To weight or not to weight? The effect of selection bias in 3 large electronic health record-linked biobanks and recommendations for practice

OBJECTIVES

To develop recommendations regarding the use of weights to reduce selection bias for commonly performed analyses using electronic health record (EHR)-linked biobank data.

MATERIALS AND METHODS

We mapped diagnosis (ICD code) data to standardized phecodes from 3 EHR-linked biobanks with varying recruitment strategies: All of Us (AOU; n = 244 071), Michigan Genomics Initiative (MGI; n = 81 243), and UK Biobank (UKB; n = 401 167). Using 2019 National Health Interview Survey data, we constructed selection weights for AOU and MGI to represent the US adult population more. We used weights previously developed for UKB to represent the UKB-eligible population. We conducted 4 common analyses comparing unweighted and weighted results.

RESULTS

For AOU and MGI, estimated phecode prevalences decreased after weighting (weighted-unweighted median phecode prevalence ratio [MPR]: 0.82 and 0.61), while UKB estimates increased (MPR: 1.06). Weighting minimally impacted latent phenome dimensionality estimation. Comparing weighted versus unweighted phenome-wide association study for colorectal cancer, the strongest associations remained unaltered, with considerable overlap in significant hits. Weighting affected the estimated log-odds ratio for sex and colorectal cancer to align more closely with national registry-based estimates.

DISCUSSION

Weighting had a limited impact on dimensionality estimation and large-scale hypothesis testing but impacted prevalence and association estimation. When interested in estimating effect size, specific signals from untargeted association analyses should be followed up by weighted analysis.

CONCLUSION

EHR-linked biobanks should report recruitment and selection mechanisms and provide selection weights with defined target populations. Researchers should consider their intended estimands, specify source and target populations, and weight EHR-linked biobank analyses accordingly.

Invited Commentary: Combining Information to Answer Epidemiologic Questions About a Target Population

Epidemiologists are attempting to address research questions of increasing complexity by developing novel methods for combining information from diverse sources. Cole et al. (Am J Epidemiol. 2023;192(3)467-474) provide 2 examples of the process of combining information to draw inferences about a population proportion. In this commentary, we consider combining information to learn about a target population as an epidemiologic activity and distinguish it from more conventional meta-analyses. We examine possible rationales for combining information and discuss broad methodological considerations, with an emphasis on study design, assumptions, and sources of uncertainty.

Comparators in Pharmacovigilance: A Quasi-Quantification Bias Analysis

BACKGROUND AND OBJECTIVE

It is unclear which comparator is the most appropriate for bias reduction in disproportionality analyses based on spontaneous reports. We conducted a quasi-quantitative bias analysis using two well-studied drug-event combinations to assess how different comparators influence the directionality of bias in pharmacovigilance.

METHODS

We used the US Food and Drug Administration Adverse Event Reporting System focusing on two drug-event combinations with a propensity for stimulated reporting: rivaroxaban and hepatotoxicity, and canagliflozin and acute kidney injury. We assessed the directionality of three disproportionality analysis estimates (reporting odds ratio, proportional reporting ratio, information component) using one unrestricted comparator (full data) and two restricted comparators (active comparator, active comparator with class exclusion). Analyses were conducted within two calendar time periods, defined based on external events (approval of direct oral anticoagulants, Food and Drug Administration safety warning on acute kidney injury with sodium-glucose cotransporter 2 inhibitors) hypothesized to alter reporting rates.

RESULTS

There were no false-positive signals for rivaroxaban and hepatotoxicity irrespective of the comparator. Restricting to the initial post-approval period led to false-positive signals, with restricted comparators performing worse. There were false-positive signals for canagliflozin and acute kidney injury, with restricted comparators performing better. Restricting to the period before the Food and Drug Administration warning weakened the false-positive signal for canagliflozin and acute kidney injury across comparators.

CONCLUSIONS

We could not identify a consistent and predictable pattern to the directionality of disproportionality analysis estimates with specific comparators. Calendar time-based restrictions anchored on relevant external events had a considerable impact.

