Methods for safety signal detection in healthcare databases: a literature review

The evolving role of disproportionality analysis in pharmacovigilance

INTRODUCTION

From 2009 to 2015, the IMI PROTECT conducted rigorous studies addressing questions about optimal implementation and significance of disproportionality analyses, leading to the development of Good Signal Detection Practices. The ensuing period witnessed the independent exploration of research paths proposed by IMI PROTECT, accumulating valuable experience and insights that have yet to be seamlessly integrated.

AREAS COVERED

This state-of-the-art review integrates IMI PROTECT recommendations with recent acquisitions and evolving challenges. It deals with defining the object of study, disproportionality methods, subgrouping, masking, drug-drug interaction, duplication, expectedness, the debated use of disproportionality results as risk measures, integration with other types of data.

EXPERT OPINION

Despite the ongoing skepticism regarding the usefulness of disproportionality analyses and individual case safety reports, their ability to timely detect safety signals regarding rare and unpredictable adverse reactions remains unparalleled. Moreover, recent exploration into their potential for characterizing safety signals revealed valuable insights concerning potential risk factors and the patient's perspective. To fully realize their potential beyond hypothesis generation and achieve a comprehensive evidence synthesis with other kinds of data and studies, each with their unique limitations and contributions, we need to investigate methods for more transparently communicating disproportionality results and mapping and addressing pharmacovigilance biases.

Optimal Significance Levels and Sample Sizes for Signal Detection Methods Based on Non-constant Hazards

BACKGROUND AND OBJECTIVES

Statistical methods for signal detection of adverse drug reactions (ADRs) in electronic health records (EHRs) need information about optimal significance levels and sample sizes to achieve sufficient power. Sauzet and Cornelius proposed tests for signal detection based on the hazard functions of Weibull type distributions (WSP tests) which use the time-to-event information available in EHRs. Optimal significance levels and sample sizes for the application of the WPS tests are derived.

METHOD

A simulation study was performed with a range of scenarios for sample size, rate of event due (ADRs), and not due to the drug and random time to ADR occurrence. Based on the area under the curve of the receiver operating characteristic graph, we obtain optimal significance levels of the different WSP tests for the implementation in a hypothesis free signal detection setting and approximate sample sizes required to reach a power of 80% or 90%.

RESULTS

The dWSP-pPWSP (combination of double WSP and power WSP) test with a significance level of 0.004 was recommended. Sample sizes needed for a power of 80% were found to start at 60 events for an ADR rate equal to the background rate of 0.1. The number of events required for a background rate of 0.05 and an ADR rate equal to a 20% increase of the background rate was 900.

CONCLUSION

Based on this study, it is recommended to use the dWSP-pWSP test combination for signal detection with a significance level of 0.004 when the same test is applied to all adverse events not depending on rates.

Sudden Sensorineural Hearing Loss after COVID-19 Vaccination: A Review of the Available Evidence through the Prism of Causality Assessment

Sudden sensorineural hearing loss (SSNHL), a rare audiological condition that accounts for 1% of all cases of sensorineural hearing loss, can cause permanent hearing damage. Soon after the launch of global COVID-19 vaccination campaigns, the World Health Organization released a signal detection about SSNHL cases following administration of various COVID-19 vaccines. Post-marketing studies have been conducted in different countries using either pharmacovigilance or medico-administrative databases to investigate SSNHL as a potential adverse effect of COVID-19 vaccines. Here, we examine the advantages and limitations of each type of post-marketing study available. While pharmacoepidemiological studies highlight the potential association between drug exposure and the event, pharmacovigilance approaches enable causality assessment. The latter objective can only be achieved if an expert evaluation is provided using internationally validated diagnostic criteria. For a rare adverse event such as SSNHL, case information and quantification of hearing loss are mandatory for assessing seriousness, severity, delay onset, differential diagnoses, corrective treatment, recovery, as well as functional sequelae. Appropriate methodology should be adopted depending on whether the target objective is to assess a global or individual risk.

High-Throughput Computing to Automate Population-Based Studies to Detect the 30-Day Risk of Adverse Outcomes After New Outpatient Medication Use in Older Adults with Chronic Kidney Disease: A Clinical Research Protocol

Background

Safety issues are detected in about one third of prescription drugs in the years following regulatory agency approval. Older adults, especially those with chronic kidney disease, are at particular risk of adverse reactions to prescription drugs. This protocol describes a new approach that may identify credible drug-safety signals more efficiently using administrative health care data.

Objective

To use high-throughput computing and automation to conduct 700+ drug-safety cohort studies in older adults in Ontario, Canada. Each study will compare 74 acute (30-day) outcomes in patients who start a new prescription drug (new users) to a group of nonusers with similar baseline health characteristics. Risks will be assessed within strata of baseline kidney function.

Design and setting

The studies will be population-based, new-user cohort studies conducted using linked administrative health care databases in Ontario, Canada (January 1, 2008, to March 1, 2020). The source population for these studies will be residents of Ontario aged 66 years or older who filled at least one outpatient prescription through the Ontario Drug Benefit (ODB) program during the study period (all residents have universal health care, and those aged 65+ have universal prescription drug coverage through the ODB).

Patients

We identified 3.2 million older adults in the source population during the study period and built 700+ initial medication cohorts, each containing mutually exclusive groups of new users and nonusers. Nonusers were randomly assigned cohort entry dates that followed the same distribution of prescription start dates as new users. Eligibility criteria included a baseline estimated glomerular filtration rate (eGFR) measurement within 12 months before the cohort entry date (median time was 71 days before cohort entry in the new user group), no prior receipt of maintenance dialysis or a kidney transplant, and no prior prescriptions for drugs in the same subclass as the study drug. New users and nonusers will be balanced on ~400 baseline health characteristics using inverse probability of treatment weighting on propensity scores within 3 strata of baseline eGFR: ≥60, 45 to <60, <45 mL/min per 1.73 m2.

Outcomes

We will compare new user and nonuser groups on 74 clinically relevant outcomes (17 composites and 57 individual outcomes) in the 30 days after cohort entry. We used a prespecified approach to identify these 74 outcomes.

Statistical analysis plan

In each cohort, we will obtain eGFR-stratum-specific weighted risk ratios and risk differences using modified Poisson regression and binomial regression, respectively. Additive and multiplicative interaction by eGFR category will be examined. Drug-outcome associations that meet prespecified criteria (identified signals) will be further examined in additional analyses (including survival, negative-control exposure, and E-value analyses) and visualizations.

Results

The initial medication cohorts had a median of 6120 new users per cohort (interquartile range: 1469-38 839) and a median of 1 088 301 nonusers (interquartile range: 751 697-1 267 009). Medications with the largest number of new users were amoxicillin trihydrate (n = 1 000 032), cephalexin (n = 571 566), prescription acetaminophen (n = 571 563), and ciprofloxacin (n = 504,374); 19% to 29% of new users in these cohorts had an eGFR <60 mL/min per 1.73 m2.

Limitations

Despite our use of robust techniques to balance baseline indicators and to control for confounding by indication, residual confounding will remain a possibility. Only acute (30-day) outcomes will be examined. Our data sources do not include nonprescription (over-the-counter) drugs or drugs prescribed in hospitals and do not include outpatient prescription drug use in children or adults <65 years.

Conclusion

This accelerated approach to conducting postmarket drug-safety studies has the potential to more efficiently detect drug-safety signals in a vulnerable population. The results of this protocol may ultimately help improve medication safety.

