms in the diagnosis of drug hypersensitivity.Methods Evaluation of 120 clinical histories of drug hypersensitivity was carried out: 60 involving -lactams (50%) and 60 involving non-steroidal anti-inflammatory drugs (50%). Each of these groups of patients underwent a standardised allergy diagnosis, which included a detailed anamnesis, skin tests and, often, provocation tests under strict hospital surveillance. Unaware of the final allergy diagnosis, scores were established for all of the cases and compared using algorithms suggested by Begaud and coworkers [2, 20], Jones [13] and Naranjo et al. [21].Results Although the methods of Jones [13] and Naranjo et al. [21] were perfectly concordant (k=1), no concordance was noted using the Begaud and coworkers [2, 20] method.Conclusions All three algorithms are dissimilar regarding the diagnosis of drug hypersensitivity.     

First experiences with a tool to measure the level of clinical information present in adverse drug reaction reports

Background: To make a proper causality assessment of an adverse drug reaction (ADR) report, a certain level of clinical information is necessary. A tool was developed to measure the level of clinical information present in ADR reports. The aim of this study was to test the validity and reliability of the clinical documentation tool (ClinDoc) in an international setting.

Methods: The tool was developed by a panel of pharmacovigilance experts. It includes four domains: ADR, chronology of the ADR, suspected drug and patient characteristics. The final score categorizes reports into: excellent, well, moderately or poorly documented.

In two rounds, eight pharmacovigilance assessors of different countries made a total of 224 assessments using the tool, with the expert panels judgement as a standard. Sensitivity and specificity were calculated.

Results: The tool with four outcome-categories demonstrated low sensitivity. A lack of distinctiveness was demonstrated between the categories moderate and well. Results for the second round were re-analysed using three categories. This demonstrated a better validity.

Conclusion: This is the first tool to give insight in the level of relevant clinical information present in ADR reports. It can be used internationally to compare reports coming from different reporting methods and different types of reporters in pharmacovigilance.     

Good Practice in the Postmarketing Surveillance of Medicines.

In addition to Good Clinical Trial Practice for the study of experimental drugs, regulations are also needed for good practice in the assessment of medicines after approval (Good PMS Practice, GPP). GPP has to protect the interests of public health at large as well as those of individual patients, investigators and pharmaceutical companies. GPP may be the natural way to solve threatening conflicts between privacy legislation and the public interest.     

Spontaneous reporting of hepatotoxicity associated with antiandrogens: data from the Spanish pharmacovigilance system

PURPOSE: To analyse the type and main features of the hepatotoxicity induced by steroidal and non-steroidal antiandrogens spontaneously reported by physicians, pharmacists and nurses. This analysis could increase the information related to these adverse reactions mainly available from the published isolated cases. METHODS: Using the Spanish Pharmacovigilance database we searched for spontaneous reports recorded since the date of approval of each antiandrogen up to the present time. We analysed the frequency of liver disorders, the preferred terms coded, the presence of other hepatotoxic drugs, and the characteristics of cases of hepatitis. RESULTS: Liver disorders were the most common adverse reactions associated with flutamide and bicalutamide, but not with cyproterone acetate. 'Hepatitis' and 'cholestatic hepatitis' were the most frequent terms coded. In 38% of the reports related to cyproterone acetate, 18% of those related to flutamide and 33% of those related to bicalutamide the patient had simultaneously received other hepatotoxic drugs. The disproportionality analysis of hepatitis showed a strong association with flutamide and a weak association with bicalutamide and cyproterone acetate. Mean doses of flutamide and bicalutamide were very close to their defined daily dose (DDD) to treat prostate cancer, although in the case of cyproterone acetate it was slightly higher. The latency period of hepatitis was between 3 and 10 months for the three antiandrogens, and the recovery period was shorter (0.5-3 months). The majority of the reported cases of hepatitis evolved favourably. CONCLUSION: Our results highlight the hepatotoxic potential of flutamide compared to cyproterone acetate. The data related to bicalutamide should be cautiously considered due to the smaller number of reports.