Estimating the impact of bias in causal epidemiological studies: the case of health outcomes following assisted reproduction

Researchers interested in causal questions must deal with two sources of error: random error (random deviation from the true mean value of a distribution), and bias (systematic deviance from the true mean value due to extraneous factors). For some causal questions, randomization is not feasible, and observational studies are necessary. Bias poses a substantial threat to the validity of observational research and can have important consequences for health policy developed from the findings. The current piece describes bias and its sources, outlines proposed methods to estimate its impacts in an observational study, and demonstrates how these methods may be used to inform debate on the causal relationship between medically assisted reproduction (MAR) and health outcomes, using cancer as an example. In doing so, we aim to enlighten researchers who work with observational data, especially regarding the health effects of MAR and infertility, on the pitfalls of bias, and how to address them. We hope that, in combination with the provided example, we can convince readers that estimating the impact of bias in causal epidemiologic research is not only important but necessary to inform the development of robust health policy and clinical practice recommendations.

Use of quantitative bias analysis to evaluate single-arm trials with real-world data external controls

PURPOSE

Use of real-world data (RWD) for external controls added to single-arm trials (SAT) is increasingly prevalent in regulatory submissions. Due to inherent differences in the data-generating mechanisms, biases can arise. This paper aims to illustrate how to use quantitative bias analysis (QBA).

METHODS

Advanced non-small cell lung cancer (NSCLC) serves as an example, where many small subsets of patients with molecular tumor subtypes exist. First, some sources of bias that may occur in oncology when comparing RWD to SAT are described. Second, using a hypothetical immunotherapy agent, a dataset is simulated based on expert input for survival analysis of advanced NSCLC. Finally, we illustrate the impact of three biases: missing confounder, misclassification of exposure, and outcome evaluation.

RESULTS

For each simulated scenario, bias was induced by removing or adding data; hazard ratios (HRs) were estimated applying conventional analyses. Estimating the bias-adjusted treatment effect and uncertainty required carefully selecting the bias model and bias factors. Although the magnitude of each biased and bias-adjusted HR appeared moderate in all three hypothetical scenarios, the direction of bias was variable.

CONCLUSION

These findings suggest that QBA can provide an intuitive framework for bias analysis, providing a key means of challenging assumptions about the evidence. However, the accuracy of bias analysis is itself dependent on correct specification of the bias model and bias factors. Ultimately, study design should reduce bias, but QBA allows us to evaluate the impact of unavoidable bias to assess the quality of the evidence.

Application of a Web-based Tool for Quantitative Bias Analysis: The Example of Misclassification Due to Self-reported Body Mass Index

BACKGROUND

We describe the use of Apisensr, a web-based application that can be used to implement quantitative bias analysis for misclassification, selection bias, and unmeasured confounding. We apply Apisensr using an example of exposure misclassification bias due to use of self-reported body mass index (BMI) to define obesity status in an analysis of the relationship between obesity and diabetes.

METHODS

We used publicly available data from the National Health and Nutrition Examination Survey. The analysis consisted of: (1) estimating bias parameter values (sensitivity, specificity, negative predictive value, and positive predictive value) for self-reported obesity by sex, age, and race-ethnicity compared to obesity defined by measured BMI, and (2) using Apisensr to adjust for exposure misclassification.

RESULTS

The discrepancy between self-reported and measured obesity varied by demographic group (sensitivity range: 75%-89%; specificity range: 91%-99%). Using Apisensr for quantitative bias analysis, there was a clear pattern in the results: the relationship between obesity and diabetes was underestimated using self-report in all age, sex, and race-ethnicity categories compared to measured obesity. For example, in non-Hispanic White men aged 40-59 years, prevalence odds ratios for diabetes were 3.06 (95% confidence inerval = 1.78, 5.30) using self-reported BMI and 4.11 (95% confidence interval = 2.56, 6.75) after bias analysis adjusting for misclassification.

CONCLUSION

Apisensr is an easy-to-use, web-based Shiny app designed to facilitate quantitative bias analysis. Our results also provide estimates of bias parameter values that can be used by other researchers interested in examining obesity defined by self-reported BMI.