Epidemiological surveillance of drug safety using cumulative sequential analysis in electronic healthcare data

BACKGROUND

Methods for safety signal detection in electronic healthcare data analysing data sequentially are being developed to meet the limitations of spontaneous reporting systems.

OBJECTIVES

This study aims to provide an overview of the literature on sequential analysis of electronic healthcare data and describe the development and testing of a novel epidemiological surveillance system.

METHODS

We searched Medline, Embase, PubMed, Scopus, Web of Science, and the Cochrane Library applying similar in- and exclusion criteria as those of a previous systematic review. The proposed system consisted of repeated cohort studies and was tested in an emulated prospective setting. Two signal evaluations were performed with several sensitivity analyses and a target trial emulation.

FINDINGS

In the literature, 11 studies analysed the data sequentially of which two applied traditional epidemiological methods. Epidemiological surveillance of several exposures and outcomes can be successfully conducted with the newly proposed sequential analysis of electronic healthcare data. Signal evaluation studies confirmed the results of the system.

CONCLUSIONS

Very few studies in the literature analysed data at multiple time points, although this seems to be a prerequisite for testing the methods in a realistic setting. We demonstrated the feasibility of a sequential surveillance system using electronic healthcare data.

Impact of the COVID-19 pandemic on the spontaneous reporting and signal detection of adverse drug events

External factors severely affecting in a short period of time the spontaneous reporting of adverse events (AEs) can significantly impact drug safety signal detection. Coronavirus disease 2019 (COVID-19) represented an enormous challenge for health systems, with over 767 million cases and massive vaccination campaigns involving over 70% of the worldwide population. This study investigates the potential masking effect on certain AEs caused by the substantial increase in reports solely related to COVID-19 vaccines within various spontaneous reporting systems (SRSs). Three SRSs were used to monitor AEs reporting before and during the pandemic, namely, the World Health Organisation (WHO) global individual case safety reports database (VigiBase®), the United States Food and Drug Administration Adverse Event Reporting System (FAERS) and the Japanese Adverse Drug Event Report database (JADER). Findings revealed a sudden over-reporting of 35 AEs (≥ 200%) during the pandemic, with an increment of the RRF value in 2021 of at least double the RRF reported in 2020. This translates into a substantial reduction in signals of disproportionate reporting (SDR) due to the massive inclusion of COVID-19 vaccine reports. To mitigate the masking effect of COVID-19 vaccines in post-marketing SRS analyses, we recommend utilizing COVID-19-corrected versions for a more accurate assessment.

Pharmacovigilance studies without a priori hypothesis: systematic review highlights inappropriate multiple testing correction procedures

OBJECTIVES

The purpose of this study was to systematically review the statistical methods used in pharmacovigilance studies without a priori hypotheses.

STUDY DESIGN AND SETTING

A systematic review was performed on studies published in the MEDLINE database between 2012 and 2021. The included studies were analyzed for database name and type, statistical methods, anatomical therapeutic chemical class for the studied drug(s), and SOC MedDRA classification for the studied adverse drug reaction.

RESULTS

Ninety-two studies were included, with pharmacovigilance databases being the most used type. Disproportionality analysis using frequentist or Bayesian methods was the most common statistical method employed. The most studied drug classes were anti-infectives, nervous system drugs, and antineoplastics and immunomodulators. However, no common procedure was implemented to correct for multiple testing.

CONCLUSION

This review highlights the limited number of statistical methods employed for pharmacovigilance studies without a priori hypotheses, with no established consensus-based method and a lack of interest in multiple testing correction. The establishment of guidelines is recommended to improve the performance of such studies.

Text mining of electronic health records can validate a register-based diagnosis of epilepsy and subgroup into focal and generalized epilepsy

OBJECTIVE

Combining population-based health registries and electronic health records offers the opportunity to create large, phenotypically detailed patient cohorts of high quality. In this study, we used text mining of clinical notes to confirm International Classification of Diseases, 10th Revision (ICD-10)-registered epilepsy diagnoses and classify patients according to focal and generalized epilepsy types.

METHODS

Using the Danish National Patient Registry, we identified patients who between 2006 and 2016 received an ICD-10 diagnosis of epilepsy. To validate the epilepsy diagnosis and stratify patients into focal and generalized epilepsy types, we constructed dictionaries for text mining-based extraction of clinical notes. Two physicians manually reviewed the clinical notes for a total of 527 patients and assigned epilepsy diagnoses, which were compared with the text-mined diagnoses.

RESULTS

We identified 23 632 patients with an ICD-10 diagnosis of epilepsy, of whom 50% were registered with an unspecified epilepsy diagnosis. In total, 11 211 patients were considered likely to have epilepsy by text mining, with an F1 measure ranging from 82% to 90%. Manual review of the electronic health records for 310 patients revealed a false discovery rate of 29%. This rate was decreased to 4% by the text mining algorithm. The weighted average F1 measure for text mining-assigned epilepsy types was 79% (82% for focal and 76% for generalized epilepsy). Text mining successfully assigned a focal or generalized epilepsy type to 92% of the text mining-eligible patients registered with unspecified epilepsy.

SIGNIFICANCE

Text mining of electronic health records can be used to establish a patient cohort with much higher likelihood of having a diagnosis of epilepsy and a focal or generalized epilepsy type compared to the cohort created from ICD-10 epilepsy codes alone. We believe the concept will be essential for future genome-wide and phenome-wide association studies and subsequently the development of precision medicine for epilepsy patients.

A New Drug Safety Signal Detection and Triage System Integrating Sequence Symmetry Analysis and Tree-Based Scan Statistics with Longitudinal Data

Purpose

Development and evaluation of a drug-safety signal detection system integrating data-mining tools in longitudinal data is essential. This study aimed to construct a new triage system using longitudinal data for drug-safety signal detection, integrating data-mining tools, and evaluate adaptability of such system.

Patients and Methods

Based on relevant guidelines and structural frameworks in Taiwan's pharmacovigilance system, we constructed a triage system integrating sequence symmetry analysis (SSA) and tree-based scan statistics (TreeScan) as data-mining tools for detecting safety signals. We conducted an exploratory analysis utilizing Taiwan's National Health Insurance Database and selecting two drug classes (sodium-glucose co-transporter-2 inhibitors (SGLT2i) and non-fluorinated quinolones (NFQ)) as chronic and episodic treatment respectively, as examples to test feasibility of the system.

Results

Under the proposed system, either cohort-based or self-controlled mining with SSA and TreeScan was selected, based on whether the screened drug had an appropriate comparator. All detected alerts were further classified as known adverse drug reactions (ADRs), events related to other causes or potential signals from the triage algorithm, building on existing drug labels and clinical judgement. Exploratory analysis revealed greater numbers of signals for NFQ with a relatively low proportion of known ADRs; most were related to indication, patient characteristics or bias. No safety signals were found. By contrast, most SGLT2i signals were known ADRs or events related to patient characteristics. Four were potential signals warranting further investigation.

Conclusion

The proposed system facilitated active and systematic screening to detect and classify potential safety signals. Countries with real-world longitudinal data could adopt it to streamline drug-safety surveillance.

Methods for drug safety signal detection using routinely collected observational electronic health care data: A systematic review

PURPOSE

Signal detection is a crucial step in the discovery of post-marketing adverse drug reactions. There is a growing interest in using routinely collected data to complement established spontaneous report analyses. This work aims to systematically review the methods for drug safety signal detection using routinely collected healthcare data and their performance, both in general and for specific types of drugs and outcomes.

