Pharmacoepidemiology Research Articles

Reports of myocarditis and pericarditis following mRNA COVID-19 vaccination: a systematic review of spontaneously reported data from the UK, Europe and the USA and of the scientific literature

ObjectivesTo combine spontaneously reported data from multiple countries to estimate reporting rate, and better understand risk factors for myocarditis and pericarditis following COVID-19 messenger RNA (mRNA) vaccines.DesignSystematic review of spontaneously...

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.

Rationale and performances of a data-driven method for computing the duration of pharmacological prescriptions using secondary data sources

The assessment of the duration of pharmacological prescriptions is an important phase in pharmacoepidemiologic studies aiming to investigate persistence, effectiveness or safety of treatments. The Sessa Empirical Estimator (SEE) is a new data-driven method which uses k-means algorithm for computing the duration of pharmacological prescriptions in secondary data sources when this information is missing or incomplete. The SEE was used to compute durations of exposure to pharmacological treatments where simulated and real-world data were used to assess its properties comparing the exposure status extrapolated with the method with the “true” exposure status available in the simulated and real-world data. Finally, the SEE was also compared to a Researcher-Defined Duration (RDD) method. When using simulated data, the SEE showed accuracy of 96% and sensitivity of 96%, while when using real-world data, the method showed sensitivity ranging from 78.0 (nortriptyline) to 95.1% (propafenone). When compared to the RDD, the method had a lower median sensitivity of 2.29% (interquartile range 1.21–4.11%). The SEE showed good properties and may represent a promising tool to assess exposure status when information on treatment duration is not available.

Mapping Outcomes and Registries Used in Current Danish Pharmacoepidemiological Research.

PurposeThere is an increasing need for national and international pharmacoepidemiological studies based on high-quality real-world data of which the Danish registries are a valuable source. In lack of a complete overview of which data are used to assess real-world drug safety and effectiveness outcomes, we aimed to map the outcomes, data sources, and the reporting of outcome quality in recent pharmacoepidemiological studies.MethodsWe conducted a systematic mapping review of pharmacoepidemiological studies based on Danish registries investigating drug safety and/or effectiveness, published in the period 2018–2019, identified in PubMed and Scopus. Extraction included: Anatomical Therapeutic Chemical level 2 code for drug exposures, outcomes, outcome data sources, and quality of outcomes.ResultsOf the 210 included studies, 96% used outcomes categorized as Clinical, 4% utilized outcomes categorized as Society-related, 5% used outcomes categorized as Healthcare cost and utilization, and 3% of the studies applied outcomes categorized as Patient-reported in which the percentages are not mutually exclusive. Diagnosis (66%) and Mortality (38%) were the two most utilized subcategories among those categorized as Clinical outcomes. Danish Health Data Authority and Statistics Denmark registries were the most reported outcome data sources (90%). Ninety-six studies (46%) reported one or more quality parameters related to their outcomes of interest with accuracy/validity being the most reported parameter (22%).ConclusionThe Danish registries support a wide range of outcomes. Across therapeutic areas, most studies investigate traditional clinical outcomes of disease and mortality based on data from a small number of available registries. In contrast, clinical and biochemical databases, despite potentially offering outcomes with high responsiveness, and the high-quality social and healthcare cost registries were rarely used as outcome data sources.

Immortal time bias for life-long conditions in retrospective observational studies using electronic health records