Preexisting Diabetes Mellitus and All-Cause Mortality in Adult Patients With Sepsis: A Population-Based Cohort Study

OBJECTIVES

We assessed the association of preexisting diabetes mellitus with all-cause mortality and organ support receipt in adult patients with sepsis.

DESIGN

Population-based cohort study.

SETTING

Ontario, Canada (2008-2019).

POPULATION

Adult patients (18 yr old or older) with a first sepsis-related hospitalization episode.

INTERVENTIONS

None.

MEASUREMENTS AND MAIN RESULTS

The main exposure of interest was preexisting diabetes (either type 1 or 2). The primary outcome was all-cause mortality by 90 days; secondary outcomes included receipt of invasive mechanical ventilation and new renal replacement therapy. We report adjusted (for baseline characteristics using standardization) risk ratios (RRs) alongside 95% CIs. A main secondary analysis evaluated the potential mediation by prior metformin use of the association between preexisting diabetes and all-cause mortality following sepsis. Overall, 503,455 adults with a first sepsis-related hospitalization episode were included; 36% had preexisting diabetes. Mean age was 73 years, and 54% of the cohort were females. Preexisting diabetes was associated with a lower adjusted risk of all-cause mortality at 90 days (RR, 0.81; 95% CI, 0.80-0.82). Preexisting diabetes was associated with an increased risk of new renal replacement therapy (RR, 1.53; 95% CI, 1.46-1.60) but not invasive mechanical ventilation (RR, 1.03; 95% CI, 1.00-1.05). Overall, 21% (95% CI, 19-28) of the association between preexisting diabetes and reduced risk of all-cause mortality was mediated by prior metformin use.

CONCLUSIONS

Preexisting diabetes is associated with a lower risk of all-cause mortality and higher risk of new renal replacement therapy among adult patients with sepsis. Future studies should evaluate the underlying mechanisms of these associations.

Infertility treatment is associated with increased risk of postpartum hospitalization due to heart disease

BACKGROUND

Cardiovascular disease is a major cause of maternal mortality, but the extent to which infertility treatment is implicated in heart disease remains unclear.

OBJECTIVE

To evaluate the association between infertility treatment and postpartum heart disease.

METHODS

We designed a retrospective cohort study of patients who delivered in the United States between 2010 and 2018. The primary outcome was hospitalization within 12-month post-delivery due to heart disease (including ischemic heart disease, atherosclerotic heart disease, cardiomyopathy, hypertensive disease, heart failure, and cardiac dysrhythmias). We estimated the rate difference (RD) of hospitalizations among patients who conceived with infertility treatment and those who conceived spontaneously. Associations were expressed as hazard ratios (HRs) and 95% confidence intervals (CIs), derived from Cox proportional hazards regression after adjustment for potential confounders.

RESULTS

Infertility treatment was recorded in 0.9% (n = 287,813) of 31,339,991 deliveries. Rates of heart disease hospitalizations with infertility treatment and with spontaneous conception were 550 and 355 per 100,000, respectively (RD 195, 95% CI: 143-247; adjusted HR 1.99, 95% CI: 1.80-2.20). The most important increase in risk was observed for hypertensive disease (adjusted HR 2.16, 95% CI: 1.92-2.42). This increased risk was apparent as early as 30-day post-delivery (HR 1.61, 95% CI: 1.39-1.86), with progressively increasing risk up to a year.

CONCLUSIONS

Although the absolute risk of postpartum heart disease hospitalization is low, infertility treatment is associated with an increased risk, especially for hypertensive disease. These findings highlight the importance of timely postpartum follow-ups in patients who received infertility treatment.

Incident Diabetes in Women With Patterns of Gestational Diabetes Occurrences Across 2 Pregnancies

Importance

Gestational diabetes is a type 2 diabetes risk indicator, and recurrence further augments risk. In women with a single occurrence across 2 pregnancies, it is unclear whether first- vs second-pregnancy gestational diabetes differ in terms of risk.

Objective

To compare the hazards of incident diabetes among those with gestational diabetes in the first, in the second, and in both pregnancies with women without gestational diabetes in either.