METHODS

We conducted a systematic review following the PRISMA guidelines, and registered a protocol in PROSPERO. MEDLINE, EMBASE, PubMed, Web of Science, Scopus, and the Cochrane Library were searched until July 13, 2021.

RESULTS

The review included 101 articles, among which there were 39 methodological works, 25 performance assessment papers, and 24 observational studies. Methods included adaptations from those used with spontaneous reports, traditional epidemiological designs, methods specific to signal detection with real-world data. More recently, implementations of machine learning have been studied in the literature. Twenty-five studies evaluated method performances, 16 of them using the area under the curve (AUC) for a range of positive and negative controls as their main measure. Despite the likelihood that performance measurement could vary by drug-event pair, only 10 studies reported performance stratified by drugs and outcomes, in a heterogeneous manner. The replicability of the performance assessment results was limited due to lack of transparency in reporting and the lack of a gold standard reference set.

CONCLUSIONS

A variety of methods have been described in the literature for signal detection with routinely collected data. No method showed superior performance in all papers and across all drugs and outcomes, performance assessment and reporting were heterogeneous. However, there is limited evidence that self-controlled designs, high dimensional propensity scores, and machine learning can achieve higher performances than other methods.

Identifying new drugs associated with pulmonary arterial hypertension: A WHO pharmacovigilance database disproportionality analysis

Since the 1960s, several drugs have been linked to the onset or aggravation of pulmonary arterial hypertension (PAH): dasatinib, some amphetamine-like appetite suppressants (aminorex, fenfluramine, dexfenfluramine, benfluorex) and recreational drugs (methamphetamine). Moreover, in numerous cases, the implication of other drugs with PAH have been suggested, but the precise identification of iatrogenic aetiologies of PAH is challenging given the scarcity of this disease and the potential long latency period between drug intake and PAH onset. In this context, we used the World Health Organization's pharmacovigilance database, VigiBase, to generate new hypotheses about drug associated PAH.

METHODS

We used VigiBase, the largest pharmacovigilance database worldwide to generate disproportionality signals through the Bayesian neural network method. All disproportionality signals were further independently reviewed by experts in pulmonary arterial hypertension, pharmacovigilance and vascular pharmacology and their plausibility ranked according to World Health Organization causality categories.

RESULTS

We included 2184 idiopathic PAH cases, yielding a total of 93 disproportionality signals. Among them, 25 signals were considered very likely, 15 probable, 28 possible and 25 unlikely. Notably, we identified 4 new protein kinases inhibitors (lapatinib, lorlatinib, ponatinib and ruxolitinib), 1 angiogenesis inhibitor (bevacizumab), and several chemotherapeutics (etoposide, trastuzumab), antimetabolites (cytarabine, fludarabine, fluorouracil, gemcitabine) and immunosuppressants (leflunomide, thalidomide, ciclosporin).

CONCLUSION

Such signals represent plausible adverse drug reactions considering the knowledge of iatrogenic PAH, the drugs' biological and pharmacological activity and the characteristics of the reported case. Although confirmatory studies need to be performed, the signals identified may help clinicians envisage an iatrogenic aetiology when faced with a patient who develops PAH.

Medicine-Induced Acute Kidney Injury Findings from Spontaneous Reporting Systems, Sequence Symmetry Analysis and a Case-Control Study with a Focus on Medicines Used in Primary Care

INTRODUCTION

Primary care provides an opportunity to prevent community acquired, medicine or drug-induced acute kidney injury. One of the barriers to proactive prevention of medicine-induced kidney injury in primary care is the lack of a list of nephrotoxic medicines that are most problematic in primary care, particularly one that provides a comparison of risks across medicines.

OBJECTIVE

The aim of this study was to consolidate evidence on the risks associated with medicines and acute kidney injury, with a focus on medicines used in primary care.

METHOD

We searched the MEDLINE and EMBASE databases to identify published studies of all medicines associated with acute kidney injury identified from spontaneous report data. For each medicine positively associated with acute kidney injury, as identified from spontaneous reports, we implemented a sequence symmetry analysis (SSA) and a case-control design to determine the association between the medicine and hospital admission with a primary diagnosis of acute kidney injury (representing community-acquired acute kidney injury). Administrative claims data held by the Australian Government Department of Veterans' Affairs for the study period 2005-2019 were used.

RESULTS

We identified 89 medicines suspected of causing acute kidney injury based on spontaneous report data and a reporting odds ratio above 2, from Japan, France and the US. Spironolactone had risk estimates of 3 or more based on spontaneous reports, SSA and case-control methods, while furosemide and trimethoprim with sulfamethoxazole had risk estimates of 1.5 or more. Positive association with SSA and spontaneous reports, but not case control, showed zoledronic acid had risk estimates above 2, while candesartan telmisartan, simvastatin, naproxen and ibuprofen all had risk estimates in SSA between 1.5 and 2. Positive associations with case-control and spontaneous reports, but not SSA, were found for amphotericin B, omeprazole, metformin, amlodipine, ramipril, olmesartan, ciprofloxacin, valaciclovir, mycophenolate and diclofenac. All with the exception of metformin and omeprazole had risk estimates above 2.

CONCLUSION

This research highlights a number of medicines that may contribute to acute injury; however, we had an insufficient sample to confirm associations of some medicines. Spironolactone, furosemide, and trimethoprim with sulfamethoxazole are medicines that, in particular, need to be used carefully and monitored closely in patients in the community at risk of acute kidney injury.

Analyzing adverse drug reaction using statistical and machine learning methods: A systematic review

BACKGROUND

Adverse drug reactions (ADRs) are unintended negative drug-induced responses. Determining the association between drugs and ADRs is crucial, and several methods have been proposed to demonstrate this association. This systematic review aimed to examine the analytical tools by considering original articles that utilized statistical and machine learning methods for detecting ADRs.

METHODS

A systematic literature review was conducted based on articles published between 2015 and 2020. The keywords used were statistical, machine learning, and deep learning methods for detecting ADR signals. The study was conducted according to the Preferred Reporting Items for Systematic Reviews and Meta-Analyses statement (PRISMA) guidelines.

RESULTS

We reviewed 72 articles, of which 51 and 21 addressed statistical and machine learning methods, respectively. Electronic medical record (EMR) data were exclusively analyzed using the regression method. For FDA Adverse Event Reporting System (FAERS) data, components of the disproportionality method were preferable. DrugBank was the most used database for machine learning. Other methods accounted for the highest and supervised methods accounted for the second highest.

CONCLUSIONS

Using the 72 main articles, this review provides guidelines on which databases are frequently utilized and which analysis methods can be connected. For statistical analysis, >90% of the cases were analyzed by disproportionate or regression analysis with each spontaneous reporting system (SRS) data or electronic medical record (EMR) data; for machine learning research, however, there was a strong tendency to analyze various data combinations. Only half of the DrugBank database was occupied, and the k-nearest neighbor method accounted for the greatest proportion.