BackgroundImmortal time bias is common in observational studies but is typically described for pharmacoepidemiology studies where there is a delay between cohort entry and treatment initiation.MethodsThis study used the Clinical Practice Research Datalink (CPRD) and linked national mortality data in England from 2000 to 2019 to investigate immortal time bias for a specific life-long condition, intellectual disability. Life expectancy (Chiang’s abridged life table approach) was compared for 33,867 exposed and 980,586 unexposed individuals aged 10+ years using five methods: (1) treating immortal time as observation time; (2) excluding time before date of first exposure diagnosis; (3) matching cohort entry to first exposure diagnosis; (4) excluding time before proxy date of inputting first exposure diagnosis (by the physician); and (5) treating exposure as a time-dependent measure.ResultsWhen not considered in the design or analysis (Method 1), immortal time bias led to disproportionately high life expectancy for the exposed population during the first calendar period (additional years expected to live: 2000–2004: 65.6 [95% CI: 63.6,67.6]) compared to the later calendar periods (2005–2009: 59.9 [58.8,60.9]; 2010–2014: 58.0 [57.1,58.9]; 2015–2019: 58.2 [56.8,59.7]). Date of entry of diagnosis (Method 4) was unreliable in this CPRD cohort. The final methods (Method 2, 3 and 5) appeared to solve the main theoretical problem but residual bias may have remained.ConclusionsWe conclude that immortal time bias is a significant issue for studies of life-long conditions that use electronic health record data and requires careful consideration of how clinical diagnoses are entered onto electronic health record systems.

Proton Pump Inhibitors and Cancer: Current State of Play.

Background: Proton pump inhibitors (PPIs) are one of the most widely used drugs worldwide and are overprescribed in patients with cancer; there is increasing evidence of their effects on cancer development and survival. The objective of this narrative review is to comprehensively identify cancer medications that have clinically meaningful drug–drug interactions (DDIs) with PPIs, including loss of efficacy or adverse effects, and to explore the association between PPIs and cancer. Methods: A PubMed search of English language studies published from 1 January 2016, to 1 June 2021 was conducted. The search terms included “proton pump inhibitors,” “cancer,” “chemotherapy,” “immunotherapy,” “hormonotherapies,” “targeted therapies,” “tyrosine kinase inhibitors,” and “gut microbiome”. Recent and relevant clinical trials, meta-analyses, and reviews were included. Results: PPIs may have pro-tumor activity by increasing plasma gastrin levels or anti-tumor activity by inhibiting V-ATPases. However, their impact on cancer survival remains unclear. PPIs may decrease the efficacy of some antineoplastic agents through direct DDIs (e.g., some tyrosine kinase inhibitors, capecitabine, irinotecan, methotrexate). More complex DDIs seem to exist for immunotherapies with indirect interactions through the microbiome. PPIs worsen hypomagnesemia, bone loss, iron, and vitamin B12 deficiencies but may have a protective effect on the renal system. Discussion/Conclusions: PPIs may interact with the cancer microbiome and the efficacy of various antineoplastic agents, although only a few DDIs involving PPIs are clinically significant. Further pharmaco-epidemiological studies are warranted, but physicians should be aware of the potential consequences of PPI use, which should be dose appropriate and prescribed according to guidelines.

ANTIBACDUS-PAN: Antibacterial Utilization among Adult Patients Before and During COVID-19 Pandemic within 12-Months Period: A Tertiary Hospital Pharmacoepidemiology Study.

Irrational use of antibacterials is a concern during the COVID-19 pandemic. Hospital pharmacoepidemiology studies are important for evaluating the rational use of medicines, especially antibacterials, during pandemics. Defined daily doses (DDD) and drug utilization 90% (DU90%) are established methods for the evaluation of drug utilization. We aimed to evaluate antibacterial utilization in a tertiary hospital setting at Koç University Hospital (KUH). This cross-sectional, descriptive study was retrospectively conducted with data extracted from KUH Inpatient Electronic Order System (CP) and was carried out for a period of one year. Antibacterial utilization of adult (aged ≥ 18 years) inpatients, who were prescribed at least one type of systemic antibacterial (ATC code J01), was evaluated using the recommended parameter DDD/100 admission and compared between 6 months before COVID-19 and during COVID-19 periods. March 11, 2020, the very first COVID-19 diagnosed case in Turkey, was set as the cutoff date of the 6-month period for the selection of the compared antibacterials using the DU90% method. Finally, 3280 of 5942 and 2605 of 4942 prescriptions for pre-COVID-19 and COVID-19 periods were included, respectively. Antibacterial utilization according to DDD/100 admissions increased from 193.96 to 201.26 DDD/100 admissions after the initiation of COVID-19 pandemic. The most utilized antibacterials were piperacillin and enzyme inhibitors in pre-COVID-19 period, whereas meropenem was utilized the most during COVID-19 period. Azithromycin utilization increased by 656.24%, whereas clarithromycin utilization decreased by 52.12%. Antibacterials were utilized most in general surgery department, with an increase of 17.57%. There is an increase in antibacterial utilization in KUH during COVID-19 pandemic, especially reserved antibacterials, which is a concern for antibacterial resistance.