Design, Setting, and Participants

This was a retrospective cohort study with cohort inception from April 1, 1990, to December 31, 2012. Follow-up was April 1, 1990, to April 1, 2019. Participants were mothers with 2 singleton deliveries between April 1, 1990, and December 31, 2012, without diabetes before or between pregnancies, who were listed in public health care insurance administrative databases and birth, stillbirth, and death registries in Quebec, Canada. Data were analyzed from July to December 2023.

Exposure

Gestational diabetes occurrence(s) across 2 pregnancies.

Main outcomes and measures

Incident diabetes from the second delivery until a third pregnancy, death, or the end of the follow-up period, whichever occurred first.

Results

The 431 980 women with 2 singleton deliveries studied had a mean (SD) age of 30.1 (4.5) years at second delivery, with a mean (SD) of 2.8 (1.5) years elapsed between deliveries; 373 415 (86.4%) were of European background, and 78 770 (18.2%) were at the highest quintile of material deprivation. Overall, 10 920 women (2.5%) had gestational diabetes in their first pregnancy, 16 145 (3.7%) in their second, and 8255 (1.9%) in both (12 205 incident diabetes events; median [IQR] follow-up 11.5 [5.3-19.4] years). First pregnancy-only gestational diabetes increased hazards 4.35-fold (95% CI, 4.06-4.67), second pregnancy-only increased hazards 7.68-fold (95% CI, 7.31-8.07), and gestational diabetes in both pregnancies increased hazards 15.8-fold (95% CI, 15.0-16.6). Compared with first pregnancy-only gestational diabetes, second pregnancy-only gestational diabetes increased hazards by 76% (95% CI, 1.63-1.91), while gestational diabetes in both pregnancies increased it 3.63-fold (95% CI, 3.36-3.93).

Conclusions and relevance

In this retrospective cohort study of nearly half a million women with 2 singleton pregnancies, both the number and ordinal pregnancy of any gestational diabetes occurrence increased diabetes risk. These considerations offer greater nuance than an ever or never gestational diabetes dichotomy.

Effectiveness of BNT162b2 COVID-19 primary series vaccination in children aged 5-17 years in the United States: a cohort study

BACKGROUND

COVID-19 vaccines are authorized for use in children in the United States; real-world assessment of vaccine effectiveness in children is needed. This study's objective was to estimate the effectiveness of receiving a complete primary series of monovalent BNT162b2 (Pfizer-BioNTech) COVID-19 vaccine in US children.

METHODS

This cohort study identified children aged 5-17 years vaccinated with BNT162b2 matched with unvaccinated children. Participants and BNT162b2 vaccinations were identified in Optum and CVS Health insurance administrative claims databases linked with Immunization Information System (IIS) COVID-19 vaccination records from 16 US jurisdictions between December 11, 2020, and May 31, 2022 (end date varied by database and IIS). Vaccinated children were followed from their first BNT162b2 dose and matched to unvaccinated children on calendar date, US county of residence, and demographic and clinical factors. Censoring occurred if vaccinated children failed to receive a timely dose 2 or if unvaccinated children received any dose. Two COVID-19 outcome definitions were evaluated: COVID-19 diagnosis in any medical setting and COVID-19 diagnosis in hospitals/emergency departments (EDs). Propensity score-weighted hazard ratios (HRs) and 95% confidence intervals (CIs) were estimated with Cox proportional hazards models, and vaccine effectiveness (VE) was estimated as 1 minus HR. VE was estimated overall, within age subgroups, and within variant-specific eras. Sensitivity, negative control, and quantitative bias analyses evaluated various potential biases.

RESULTS

There were 453,655 eligible vaccinated children one-to-one matched to unvaccinated comparators (mean age 12 years; 50% female). COVID-19 hospitalizations/ED visits were rare in children, regardless of vaccination status (Optum, 41.2 per 10,000 person-years; CVS Health, 44.1 per 10,000 person-years). Overall, vaccination was associated with reduced incidence of any medically diagnosed COVID-19 (meta-analyzed VE = 38% [95% CI, 36-40%]) and hospital/ED-diagnosed COVID-19 (meta-analyzed VE = 61% [95% CI, 56-65%]). VE estimates were lowest among children 5-11 years and during the Omicron-variant era.