Engaging Patients via Online Healthcare Fora: Three Pharmacovigilance Use Cases

Increasingly, patient-generated safety insights are shared online, via general social media platforms or dedicated healthcare fora which give patients the opportunity to discuss their disease and treatment options. We evaluated three areas of potential interest for the use of social media in pharmacovigilance. To evaluate how social media may complement existing safety signal detection capabilities, we identified two use cases (drug/adverse event [AE] pairs) and then evaluated the frequency of AE discussions across a range of social media channels. Changes in frequency over time were noted in social media, then compared to frequency changes in Food and Drug Administration Adverse Event Reporting System (FAERS) data over the same time period using a traditional disproportionality method. Although both data sources showed increasing frequencies of AE discussions over time, the increase in frequency was greater in the FAERS data as compared to social media. To demonstrate the robustness of medical/AE insights of linked posts we manually reviewed 2,817 threads containing 21,313 individual posts from 3,601 unique authors. Posts from the same authors were linked together. We used a quality scoring algorithm to determine the groups of linked posts with the highest quality and manually evaluated the top 16 groups of posts. Most linked posts (12/16; 75%) contained all seven relevant medical insights assessed compared to only one (of 1,672) individual post. To test the capability of actively engage patients via social media to obtain follow-up AE information we identified and sent consents for follow-up to 39 individuals (through a third party). We sent target follow-up questions (identified by pharmacovigilance experts as critical for causality assessment) to those who consented. The number of people consenting to follow-up was low (20%), but receipt of follow-up was high (75%). We observed completeness of responses (37 out of 37 questions answered) and short average time required to receive the follow-up (1.8 days). Our findings indicate a limited use of social media data for safety signal detection. However, our research highlights two areas of potential value to pharmacovigilance: obtaining more complete medical/AE insights via longitudinal post linking and actively obtaining rapid follow-up information on AEs.

Evaluating the performance of temporal pattern discovery: new application using statins and rhabdomyolysis in OMOP databases

Background

Temporal pattern discovery (TPD) is a method of signal detection using electronic healthcare databases, serving as an alternative to spontaneous reporting of adverse drug events. Here, we aimed to replicate and optimise a TPD approach previously used to assess temporal signals of statins with rhabdomyolysis (in The Health Improvement Network (THIN) database) by using the OHDSI tools designed for OMOP data sources.

Methods

We used data from the Truven MarketScan US Commercial Claims and the Commercial Claims and Encounters (CCAE). Using an extension of the OHDSI ICTemporalPatternDiscovery package, we ran positive and negative controls through four analytical settings and calculated sensitivity, specificity, bias and AUC to assess performance.

Results

Similar to previous findings, we noted an increase in the Information Component (IC) for simvastatin and rhabdomyolysis following initial exposure and throughout the surveillance window. For example, the change in IC was 0.266 for the surveillance period of 1–30 days as compared to the control period of − 180 to − 1 days. Our modification of the existing OHDSI software allowed for faster queries and more efficient generation of chronographs.

Conclusion

Our OMOP replication matched the we can account forwe can account for of the original THIN study, only simvastatin had a signal. The TPD method is a useful signal detection tool that provides a single statistic on temporal association and a graphical depiction of the temporal pattern of the drug outcome combination. It remains unclear if the method works well for rare adverse events, but it has been shown to be a useful risk identification tool for longitudinal observational databases. Future work should compare the performance of TPD with other pharmacoepidemiology methods and mining techniques of signal detection. In addition, it would be worth investigating the relative TPD performance characteristics using a variety of observational data sources.

Applying Machine Learning in Distributed Data Networks for Pharmacoepidemiologic and Pharmacovigilance Studies: Opportunities, Challenges, and Considerations

Increasing availability of electronic health databases capturing real-world experiences with medical products has garnered much interest in their use for pharmacoepidemiologic and pharmacovigilance studies. The traditional practice of having numerous groups use single databases to accomplish similar tasks and address common questions about medical products can be made more efficient through well-coordinated multi-database studies, greatly facilitated through distributed data network (DDN) architectures. Access to larger amounts of electronic health data within DDNs has created a growing interest in using data-adaptive machine learning (ML) techniques that can automatically model complex associations in high-dimensional data with minimal human guidance. However, the siloed storage and diverse nature of the databases in DDNs create unique challenges for using ML. In this paper, we discuss opportunities, challenges, and considerations for applying ML in DDNs for pharmacoepidemiologic and pharmacovigilance studies. We first discuss major types of activities performed by DDNs and how ML may be used. Next, we discuss practical data-related factors influencing how DDNs work in practice. We then combine these discussions and jointly consider how opportunities for ML are affected by practical data-related factors for DDNs, leading to several challenges. We present different approaches for addressing these challenges and highlight efforts that real-world DDNs have taken or are currently taking to help mitigate them. Despite these challenges, the time is ripe for the emerging interest to use ML in DDNs, and the utility of these data-adaptive modeling techniques in pharmacoepidemiologic and pharmacovigilance studies will likely continue to increase in the coming years.

"Artificial Intelligence" for Pharmacovigilance: Ready for Prime Time?

There is great interest in the application of 'artificial intelligence' (AI) to pharmacovigilance (PV). Although US FDA is broadly exploring the use of AI for PV, we focus on the application of AI to the processing and evaluation of Individual Case Safety Reports (ICSRs) submitted to the FDA Adverse Event Reporting System (FAERS). We describe a general framework for considering the readiness of AI for PV, followed by some examples of the application of AI to ICSR processing and evaluation in industry and FDA. We conclude that AI can usefully be applied to some aspects of ICSR processing and evaluation, but the performance of current AI algorithms requires a 'human-in-the-loop' to ensure good quality. We identify outstanding scientific and policy issues to be addressed before the full potential of AI can be exploited for ICSR processing and evaluation, including approaches to quality assurance of 'human-in-the-loop' AI systems, large-scale, publicly available training datasets, a well-defined and computable 'cognitive framework', a formal sociotechnical framework for applying AI to PV, and development of best practices for applying AI to PV. Practical experience with stepwise implementation of AI for ICSR processing and evaluation will likely provide important lessons that will inform the necessary policy and regulatory framework to facilitate widespread adoption and provide a foundation for further development of AI approaches to other aspects of PV.

Use of State Sequence Analysis in Pharmacoepidemiology: A Tutorial

While state sequence analysis (SSA) has been long used in social sciences, its use in pharmacoepidemiology is still in its infancy. Indeed, this technique is relatively easy to use, and its intrinsic visual nature may help investigators to untangle the latent information within prescription data, facilitating the individuation of specific patterns and possible inappropriate use of medications. In this paper, we provide an educational primer of the most important learning concepts and methods of SSA, including measurement of dissimilarities between sequences, the application of clustering methods to identify sequence patterns, the use of complexity measures for sequence patterns, the graphical visualization of sequences, and the use of SSA in predictive models. As a worked example, we present an application of SSA to opioid prescription patterns in patients with non-cancer pain, using real-world data from Italy. We show how SSA allows the identification of patterns in prescriptions in these data that might not be evident using standard statistical approaches and how these patterns are associated with future discontinuation of opioid therapy.

Identifying Actionability as a Key Factor for the Adoption of 'Intelligent' Systems for Drug Safety: Lessons Learned from a User-Centred Design Approach

Introduction

Information technology (IT) plays an important role in the healthcare landscape via the increasing digitization of medical data and the use of modern computational paradigms such as machine learning (ML) and knowledge graphs (KGs). These ‘intelligent’ technical paradigms provide a new digital ‘toolkit’ supporting drug safety and healthcare processes, including ‘active pharmacovigilance’. While these technical paradigms are promising, intelligent systems (ISs) are not yet widely adopted by pharmacovigilance (PV) stakeholders, namely the pharma industry, academia/research community, drug safety monitoring organizations, regulatory authorities, and healthcare institutions. The limitations obscuring the integration of ISs into PV activities are multifaceted, involving technical, legal and medical hurdles, and thus require further elucidation.

Objective

We dissect the abovementioned limitations by describing the lessons learned during the design and implementation of the PVClinical platform, a web platform aiming to support the investigation of potential adverse drug reactions (ADRs), emphasizing the use of knowledge engineering (KE) as its main technical paradigm.