Colistin-associated Stevens-Johnson syndrome and toxic epidermal necrolysis reactions: a retrospective case-non-case pharmacovigilance study

ABSTRACT Background Stevens-Johnson syndrome (SJS) and toxic epidermal necrolysis (TEN) are life-threatening skin reactions. Colistin is a last resort antibiotic with a historically poor safety profile. The association between colistin and SJS/TEN has not been previously quantified. Research design and methods We identified colistin and SJS/TEN adverse event reports from the Food and Drug Administration Adverse Event Reporting System (FAERS) and calculated effect estimates using OpenEpi. Results From January 2013 through March 2021, 964 adverse events were reported for colistin. Colistin was listed as a secondary suspect drug in 13 SJS/TEN adverse event reports (1.3%), with a reporting odds ratio of 29.6 (95% confidence interval [CI] 17.1–51.1), and proportional reporting ratio of 29.2 (95% CI 17.0–50.2). Limitations of any FAERS study include the voluntary nature of reporting, unclear causal relationship between drug and adverse reaction, underreporting, and wide confidence intervals for rare adverse events like SJS/TEN. Conclusions Colistin was not the primary suspect drug in any SJS/TEN adverse event reports. We did identify a statistically significant safety signal for SJS/TEN with colistin as a secondary suspect drug. SJS/TEN is not currently included in the colistin product label. This association should be further explored in other pharmacoepidemiologic drug safety studies.

Validity of identification algorithms combining diagnostic codes with other measures for acute ischemic stroke in MID-NET®.

We aimed to develop a reliable identification algorithm combining diagnostic codes with several treatment factors for inpatients with acute ischemic stroke (AIS) to conduct pharmacoepidemiological studies using the administrative database MID-NET® in Japan. We validated 11 identification algorithms based on 56 different diagnostic codes (International Classification of Diseases, Tenth Revision; ICD-10) using Diagnosis Procedure Combination (DPC) data combined with information on AIS therapeutic procedures added as "AND" condition or "OR" condition. The target population for this study was 366 randomly selected hospitalized patients with possible cases of AIS, defined as relevant ICD-10 codes and diagnostic imaging and prescription or surgical procedure, in three institutions between April 1, 2015 and March 31, 2017. We determined the positive predictive values (PPVs) of these identification algorithms based on comparisons with a gold standard consisting of chart reviews by experienced specialist physicians. Additionally, the sensitivities of them among 166 patients with the possible cases of AIS at a single institution were evaluated. The PPVs were 0.618 (95% confidence interval [CI]: 0.566-0.667) to 0.909 (95% CI: 0.708-0.989) and progressively increased with adding or limiting information on AIS therapeutic procedures as "AND" condition in the identification algorithms. The PPVs for identification algorithms based on diagnostic codes I63.x were >0.8. However, the sensitivities progressively decreased to a maximum of ~0.2 after adding information on AIS therapeutic procedures as "AND" condition. The identification algorithms based on the combination of appropriate ICD-10 diagnostic codes in DPC data and other AIS treatment factors may be useful to studies for AIS at a national level using MID-NET®.

Melatonin dispensing and polypharmacy rates for New Zealand children with autism or attention deficit hyperactivity disorders: A nationwide pharmacoepidemiological study