CONCLUSIONS

Receipt of a complete BNT162b2 vaccine primary series was associated with overall reduced medically diagnosed COVID-19 and hospital/ED-diagnosed COVID-19 in children; observed VE estimates differed by age group and variant era.

REGISTRATION

The study protocol was publicly posted on the BEST Initiative website ( https://bestinitiative.org/wp-content/uploads/2022/03/C19-VX-Effectiveness-Protocol_2022_508.pdf ).

Association between Empirical Anti-Pseudomonal Antibiotics and Progression to Thoracic Surgery and Death in Empyema: Database Research

Evidence on the optimal antibiotic strategy for empyema is lacking. Our database study aimed to evaluate the effectiveness of empirical anti-pseudomonal antibiotics in patients with empyema. We utilised a Japanese real-world data database, focusing on patients aged ≥40 diagnosed with empyema, who underwent thoracostomy and received intravenous antibiotics either upon admission or the following day. Patients administered intravenous vasopressors were excluded. We compared thoracic surgery and death within 90 days after admission between patients treated with empirical anti-pseudomonal and non-anti-pseudomonal antibiotics. Cause-specific hazard ratios for thoracic surgery and death were estimated using Cox proportional hazards models, with adjustment for clinically important confounders. Subgroup analyses entailed the same procedures for patients exhibiting at least one risk factor for multidrug-resistant organisms. Between March 2014 and March 2023, 855 patients with empyema meeting the inclusion criteria were enrolled. Among them, 271 (31.7%) patients received anti-pseudomonal antibiotics. The Cox proportional hazards models indicated that compared to empirical non-anti-pseudomonal antibiotics, empirical anti-pseudomonal antibiotics were associated with higher HRs for thoracic surgery and death within 90 days, respectively. Thus, regardless of the risks of multidrug-resistant organisms, empirical anti-pseudomonal antibiotics did not extend the time to thoracic surgery or death within 90 days.

Prenatal psychological distress and 11β-HSD2 gene expression in human placentas: Systematic review and meta-analysis

BACKGROUND

The placenta acts as a buffer to regulate the degree of fetal exposure to maternal cortisol through the 11-Beta Hydroxysteroid Dehydrogenase isoenzyme type 2 (11-β HSD2) enzyme. We conducted a systematic review and meta-analysis to assess the effect of prenatal psychological distress (PPD) on placental 11-β HSD2 gene expression and explore the related mechanistic pathways involved in fetal neurodevelopment.

METHODS

We searched PubMed, Embase, Scopus, APA PsycInfo®, and ProQuest Dissertations for observational studies assessing the association between PPD and 11-β HSD2 expression in human placentas. Adjusted regression coefficients (β) and corresponding 95% confidence intervals (CIs) were pooled based on three contextual PPD exposure groups: prenatal depression, anxiety symptoms, and perceived stress.

RESULTS

Of 3159 retrieved records, sixteen longitudinal studies involving 1869 participants across seven countries were included. Overall, exposure to PPD disorders showed weak negative associations with the placental 11-β HSD2 gene expression as follows: prenatal depression (β -0.01, 95% CI 0.05-0.02, I2=0%), anxiety symptoms (β -0.02, 95% CI 0.06-0.01, I2=0%), and perceived stress (β -0.01 95% CI 0.06-0.04, I2=62.8%). Third-trimester PPD exposure was more frequently associated with lower placental 11-β HSD2 levels. PPD and placental 11-β HSD2 were associated with changes in cortisol reactivity and the development of adverse health outcomes in mothers and children. Female-offspring were more vulnerable to PPD exposures.

CONCLUSION

The study presents evidence of a modest role of prenatal psychological distress in regulating placental 11-β HSD2 gene expression. Future prospective cohorts utilizing larger sample sizes or advanced statistical methods to enhance the detection of small effect sizes should be planned. Additionally, controlling for key predictors such as the mother's ethnicity, trimester of PPD exposure, mode of delivery, and infant sex is crucial for valid exploration of PPD effects on fetal programming.