Results

To this end, we elaborate on the related ‘business processes’ (i.e. operational processes) and ‘user goals’ identified as part of the PVClinical platform design process based on Design Thinking principles. We also elaborate on key challenges restricting the adoption of such ISs and their integration in the clinical setting and beyond.

Conclusions

We highlight the fact that beyond providing analytics and useful statistics to the end user, ‘actionability’ has emerged as the operational priority identified through the whole process. Furthermore, we focus on the needs for valid, reproducible, explainable and human-interpretable results, stressing the need to emphasize on usability.

Repurposing haloperidol for the treatment of rheumatoid arthritis: an integrative approach using data mining techniques

Introduction

Treatment of rheumatoid arthritis (RA) has advanced with the introduction of biological disease-modifying antirheumatic drugs. However, more than 20% of patients with RA still have moderate or severe disease activity. Hence, novel antirheumatic drugs are required. Recently, drug repurposing, a process of identifying new indications for existing drugs, has received great attention. Furthermore, a few reports have shown that antipsychotics are capable of affecting several cytokines that are also modulated by existing antirheumatic drugs. Therefore, we investigated the association between antipsychotics and RA by data mining using real-world data and bioinformatics databases.

Methods

Disproportionality and sequence symmetry analyses were employed to identify the associations between the investigational drugs and RA using the US Food and Drug Administration Adverse Event Reporting System (2004-2016) and JMDC administrative claims database (January 2005-April 2017; JMDC Inc., Tokyo, Japan), respectively. The reporting odds ratio (ROR) and information component (IC) were used in the disproportionality analysis to indicate a signal. The adjusted sequence ratio (SR) was used in the sequence symmetry analysis to indicate a signal. The bioinformatics analysis suite, BaseSpace Correlation Engine (Illumina, CA, USA) was employed to explore the molecular mechanisms associated with the potential candidates identified by the drug-repurposing approach.

Results

A potential inverse association between the antipsychotic haloperidol and RA, which exhibited significant inverse signals with ROR, IC, and adjusted SR, was found. Furthermore, the results suggested that haloperidol may exert antirheumatic effects by modulating various signaling pathways, including cytokine and chemokine signaling, major histocompatibility complex class-II antigen presentation, and Toll-like receptor cascade pathways.

Conclusion

Our drug-repurposing approach using data mining techniques identified haloperidol as a potential antirheumatic drug candidate.

The concept of justifiable healthcare and how big data can help us to achieve it

Over the last decades, the face of health care has changed dramatically, with big improvements in what is technically feasible. However, there are indicators that the current approach to evaluating evidence in health care is not holistic and hence in the long run, health care will not be sustainable. New conceptual and normative frameworks for the evaluation of health care need to be developed and investigated. The current paper presents a novel framework of justifiable health care and explores how the use of artificial intelligence and big data can contribute to achieving the goals of this framework.

Machine Learning for Detection of Safety Signals From Spontaneous Reporting System Data: Example of Nivolumab and Docetaxel

Introduction: Various methods have been implemented to detect adverse drug reaction (ADR) signals. However, the applicability of machine learning methods has not yet been fully evaluated.

Objective: To evaluate the feasibility of machine learning algorithms in detecting ADR signals of nivolumab and docetaxel, new and old anticancer agents.

Methods: We conducted a safety surveillance study of nivolumab and docetaxel using the Korea national spontaneous reporting database from 2009 to 2018. We constructed a novel input dataset for each study drug comprised of known ADRs that were listed in the drug labels and unknown ADRs. Given the known ADRs, we trained machine learning algorithms and evaluated predictive performance in generating safety signals of machine learning algorithms (gradient boosting machine [GBM] and random forest [RF]) compared with traditional disproportionality analysis methods (reporting odds ratio [ROR] and information component [IC]) by using the area under the curve (AUC). Each method then was implemented to detect new safety signals from the unknown ADR datasets.

Results: Of all methods implemented, GBM achieved the best average predictive performance (AUC: 0.97 and 0.93 for nivolumab and docetaxel). The AUC achieved by each method was 0.95 and 0.92 (RF), 0.55 and 0.51 (ROR), and 0.49 and 0.48 (IC) for respective drug. GBM detected additional 24 and nine signals for nivolumab and 82 and 76 for docetaxel compared to ROR and IC, respectively, from the unknown ADR datasets.

Conclusion: Machine learning algorithm based on GBM performed better and detected more new ADR signals than traditional disproportionality analysis methods.

Bayesian hierarchical model for safety signal detection in multiple clinical trials

Clinical safety signal detection is of great importance in establishing the safety profile of new drugs and biologics during drug development. Bayesian hierarchical meta-analysis has proven to be a very effective method of identifying potential safety signals by considering the hierarchical structure of clinical safety data from multiple randomized clinical trials conducted under an Investigational New Drug (IND) application or Biological License Application (BLA). This type of model can integrate information across studies, for instance by grouping related adverse events using the MedDRA system-organ-class (SOC) and preferred terms (PT). It therefore improves the precision of parameter estimates compared to models that do not consider the hierarchical structure of the safety data. We propose to extend an existing four-stage Bayesian hierarchical model and consider the exposure adjusted incidence rate, assuming the number of adverse events (AEs) follows a Poisson distribution. The proposed model is applied to a real-world example, using data from three randomized clinical trials of a neuroscience drug and examine in three simulation studies motivated by real-world examples. Comparison is made between the proposed method and other existing methods. The simulation results indicate that our proposed model outperforms other two candidate models in terms of power and false detection rate.

Signals of adverse drug reactions communicated by pharmacovigilance stakeholders: protocol for a scoping review of the global literature

BACKGROUND

Signals of adverse drug reactions (ADRs) form the basis of some regulatory risk-minimization actions in pharmacovigilance. Reviews of limited scope have highlighted that such signals are mostly supported by reports of ADRs or multiple types of evidence. The time that elapses between a report of a suspected ADR and the communication of a signal has not been systematically characterized. Neither has the features of reports of suspected ADRs that authors used to support putative causal relationships, although difficulties with establishing causal relationships between medicinal products and adverse events have been highlighted. The objectives of this study will be to describe the evidence underpinning signals in pharmacovigilance, the features of reports of ADRs supporting signals, and the time that it takes to communicate a signal.

METHODS

We shall retrieve records from PubMed, EMBASE, Web of Science, and PsycINFO (from inception onwards), without language/design restrictions, and apply backward citation screening. We shall hand-search the websites of 35 regulatory agencies/authorities, restricted publications from the Uppsala Monitoring Centre, and drug bulletins. Signals will be requested from the competent stakeholder, if absent from websites. We shall use VigiBase, the World Health Organization's Global Individual Case Safety Report database, to determine the dates on which ADRs were reported. We shall manage records using EndNote (v. 8.2); one reviewer will screen titles/abstracts and full texts, a second will cross-validate the findings, and a third will arbitrate disagreements. Data will be charted via the Systematic Reviews Data Repository, following the same procedures as for data retrieval. Evidence will be categorized according to the Oxford Centre for Evidence-Based Medicine Levels of Evidence. Features of reports of ADRs will be coded. Tables will display frequencies of types of evidence and features of reports of ADRs. We shall use plots or pictograms (if appropriate) to represent the time from the first report of a suspected ADR to a signal.

DISCUSSION

We expect the findings from this review will allow a better understanding of global patterns of similarities or differences in terms of supporting evidence and timing of communications and identify relevant research questions for future systematic reviews. SYSTEMATIC REVIEW REGISTRATION: osf.io/a4xns.