BackgroundChronic insomnia is common in children with autism and Attention Deficit Hyperactivity Disorder (ADHD). Melatonin is often used to treat childhood insomnia. However, it may interact with other medications being used to manage other symptoms. This pharmacoepidemiological study examined the rates of general and psychotropic polypharmacy among children with autism and/or ADHD, stratified by melatonin dispensing. The impact of sociodemographic and child characteristics on such dispensing was also examined. MethodLinked national health and pharmaceutical administrative data for children aged 0–18 years in 2019 was utilized. Overall and melatonin dispensing stratified polypharmacy rates were calculated. Ordinal logistic regression models were employed to compare groups and adjust for confounders. ResultsData were acquired for 10,209 children with autism (18.5% were dispensed melatonin), 5970 with ADHD (22.3% were dispensed melatonin), 2064 with autism and ADHD (29.9% were dispensed melatonin), and 1156,296 without a diagnosis of autism or ADHD (a control group; 0.5% dispensed melatonin). Relative to controls, rates of melatonin dispensing and polypharmacy were higher in children with autism and ADHD, and highest among those with both conditions. Children dispensed melatonin experienced the greatest rates of polypharmacy, especially if they had both autism and ADHD. ConclusionsChildren with autism and ADHD experience significant medication burden and potentially adverse interactions between psychotropic and sleep-related medication, raising important questions regarding their clinical care.

Exposure to High Precariousness Prevalence Negatively Impacts Drug Prescriptions of General Practitioners to Precarious and Non-Precarious Populations: A Retrospective Pharmaco-Epidemiological Study.

(1) Background: Precarious patients are more difficult to care for due to low literacy rates and poor adherence to treatment and hospitalization. These difficulties have detrimental effects on general practitioners (GPs), deteriorating medical communication, advice, diagnoses, and drug prescriptions. To better understand how precariousness affects primary care, we tested whether, among GPs, exposure to high precariousness prevalence more severely impacts drug prescriptions to precarious and non-precarious populations compared to low precariousness prevalence. Materials and methods: This pharmaco-epidemiological study, using linear regression analyses, compared the defined daily dose of 20 drugs prescribed by GPs to precarious and non-precarious patients in four French regions with low and high precariousness prevalence in 2015. (2) Findings: Exposure to high precariousness prevalence significantly impacted the prescriptions of nine medications to precarious patients and two medications to non-precarious patients, and distributed into three interaction patterns. (3) Interpretation: The selective over-prescription of drugs with easy intake modalities to precarious patients probably reflects GPs’ attempts to compensate for poor patient compliance. In contrast, the under-prescription of drugs targeting fungal infections in precarious populations and diabetes and cardiovascular diseases in non-precarious populations was seemingly due to a breakdown of empathy and professional exhaustion, causing medical neglect.

The rationality of recommendations of drugs for patients with cephalgia in simulated situations in the pharmacy

Relevance. According to statistics of the World Health Organization, at least half of the world’s adult population experienced headaches at least once a year. Patients suffering from headaches do not always have the opportunity to visit a doctor, therefore, an important role in helping these patients is assigned to pharmacists. Purpose. Based on the pharmacoepidemiological study to identify the features of providing pharmaceutical care to patients with headache in pharmacies in the issue of rationality of recommendations of pharmaceutical workers. Materials and methods. A pharmacoepidemiological study was conducted based on a survey of 153 pharmaceutical workers (Saratov) using a specially developed questionnaire that includes free-form questions with a simulated situation. The data was processed using Microsoft Excel 2010 and STATISTIKA 6.0. Results Of the groups recommended to visitors, INNs in the group «Nonsteroidal anti-inflammatory and antirheumatic drugs [M01A]» account for 45.0 % of all INNs, and the group «Analgesics [N02]» - 55.0 %. Ibuprofen (62.8 %) and the combination of drotaverine + caffeine + naproxen + paracetamol + phenylephrine (41.2 %) were noted as the most frequently recommended. A visit to a doctor was recommended by the majority of respondents (52.3 %, n=153) in the case of a patient with a headache accompanied by nausea and vomiting. Ibuprofen (90.8 %) or paracetamol (45.8 %), moreover in the form of suspension or suppositories, are the preferred recommendations for headache in children under 3 years of age (n=131). Paracetamol in combinations (69.2 %) prevails as the first recommendation for headache associated with acute respiratory viral infections (n??=146). Non-selective NSAIDs for a patient with gastrointestinal complications could be recommended by 43.4 % of specialists (n=113). The NSAID recommendation is the most frequent (51.8 %, n=110) in the case of an 80-year-old patient without concomitant diseases. Paracetamol (50.0 %) or ibuprofen (34.7 %) would be more often recommended by pharmaceutical workers for a pregnant woman with headache (n=98). In case of headache after trauma, 47.1 % recommend a visit to a doctor (n=153). Conclusion. The frequent recommendation of combined analgesics and their association with the development of drug-induced headaches indicates the need for careful recommendation of these drugs. Not always correct recommendations indicate the need to increase the level of professional knowledge, study clinical recommendations for the therapy and diagnosis of cephalgia, which can make it possible to make a rational choice of the drug, initially assume that the visitor has conditions that require a visit to a doctor, and ultimately improve the quality of pharmaceutical care.