The PRECISE-DYAD protocol: linking maternal and infant health trajectories in sub-Saharan Africa

Background

PRECISE-DYAD is an observational cohort study of mother-child dyads running in urban and rural communities in The Gambia and Kenya. The cohort is being followed for two years and includes uncomplicated pregnancies and those that suffered pregnancy hypertension, fetal growth restriction, preterm birth, and/or stillbirth.

Methods

The PRECISE-DYAD study will follow up ~4200 women and their children recruited into the original PRECISE study. The study will add to the detailed pregnancy information and samples in PRECISE, collecting additional biological samples and clinical information on both the maternal and child health.Women will be asked about both their and their child's health, their diets as well as undertaking a basic cardiology assessment. Using a case-control approach, some mothers will be asked about their mental health, their experiences of care during labour in the healthcare facility. In a sub-group, data on financial expenditure during antenatal, intrapartum, and postnatal periods will also be collected. Child development will be assessed using a range of tools, including neurodevelopment assessments, and evaluating their home environment and quality of life. In the event developmental milestones are not met, additional assessments to assess vision and their risk of autism spectrum disorders will be conducted. Finally, a personal environmental exposure model for the full cohort will be created based on air and water quality data, combined with geographical, demographic, and behavioural variables.

Conclusions

The PRECISE-DYAD study will provide a greater epidemiological and mechanistic understanding of health and disease pathways in two sub-Saharan African countries, following healthy and complicated pregnancies. We are seeking additional funding to maintain this cohort and to gain an understanding of the effects of pregnancies outcome on longer-term health trajectories in mothers and their children.

Avoidance of causality outside experiments: Hypotheses from cognitive dissonance reduction

The avoidance of causality in the design, analysis and interpretation of non-experimental studies has often been criticised as an untenable scientific stance, because theories are based on causal relations (and not associations) and a rich set of methodological tools for causal analysis has been developed in recent decades. Psychology researchers (n = 106 with complete data) participated in an online study presenting a causal statement about the results of a fictitious paper on the potential effect of drinking clear water for years on the risk of dementia. Two randomised groups of participants were then asked to reflect on the conflict between the goal of approaching a causal answer and the prevailing norm of avoiding doing so. One of the two groups was also instructed to think about possible benefits of addressing causality. Both groups then responded to a list of 19 items about attitudes to causal questions in science. A control group did this without reflecting on conflict or benefits. Free-text assessments were also collected during reflection, giving some indication of how and why causality is avoided. We condense the exploratory findings of this study into five new hypotheses about the how and why, filtered through what can be explained by cognitive dissonance reduction theory. These concern the cost of addressing causality, the variety of ways in which dissonance can be reduced, the need for profound intervention through teaching and social aspects. Predictions are derived from the hypotheses for confirmation trials in future studies and recommendations for teaching causality. Open data are provided for researchers' own analyses.

Challenges in and Opportunities for Electronic Health Record-Based Data Analysis and Interpretation

Electronic health records (EHRs) have been increasingly adopted in clinical practices across the United States, providing a primary source of data for clinical research, particularly observational cohort studies. EHRs are a high-yield, low-maintenance source of longitudinal real-world data for large patient populations and provide a wealth of information and clinical contexts that are useful for clinical research and translation into practice. Despite these strengths, it is important to recognize the multiple limitations and challenges related to the use of EHR data in clinical research. Missing data are a major source of error and biases and can affect the representativeness of the cohort of interest, as well as the accuracy of the outcomes and exposures. Here, we aim to provide a critical understanding of the types of data available in EHRs and describe the impact of data heterogeneity, quality, and generalizability, which should be evaluated prior to and during the analysis of EHR data. We also identify challenges pertaining to data quality, including errors and biases, and examine potential sources of such biases and errors. Finally, we discuss approaches to mitigate and remediate these limitations. A proactive approach to addressing these issues can help ensure the integrity and quality of EHR data and the appropriateness of their use in clinical studies.

Gestational diabetes mellitus in pregnancies conceived after infertility treatment: a population-based study in the United States, 2015-2020

Objective

To evaluate the risk of gestational diabetes mellitus (GDM) in singleton pregnancies conceived using infertility treatment and examine the influence of race and ethnicity as well as prepregnancy body mass index (BMI).