Safety Surveillance of Pneumococcal Vaccine Using Three Algorithms: Disproportionality Methods, Empirical Bayes Geometric Mean, and Tree-Based Scan Statistic

Introduction: Diverse algorithms for signal detection exist. However, inconsistent results are often encountered among the algorithms due to different levels of specificity used in defining the adverse events (AEs) and signal threshold. We aimed to explore potential safety signals for two pneumococcal vaccines in a spontaneous reporting database and compare the results and performances among the algorithms. Methods: Safety surveillance was conducted using the Korea national spontaneous reporting database from 1988 to 2017. Safety signals for pneumococcal vaccine and its subtypes were detected using the following the algorithms: disproportionality methods comprising of proportional reporting ratio (PRR), reporting odds ratio (ROR), and information component (IC); empirical Bayes geometric mean (EBGM); and tree-based scan statistics (TSS). Moreover, the performances of these algorithms were measured by comparing detected signals with the known AEs or pneumococcal vaccines (reference standard). Results: Among 10,380 vaccine-related AEs, 1135 reports and 101 AE terms were reported following pneumococcal vaccine. IC generated the most safety signals for pneumococcal vaccine (40/101), followed by PRR and ROR (19/101 each), TSS (15/101), and EBGM (1/101). Similar results were observed for its subtypes. Cellulitis was the only AE detected by all algorithms for pneumococcal vaccine. TSS showed the best balance in the performance: the highest in accuracy, negative predictive value, and area under the curve (70.3%, 67.4%, and 64.2%). Conclusion: Discrepancy in the number of detected signals was observed between algorithms. EBGM and TSS calibrated noise better than disproportionality methods, and TSS showed balanced performance. Nonetheless, these results should be interpreted with caution due to a lack of a gold standard for signal detection.

Characterizing non-heroin opioid overdoses using electronic health records

INTRODUCTION

The opioid epidemic is a modern public health emergency. Common interventions to alleviate the opioid epidemic aim to discourage excessive prescription of opioids. However, these methods often take place over large municipal areas (state-level) and may fail to address the diversity that exists within each opioid case (individual-level). An intervention to combat the opioid epidemic that takes place at the individual-level would be preferable.

METHODS

This research leverages computational tools and methods to characterize the opioid epidemic at the individual-level using the electronic health record data from a large, academic medical center. To better understand the characteristics of patients with opioid use disorder (OUD) we leveraged a self-controlled analysis to compare the healthcare encounters before and after an individual's first overdose event recorded within the data. We further contrast these patients with matched, non-OUD controls to demonstrate the unique qualities of the OUD cohort.

RESULTS

Our research confirms that the rate of opioid overdoses in our hospital significantly increased between 2006 and 2015 (P < 0.001), at an average rate of 9% per year. We further found that the period just prior to the first overdose is marked by conditions of pain or malignancy, which may suggest that overdose stems from pharmaceutical opioids prescribed for these conditions.

CONCLUSIONS

Informatics-based methodologies, like those presented here, may play a role in better understanding those individuals who suffer from opioid dependency and overdose, and may lead to future research and interventions that could successfully prevent morbidity and mortality associated with this epidemic.

Risk of excessive sleepiness in sleep restriction therapy and cognitive behavioral therapy for insomnia: a randomized controlled trial

STUDY OBJECTIVES

Sleep restriction therapy (SRT) has been shown to be comparably effective relative to cognitive behavioral therapy for insomnia (CBT-I), but with lower requirements for patient contact. As such, SRT appears to be a viable alternate treatment for those who cannot complete a full course of CBT-I. However, it is unclear whether SRT-a treatment solely focusing on restricting time in bed-increases risk for sleepiness comparably to CBT-I. The current study tested objective sleepiness as an outcome in a randomized controlled trial comparing SRT, CBT-I, and attention control in a sample of postmenopausal women in whom insomnia was diagnosed according to criteria of the Diagnostic and Statistical Manual of Mental Disorders, Fifth Edition.

METHODS

Single-site, randomized controlled trial. A total of 150 postmenopausal women (56.44 ± 5.64 years) with perimenopausal or postmenopausal onset of Diagnostic and Statistical Manual of Mental Disorders, Fifth Edition insomnia disorder were randomized to 3 treatment conditions: sleep education control (6 sessions); SRT (2 sessions with interim phone contact); and CBT-I (6 sessions). Blinded assessments were performed at pretreatment and posttreatment. Risk of excessive sleepiness was evaluated using a symmetry analysis of sleepiness measured through the Multiple Sleep Latency Test (MSLT).

RESULTS

The odds ratios (ORs) of being excessively sleepy versus nonsleepy were not different than 1.0 for both SRT (OR = 0.94, 95% confidence interval [0.13-6.96]) and CBT-I (OR = 0.62, 95% confidence interval [0.09-4.46]), indicating that the odds of becoming excessively sleepy following treatment was not different from the odds of being nonsleepy. This suggests that excessive sleepiness is not of unique concern following SRT relative to CBT-I or sleep education.

CONCLUSIONS

SRT appears to have a comparable risk profile for excessive sleepiness as CBT-I, and thus may be considered a safe alternative to CBT-I. Future research should characterize objective measures of excessive sleepiness immediately following sleep restriction.

CLINICAL TRAIL REGISTRATION

Registry: ClinicalTrials.gov; Name: Behavioral Treatment of Menopausal Insomnia; Sleep and Daytime Outcomes; Identifier: NCT01933295.

An Implementation and Visualization of the Tree-Based Scan Statistic for Safety Event Monitoring in Longitudinal Electronic Health Data

INTRODUCTION

Longitudinal electronic healthcare data hold great potential for drug safety surveillance. The tree-based scan statistic (TBSS), as implemented by the TreeScan^® software, allows for hypothesis-free signal detection in longitudinal data by grouping safety events according to branching, hierarchical data coding systems, and then identifying signals of disproportionate recording (SDRs) among the singular events or event groups.

OBJECTIVE

The objective of this analysis was to identify and visualize SDRs with the TBSS in historical data from patients using two antifungal drugs, itraconazole or terbinafine. By examining patients who used either itraconazole or terbinafine, we provide a conceptual replication of a previous TBSS analyses by varying methodological choices and using a data source that had not been previously used with the TBSS, i.e., the Optum Clinformatics™ claims database. With this analysis, we aimed to test a parsimonious design that could be the basis of a broadly applicable method for multiple drug and safety event pairs.

METHODS

The TBSS analysis was used to examine incident events and any itraconazole or terbinafine use among US-based patients from 2002 through 2007. Event frequencies before and after the first day of drug exposure were compared over 14- and 56-day periods of observation in a Bernoulli model with a self-controlled design. Safety events were classified into a hierarchical tree structure using the Clinical Classifications Software (CCS) which mapped International Classification of Diseases, 9th Revision (ICD-9) codes to 879 diagnostic groups. Using the TBSS, the log likelihood ratio of observed versus expected events in all groups along the CCS hierarchy were compared, and groups of events that occurred at disproportionally high frequencies were identified as potential SDRs; p-values for the potential SDRs were estimated with Monte-Carlo permutation based methods. Output from TreeScan^® was visualized and plotted as a network which followed the CCS tree structure.