Transparency of high-dimensional propensity score analyses: Guidance for diagnostics and reporting.

PurposeThe high‐dimensional propensity score (HDPS) is a semi‐automated procedure for confounder identification, prioritisation and adjustment in large healthcare databases that requires investigators to specify data dimensions, prioritisation strategy and tuning parameters. In practice, reporting of these decisions is inconsistent and this can undermine the transparency, and reproducibility of results obtained. We illustrate reporting tools, graphical displays and sensitivity analyses to increase transparency and facilitate evaluation of the robustness of analyses involving HDPS.MethodsUsing a study from the UK Clinical Practice Research Datalink that implemented HDPS we demonstrate the application of the proposed recommendations.ResultsWe identify seven considerations surrounding the implementation of HDPS, such as the identification of data dimensions, method for code prioritisation and number of variables selected. Graphical diagnostic tools include assessing the balance of key confounders before and after adjusting for empirically selected HDPS covariates and the identification of potentially influential covariates. Sensitivity analyses include varying the number of covariates selected and assessing the impact of covariates behaving empirically as instrumental variables. In our example, results were robust to both the number of covariates selected and the inclusion of potentially influential covariates. Furthermore, our HDPS models achieved good balance in key confounders.ConclusionsThe data‐adaptive approach of HDPS and the resulting benefits have led to its popularity as a method for confounder adjustment in pharmacoepidemiological studies. Reporting of HDPS analyses in practice may be improved by the considerations and tools proposed here to increase the transparency and reproducibility of study results.

Patterns of opioid analgesic prescribing in cancer outpatients during the last year of life in France: A pharmacoepidemiological cohort study based on the French health insurance database

Cancer pain management with adequate analgesics for cancer outpatients can be particularly challenging. This representative retrospective cohort study aimed to investigate the prevalence and timing of weak and strong opioid analgesic prescriptions in cancer outpatients during their last year of life, with a focus on factors associated to potential late strong opioid initiation. Factors associated with late strong opioid initiation were investigated through multivariate logistic regression analyses stratified by place of death. A retrospective cohort of cancer outpatients, who died between 2014 and 2016, was identified from the general sample of beneficiaries. Among N=4704 cancer patients (median age 76 years, 42.7% women), 3002 (63.8%) were prescribed and dispensed ≥1 weak or strong opioid analgesic during their last year of life; of whom, 2458 (52.3%) received ≥1 weak opioid analgesic (tramadol as single-ingredient accounting for 25.9%) and 1733 (36.8%) ≥1 strong opioid analgesic dispensation (fentanyl 21.6%). Median interval between the first prescription for any strong opioid and death was 18 weeks (interquartile range: 8-38), and for weak opioids 33 weeks (interquartile range: 20-47). Among weak opioid users, 1229 (50.0%) patients had received ≥1 weak opioid analgesic dispensation during the year n-2 before death. Among strong opioid users, 986 (56.9%) patients had received ≥1 weak opioid analgesic dispensation during the year n-2 before death and 381 (21.9%) patients ≥1 strong opioid analgesic dispensation. Patients with an outpatient death were more likely to have a late strong opioid initiation compared to patients with an inpatient death. Late strong opioid initiation (<18 weeks before death) was significantly associated with a lower number of hospitalization days and prior weak opioid exposure for patients with an inpatient death and, with older age, social, prior weak opioid exposure, and a prescription initiation by general practitioner for patients with an outpatient death. Our gained knowledge of opioid prescribing patterns in cancer patients during the last year of life might help to progress opioid analgesic treatment and to improve patient outcomes.