Design

Cross-sectional study using the US vital records data of women that delivered singleton births.

Setting

United States, 2015-2020.

Interventions

Any infertility treatment was divided into two groups: those that used fertility-enhancing drugs, artificial insemination, or intrauterine insemination, and those that used assisted reproductive technology (ART).

Main Outcome Measuress

Gestational diabetes mellitus, defined as a diagnosis of diabetes mellitus during pregnancy, includes both diet-controlled GDM and medication-controlled GDM in singleton pregnancies conceived with infertility treatment or spontaneously and delivered between 20- and 44-weeks' gestation. We also examined whether the infertility treatment-GDM association was modified by maternal race and ethnicity as well as prepregnancy BMI. Associations were expressed as a rate ratio (RR) and 95% confidence interval (CI), derived from log-linear models after adjustment for potential confounders.

Results

A total of 21,943,384 singleton births were included, with 1.5% (n = 318,086) undergoing infertility treatment. Rates of GDM among women undergoing infertility treatment and those who conceived spontaneously were 11.0% (n = 34,946) and 6.5% (n = 1,398,613), respectively (adjusted RR 1.24, 95% CI 1.23, 1.26). The RRs were adjusted for maternal age, parity, education, race and ethnicity, smoking, BMI, chronic hypertension, and year of delivery. The risk of GDM was modestly increased for those using fertility-enhancing drugs (adjusted RR 1.28, 95% CI 1.27, 1.30) compared with ART (adjusted RR 1.18, 95% CI 1.17, 1.20), and this risk was especially apparent for non-Hispanic White women (adjusted RR 1.29, 95% CI 1.26, 1.31) and Hispanic women (adjusted RR 1.35, 95% CI 1.29, 1.41). The number of women who needed to be exposed to infertility treatment to diagnose one case of GDM was 46. Prepregnancy BMI did not modify the infertility treatment-GDM association overall and within strata of race and ethnicity. These general patterns were stronger after potential corrections for misclassification of infertility treatment and unmeasured confounding.

Conclusions

Infertility treatment, among those who received fertility-enhancing drugs, is associated with an increased GDM risk. The persistently higher risk of GDM among women who seek infertility treatment, irrespective of prepregnancy weight classification, deserves attention. Infertility specialists must be vigilant with preconception counseling and ensure that all patients, regardless of race and ethnicity or BMI, are adequately tested for GDM early in pregnancy using a fasting blood glucose level or a traditional 50-g oral glucose tolerance test. Testing may be completed by the infertility specialist or deferred to the primary prenatal care provider at the first prenatal visit.

Assessing cardiovascular disease risk in women with a history of hypertensive disorders of pregnancy: A guidance paper for studies using administrative data

BACKGROUND

Hypertensive disorders of pregnancy (HDP) are a major cause of maternal morbidity and mortality, and their association with increased cardiovascular disease (CVD) risk represents a major public health concern. However, assessing CVD risk in women with a history of these conditions presents unique challenges, especially when studies are carried out using routinely collected data.

OBJECTIVES

To summarise and describe key challenges related to the design and conduct of administrative studies assessing CVD risk in women with a history of HDP and provide concrete recommendations for addressing them in future research.

METHODS

This is a methodological guidance paper.

RESULTS

Several conceptual and methodological factors related to the data-generating mechanism and study conceptualisation, design/data management and analysis, as well as the interpretation and reporting of study findings should be considered and addressed when designing and carrying out administrative studies on this topic. Researchers should develop an a priori conceptual framework within which the research question is articulated, important study variables are identified and their interrelationships are carefully considered.

CONCLUSIONS

To advance our understanding of CVD risk in women with a history of HDP, future studies should carefully consider and address the conceptual and methodological considerations outlined in this guidance paper. In highlighting these challenges, and providing specific recommendations for how to address them, our goal is to improve the quality of research carried out on this topic.