RESULTS

Terbinafine use (n = 223,968) was associated with SDRs for diseases of the circulatory system (14- and 56-day p = 0.001) and heart (14-day p = 0.026 and 56-day p = 0.001) as well as coronary atherosclerosis and other heart disease (14-day p = 0.003 and 56-day p = 0.004). For itraconazole use (n = 36,025), the TBSS identified SDRs for coronary atherosclerosis and other heart disease (p = 0.002) and complications of an implanted or grafted device (14-day p = 0.001 and 56-day p < 0.05). Use of both drugs was associated with SDRs for diseases of the digestive system at 14 days (p < 0.05) and this SDR had been observed among terbinafine users in a previous TBSS analysis with a different data source. The TreeScan^® visualization facilitated the identification of the atherosclerosis and other heart disease SDRs as well as highlighting the consistency of the SDR for diseases of the digestive system across drugs and data sources.

CONCLUSION

With the TBSS, we identified potential SDRs related to the circulatory system that may reflect the cardiac risk that was described in the itraconazole product label. SDRs for diseases of the digestive system among terbinafine users were also reported in a previous signal detection analysis, although other SDRs from the previous publications were not replicated. The TBSS visualizations aided in the understanding and interpretation of the TBSS output, including the comparisons to the previous publications. In this conceptual replication, differences in the results observed in our analysis and the previous analyses could be attributable to variation in modeling and design choices as well as factors that were intrinsic to the underlying data sources. The broad consistency, but far from perfect concordance, of our results with the known safety profile of these antifungals including the risks from the itraconazole product label supports the rationale for continued investigations of signal detection methods across differing data sources and populations.

Signal Detection of Potentially Drug-Induced Liver Injury in Children Using Electronic Health Records

Background: This study proposes a quantitative 2-stage procedure to detect potential drug-induced liver injury (DILI) signals in pediatric inpatients using an data warehouse of electronic health records (EHRs). Methods: Eight years of medical data from a constructed database were used. A two-stage procedure was adopted: (i) stage 1: the drugs suspected of inducing DILI were selected and (ii) stage 2: the associations between the drugs and DILI were identified in a retrospective cohort study. Results: 1,196 drugs were filtered initially and 12 drugs were further potentially identified as suspect drugs inducing DILI. Eleven drugs (fluconazole, omeprazole, sulfamethoxazole, vancomycin, granulocyte colony-stimulating factor (G-CSF), acetaminophen, nifedipine, fusidine, oseltamivir, nystatin and meropenem) were showed to be associated with DILI. Of these, two drugs, nystatin [odds ratio[OR]=1.39, 95%CI:1.10-1.75] and G-CSF (OR = 1.91, 95%CI:1.55-2.35), were found to be new potential signals in adults and children. Three drugs [nifedipine [OR = 1.77, 95%CI:1.26-2.46], fusidine [OR = 1.43, 95%CI:1.08-1.86], and oseltamivi r [OR = 1.64, 95%CI:1.23-2.18]] were demonstrated to be new signals in pediatrics. The other drug-DILI associations had been confirmed in previous studies. Conclusions: A quantitative algorithm to detect potential signals of DILI has been described. Our work promotes the application of EHR data in pharmacovigilance and provides candidate drugs for further causality assessment studies.

Early signal detection of adverse events following influenza vaccination using proportional reporting ratio, Victoria, Australia

INTRODUCTION

Timely adverse event following immunisation (AEFI) signal event detection is essential to minimise further vaccinees receiving unsafe vaccines. We explored the proportional reporting ratio (PRR) ability to detect two known signal events with influenza vaccines with the aim of providing a model for prospective routine signal detection and improving vaccine safety surveillance in Australia.

METHODS

Passive AEFI surveillance reports from 2008-2017 relating to influenza vaccines were accessed from the Australian SAEFVIC (Victoria) database. Proportional reporting ratios were calculated for two vaccine-event categories; fever and allergic AEFI. Signal detection sensitivity for two known signal events were determined using weekly data; cumulative data by individual year and; cumulative for all previous years. Signal event thresholds of PRR ≥2 and Chi-square ≥4 were applied.

RESULTS

PRR provided sensitive signal detection when calculated cumulatively by individual year or by all previous years. Known signal events were detected 15 and 11 days earlier than traditional methods used at the time of the actual events.

CONCLUSION

Utilising a single jurisdiction's data, PRR improved vaccine pharmacovigilance and showed the potential to detect important safety signals much earlier than previously. It has potential to maximise immunisation safety in Australia. This study progresses the necessary work to establish national cohesion for passive surveillance signal detection and strengthen routine Australian vaccine pharmacovigilance.

Suspected Adverse Drug Reactions Related to Breast Cancer Chemotherapy: Disproportionality Analysis of the Brazilian Spontaneous Reporting System

Spontaneous reporting systems may generate a large volume of information in real world conditions with a relatively low cost. Disproportionality measures are useful to indicate and quantify unexpected safety issues associated with a given drug-event pair (signals of disproportionality), based upon differences compared to the background reporting frequency. This cross-sectional study (2008 to 2013) aimed to analyse the feasibility of detecting such signals in the Brazilian Pharmacovigilance Database comprising suspected adverse drug reactions related to the use of doxorubicin, cyclophosphamide, carboplatin, trastuzumab, docetaxel, and paclitaxel for breast cancer chemotherapy. We first accessed overall database features (patient information and suspected adverse drug reactions) and further conducted a disproportionality analysis based on Reporting Odds Ratios with a confidence interval of 95% in order to identify possible signals of disproportionate reporting, only among serious suspected adverse drug reactions. Of all data reports of adverse reactions (n = 2603), 83% were classified as serious, with the highest prevalence with docetaxel (78.1%). The final analysis was performed using 1,309 reports with 3,139 drug-reaction pairs. The following signals of disproportionate reporting, some rare or not mentioned on labels, were observed: tachypnea with docetaxel; bronchospasm, syncope, cyanosis, and anaphylactic reaction with paclitaxel; and anaphylactic shock with trastuzumab. Structured management of spontaneous adverse drug reaction reporting is essential for monitoring the safe use of drugs and detecting early safety signals. Disproportionality signal analysis represents a viable and applicable strategy for oncology signal screening in the Brazilian Pharmacovigilance Database.

Big data in IBD: a look into the future

Big data methodologies, made possible with the increasing generation and availability of digital data and enhanced analytical capabilities, have produced new insights to improve outcomes in many disciplines. Application of big data in the health-care sector is in its early stages, although the potential for leveraging underutilized data to gain a better understanding of disease and improve quality of care is enormous. Owing to the intrinsic characteristics of inflammatory bowel disease (IBD) and the management dilemmas that it imposes, the implementation of big data research strategies not only can complement current research efforts but also could represent the only way to disentangle the complexity of the disease. In this Review, we explore important potential applications of big data in IBD research, including predictive models of disease course and response to therapy, characterization of disease heterogeneity, drug safety and development, precision medicine and cost-effectiveness of care. We also discuss the strengths and limitations of potential data sources that big data analytics could draw from in the field of IBD, including electronic health records, clinical trial data, e-health applications and genomic, transcriptomic, proteomic, metabolomic and microbiomic data.

Pharmacoepidemiology in older people: Purposes and future directions

Knowledge of the benefit/risk ratio of drugs in older adults is essential to optimise medication use. While randomised controlled trials are fundamental to the process of drug development and bringing new drugs to the market, they often exclude older adults, especially those suffering from frailty, multimorbidity and/or receiving polypharmacy. Therefore, it is generally unknown whether the benefits and harms of drugs established through pre-marketing clinical trials are translatable to the real-word population of older adults. Pharmacoepidemiology can provide real-world data on drug utilisation and drug effects in older people with multiple comorbidities and polypharmacy and can greatly contribute towards the goal of high quality use of drugs and well-being in older adults. A wide variety of pharmacoepidemiology studies can be used and exciting progress is being made with the use of novel and advanced statistical methods to improve the robustness of data. Coordinated and strategic initiatives are required internationally in order for this field to reach its full potential of optimising drug use in older adults so as to improve health care outcomes.