Changes in the use of antiseizure medications in children and adolescents in Norway, 2009–2018

BackgroundThe use of antiseizure medications (ASMs) in the pediatric population is poorly studied. The purpose of this study was to investigate changes in the use of ASMs in children and adolescents compared to adults, and to elucidate safety considerations of certain drugs. MethodIn this population-based pharmacoepidemiological study we used the Norwegian Prescription Database (NorPD), 2009–2018. The use of ASMs is presented as 1-year prevalence per 1000: number of ASM users in a year * 1000 / number of inhabitants that year. Variables included predetermined 5-year age groups, gender, ASMs, diagnosis-specific reimbursement codes, user, and population numbers. Selected ASMs used for specific indications or subgroups included ethosuximide, sulthiame, rufinamide, stiripentol, and clobazam. Gender differences in the use of valproate was examined due to safety considerations in girls/women. ResultsThe total number of ASM users (all indications) for the age groups 0–19 and 20–59 years was 5807 and 47,481 respectively in 2009, and 5906 and 61,447 respectively in 2018. The 1-year prevalence for children/adolescents (0–19 years) using ASMs in epilepsy remained stable from 2008 to 2018, 4.3–4.2/1000 inhabitants, as compared to 8.2–7.6/1000 in adults (20–59 years). Valproate, lamotrigine, and levetiracetam were the three most used ASMs in epilepsy in children/adolescents, similar to adults. The selected ASMs were mainly used in children/ adolescents, accounting for 0.74/1000 in 2018 versus 0.17/1000 in adults. A significant increase was seen for sulthiame (8-fold), ethosuximide (4-fold), clobazam (3-fold), and stiripentol (2-fold). The use of ASMs in non-epilepsy indications was limited and stable (17% in 2018); mainly lamotrigine in psychiatry in adolescents (15–19 years). This finding was in contrast to extensive non-epilepsy use in adults (71% in 2018). ConclusionChanges in the use of ASMs in children/adolescents differ as compared to adults, most notably extensive and increasing use of selected ASMs and limited non-epilepsy. This is an important part of pharmacovigilance and patient safety evaluations.

Systemic quinolones and risk of retinal detachment I: analysis of data from the US FDA Adverse Event Reporting System

Background Quinolones comprise a class of antibiotics that are globally preferred for treating a wide range of bacterial infections due to their potency, broad coverage, favorable pharmacologic profile, and mostly mild to moderate adverse reactions. Spontaneous reports on adverse drug events (ADE) and data from some pharmacoepidemiologic studies have raised concerns regarding quinolones and risk of retinal detachment (RD). This study examined ADE reports submitted to FDA adverse event reporting system (FAERS) for evidence on quinolone-associated RD risk. Research design and methods We identified all reports of RD in FAERS between 2010-2019. We compared ADE signals between quinolones and selected medications that were previously associated with RD, and with reference drugs not known to cause RD. For signal detection, we used two disproportionality techniques: the proportional reporting ratio (PRR) and multi-item gamma Poisson shrinker (MGPS), which are known for their higher sensitivity and specificity for ADE signal detection, respectively. Results Moxifloxacin showed a positive and significant PRR signal for RD [PRR: 2.54 (1.60, 4.04)], and a marginally significant EBGM signal [EBGM: 2.21 (1.41, 3.02)]. Conclusion Moxifloxacin is the only quinolone showing a positive disproportionality signal for RD. Further epidemiologic research is needed to clarify the association between moxifloxacin and the risk of RD.

Research and scholarly methods: Propensity scores.

Propensity score methods are increasingly used as a tool to control for confounding bias in pharmacoepidemiologic studies. The propensity score is a dimension reducing balancing score, creating treatment and reference groups that have comparable distributions of measured covariates. The purpose of this methods review is to provide an overview of the use of propensity score methods, including a summary of important data assumptions, various applications of the propensity score, and how to evaluate covariate balance. This article is intended for pharmacists and researchers who wish to receive an introduction to propensity score methods and be able to engage in high-level discussions on application and reporting.