Time Series Disturbance Detection for Hypothesis-Free Signal Detection in Longitudinal Observational Databases

INTRODUCTION

Signal detection remains a cornerstone activity of pharmacovigilance. Routine quantitative signal detection primarily focuses on screening of spontaneous reports. In striving to enhance quantitative signal detection capability further, other data streams are being considered for their potential contribution as sources of emerging signals, one of which is longitudinal observational databases, including electronic medical record (EMR) and transactional insurance claims databases. Quantitative signal detection on such databases is a nascent field-with published methods being primarily based either on individual metrics, which may not effectively represent the complexity of the longitudinal records and their necessary variation for analysis for drug-outcome pairs, or on visualization discovery approaches leveraging multiple aspects of the records, which are not particularly tractable to high-throughput hypothesis-free signal detection. One extensively tested example of the latter is chronographs.

METHODS

We apply a disturbance detection algorithm to chronographs using UK EMR The Health Improvement Network (THIN) data. The algorithm utilizes autoregressive integrated moving average (ARIMA)-based time series methodology designed to find disturbances that occur outside the normal pattern trends of the ARIMA model for the chronograph. Chronographs currently highlight drug-event pairs as potentially worthy of further clinical assessment, via filter-based increases in disproportionality scores from before to after the index drug exposure, tested across a range of case and control windows.

RESULTS

We replicate the findings on six exemplar chronographs from a previous publication, and show how disturbances can be effectively detected across this set of pairs. Further, 692 disturbances were detected in analysis of all 384 individual READ code outcomes ever recorded 50 or more times for patients prescribed sibutramine. The disturbances are algorithmically further broken into subsets of clinical interest.

CONCLUSION

Overall, the disturbance algorithm approach shows promising capacity for detecting outliers, and shows tractability of the algorithmic approach for large-scale screening. The method offers an array of pattern types for detection and clinical review.

Assessment of Medication Safety Using Only Dispensing Data

PURPOSE OF REVIEW

The purpose of this review is to provide an overview of the published studies that have been used to generate evidence on the safety of medicine use when only medication dispensing data are available.

RECENT FINDINGS

Medication dispensing databases are increasingly available for research on large populations, particularly in countries that provide universal coverage for medicines. These data are often used for drug utilisation studies to identify inappropriate medicine use at the population level that may be associated with known safety issues. Lack of coded diagnoses, to identify outcomes, and lack of data on confounders can limit use of these data in practice for medication safety assessment. To overcome these issues, studies have exploited the fact that symptoms of adverse effects of medications can be treated with other medications, for example antidepressants to treat depression or oxybutynin to treat urinary incontinence. The challenge of unmeasured confounding has been addressed by implementing self-controlled study designs that use within-person comparisons and provide inherent control for confounding. Prescription sequence symmetry analysis (SSA) is a within-person study design that has been demonstrated as a useful tool for safety signal generation in dispensing data.

SUMMARY

Using medicine initiation as a proxy for the development of adverse events can help to generate evidence of the safety of medicines when only medication dispensing data are available. Careful consideration, however, should be given to the sensitivity and specificity of the proxy medicine for the adverse event and potential for time-varying confounding due to trends in medicine utilisation. Data-mining approaches using dispensing data have the potential to improve safety assessments; however, the challenge of unmeasured confounding with these methods remains to be investigated.

Nurse-led medicines' monitoring in care homes study protocol: a process evaluation of the impact and sustainability of the adverse drug reaction (ADRe) profile for mental health medicines

INTRODUCTION

Improved medicines' management could lead to real and sustainable improvements to the care of older adults. The overuse of mental health medicines has featured in many reports, and insufficient patient monitoring has been identified as an important cause of medicine-related harms. Nurse-led monitoring using the structured adverse drug reaction (ADRe) profile identifies and addresses the adverse effects of mental health medicines. Our study investigates clinical impact and what is needed to sustain utilisation in routine practice in care homes.

METHODS AND ANALYSIS

This process evaluation will use interviews and observations with the participants of all five homes involved in earlier research, and five newly recruited homes caring for people prescribed mental health medicines. The ADRe profile is implemented by nurses, within existing resources, to check for signs and symptoms of ADRs, initiate amelioration and share findings with pharmacists and prescribers for medication review. Outcome measures are the numbers and nature of problems addressed and understanding of changes needed to optimise clinical gain and sustain implementation. Data will be collected by 30 observations and 30 semistructured interviews. Clinical gains will be described and narrated. Interview analysis will be based on the constant comparative method.

ETHICS AND DISSEMINATION

Ethical approval was conferred by the National Health Service Wales Research Ethics Committee. If the ADRe profile can be sustained in routine practice, it has potential to (1) improve the lives of patients, for example, by reducing pain and sedation, and (2) assist in early identification of problems caused by ADRs. Therefore, in addition to peer-reviewed publications and conferences, we shall communicate our findings to healthcare professionals, policy-makers and sector regulators.

TRIAL REGISTRATION NUMBER

NCT03110471.

Supervised signal detection for adverse drug reactions in medication dispensing data

MOTIVATION

Adverse drug reactions (ADRs) are one of the leading causes of morbidity and mortality and thus should be detected early to reduce consequences on health outcomes. Medication dispensing data are comprehensive sources of information about medicine uses that can be utilized for the signal detection of ADRs. Sequence symmetry analysis (SSA) has been employed in previous studies to detect signals of ADRs from medication dispensing data, but it has a moderate sensitivity and tends to miss some ADR signals. With successful applications in various areas, supervised machine learning (SML) methods are promising in detecting ADR signals. Gold standards of known ADRs and non- ADRs from previous studies create opportunities to take into account additional domain knowledge to improve ADR signal detection with SML.

OBJECTIVE

We assess the utility of SML as a signal detection tool for ADRs in medication dispensing data with the consideration of domain knowledge from DrugBank and MedDRA. We compare the best performing SML method with SSA.

METHODS

We model the ADR signal detection problem as a supervised machine learning problem by linking medication dispensing data with domain knowledge bases. Suspected ADR signals are extracted from the Australian Pharmaceutical Benefit Scheme (PBS) medication dispensing data from 2013 to 2016. We construct predictive features for each signal candidate based on its occurrences in medication dispensing data as well as its pharmacological properties. Pharmaceutical knowledge bases including DrugBank and MedDRA are employed to provide pharmacological features for a signal candidate. Given a gold standard of known ADRs and non-ADRs, SML learns to differentiate between known ADRs and non-ADRs based on their combined predictive features from linked sources, and then predicts whether a new case is a potential ADR signal.

RESULTS

We evaluate the performance of six widely used SML methods with two gold standards of known ADRs and non-ADRs from previous studies. On average, gradient boosting classifier achieves the sensitivity of 77%, specificity of 81%, positive predictive value of 76%, negative predictive value of 82%, area under precision-recall curve of 81%, and area under receiver operating characteristic curve of 82%, most of which are higher than in other SML methods. In particular, gradient boosting classifier has 21% higher sensitivity than and comparable specificity with SSA. Furthermore, gradient boosting classifier detects 10% more unknown potential ADR signals than SSA.

CONCLUSIONS

Our study demonstrates that gradient boosting classifier is a promising supervised signal detection tool for ADRs in medication dispensing data to complement SSA.