Kidney and Mortality Outcomes Associated with Ondansetron in Critically Ill Patients

Background: Ondansetron is a preferred anti-emetic in critical care to treat nausea and vomiting, and has historically been considered a largely safe option. A recent pharmacoepidemiology study reported that ondansetron may be associated with an increased risk for acute kidney injury (AKI). Methods: We interrogated the High-Density Intensive Care (HiDenIC-15) database containing intensive care data for 13 hospitals across Western Pennsylvania between Oct 2008-Dec 2014. AKI was defined using the Kidney Disease, Improving Global Outcomes 2012 guidelines. Ondansetron use was considered as receiving any form of ondansetron within 24 h of admission. The subsequent 48 h (hours 25-72 after admission) were analyzed for outcomes. Primary outcome was development of AKI; secondary outcomes included 90-day mortality and time to AKI. Propensity-matched, multivariate logistic regression was applied for both outcomes. Comparator groups were metoclopramide and prochlorperazine using the same exposure criteria. Results:AKI occurred in 965 (5.6%), 12 (3.0%), and 61 (6.5%) patients receiving ondansetron, prochlorperazine, and metoclopramide, respectively. In the adjusted analysis, no anti-emetic was associated with a significant change in the odds of developing AKI. Ondansetron was associated with a 5.48% decrease (CI -6.17--4.79) in death within 90 days of ICU-admission, which was independent of AKI status; an effect not seen with other anti-emetics. Anti-emetic usage was not associated with a change in the time to first AKI. Conclusion:Anti-emetic usage did not alter AKI risk. Ondansetron was associated with a significant decrease in 90-day mortality that was not seen by other anti-emetics, which requires further exploration.

Use of Linked Databases for Improved Confounding Control: Considerations for Potential Selection Bias.

Pharmacoepidemiologic studies are increasingly conducted within linked databases, often to obtain richer confounder data. However, the potential for selection bias is frequently overlooked when linked data is available only for a subset of patients. We highlight the importance of accounting for potential selection bias by evaluating the association between antipsychotics and type 2 diabetes in youths within a claims database linked to a smaller laboratory database. We used inverse probability of treatment weights (IPTW) to control for confounding. In analyses restricted to the linked cohorts, we applied inverse probability of selection weights (IPSW) to create a population representative of the full cohort. We used pooled logistic regression weighted by IPTW only or IPTW and IPSW to estimate treatment effects. Metabolic conditions were more prevalent in linked cohorts compared with the full cohort. Within the full cohort, the confounding-adjusted hazard ratio was 2.26 (95% CI: 2.07, 2.49) comparing initiation of antipsychotics with initiation of control medications. Within the linked cohorts, a different magnitude of association was obtained without adjustment for selection, whereas applying IPSW resulted in point estimates similar to the full cohort's (e.g., an adjusted hazard ratio of 1.63 became 2.12). Linked database studies may generate biased estimates without proper adjustment for potential selection bias.

A potential outcomes approach to defining and estimating gestational age-specific exposure effects during pregnancy.

Many studies seek to evaluate the effects of potentially harmful pregnancy exposures during specific gestational periods. We consider an observational pregnancy cohort where pregnant individuals can initiate medication usage or become exposed to a drug at various times during their pregnancy. An important statistical challenge involves how to define and estimate exposure effects when pregnancy loss or delivery can occur over time. Without proper consideration, the results of standard analysis may be vulnerable to selection bias, immortal time-bias, and time-dependent confounding. In this study, we apply the “target trials” framework of Hernán and Robins in order to define effects based on the counterfactual approach often used in causal inference. This effect is defined relative to a hypothetical randomized trial of timed pregnancy exposures where delivery may precede and thus potentially interrupt exposure initiation. We describe specific implementations of inverse probability weighting, G-computation, and Targeted Maximum Likelihood Estimation to estimate the effects of interest. We demonstrate the performance of all estimators using simulated data and show that a standard implementation of inverse probability weighting is biased. We then apply our proposed methods to a pharmacoepidemiology study to evaluate the potentially time-dependent effect of exposure to inhaled corticosteroids on birthweight in pregnant people with mild asthma.