Quantitative bias analysis for external control arms using real-world data in clinical trials: a primer for clinical researchers

Development of medicines in rare oncologic patient populations are growing, but well-powered randomized controlled trials are typically extremely challenging or unethical to conduct in such settings. External control arms using real-world data are increasingly used to supplement clinical trial evidence where no or little control arm data exists. The construction of an external control arm should always aim to match the population, treatment settings and outcome measurements of the corresponding treatment arm. Yet, external real-world data is typically fraught with limitations including missing data, measurement error and the potential for unmeasured confounding given a nonrandomized comparison. Quantitative bias analysis (QBA) comprises a collection of approaches for modelling the magnitude of systematic errors in data which cannot be addressed with conventional statistical adjustment. Their applications can range from simple deterministic equations to complex hierarchical models. QBA applied to external control arm represent an opportunity for evaluating the validity of the corresponding comparative efficacy estimates. We provide a brief overview of available QBA approaches and explore their application in practice. Using a motivating example of a comparison between pralsetinib single-arm trial data versus pembrolizumab alone or combined with chemotherapy real-world data for RET fusion-positive advanced non-small cell lung cancer (aNSCLC) patients (1-2% among all NSCLC), we illustrate how QBA can be applied to external control arms. We illustrate how QBA is used to ascertain robustness of results despite a large proportion of missing data on baseline ECOG performance status and suspicion of unknown confounding. The robustness of findings is illustrated by showing that no meaningful change to the comparative effect was observed across several 'tipping-point' scenario analyses, and by showing that suspicion of unknown confounding was ruled out by use of E-values. Full R code is also provided.

Considerations for master protocols using external controls

There has been an increasing use of master protocols in oncology clinical trials because of its efficiency to accelerate cancer drug development and flexibility to accommodate multiple substudies. Depending on the study objective and design, a master protocol trial can be a basket trial, an umbrella trial, a platform trial, or any other form of trials in which multiple investigational products and/or subpopulations are studied under a single protocol. Master protocols can use external data and evidence (e.g. external controls) for treatment effect estimation, which can further improve efficiency of master protocol trials. This paper provides an overview of different types of external controls and their unique features when used in master protocols. Some key considerations in master protocols with external controls are discussed including construction of estimands, assessment of fit-for-use real-world data, and considerations for different types of master protocols. Similarities and differences between regular randomized controlled trials and master protocols when using external controls are discussed. A targeted learning-based causal roadmap is presented which constitutes three key steps: (1) define a target statistical estimand that aligns with the causal estimand for the study objective, (2) use an efficient estimator to estimate the target statistical estimand and its uncertainty, and (3) evaluate the impact of causal assumptions on the study conclusion by performing sensitivity analyses. Two illustrative examples for master protocols using external controls are discussed for their merits and possible improvement in causal effect estimation.

To weight or not to weight? Studying the effect of selection bias in three large EHR-linked biobanks

Objective

To explore the role of selection bias adjustment by weighting electronic health record (EHR)-linked biobank data for commonly performed analyses.

Materials and methods

We mapped diagnosis (ICD code) data to standardized phecodes from three EHR-linked biobanks with varying recruitment strategies: All of Us (AOU; n=244,071), Michigan Genomics Initiative (MGI; n=81,243), and UK Biobank (UKB; n=401,167). Using 2019 National Health Interview Survey data, we constructed selection weights for AOU and MGI to be more representative of the US adult population. We used weights previously developed for UKB to represent the UKB-eligible population. We conducted four common descriptive and analytic tasks comparing unweighted and weighted results.

Results

For AOU and MGI, estimated phecode prevalences decreased after weighting (weighted-unweighted median phecode prevalence ratio [MPR]: 0.82 and 0.61), while UKB's estimates increased (MPR: 1.06). Weighting minimally impacted latent phenome dimensionality estimation. Comparing weighted versus unweighted PheWAS for colorectal cancer, the strongest associations remained unaltered and there was large overlap in significant hits. Weighting affected the estimated log-odds ratio for sex and colorectal cancer to align more closely with national registry-based estimates.

Discussion

Weighting had limited impact on dimensionality estimation and large-scale hypothesis testing but impacted prevalence and association estimation more. Results from untargeted association analyses should be followed by weighted analysis when effect size estimation is of interest for specific signals.

Conclusion

EHR-linked biobanks should report recruitment and selection mechanisms and provide selection weights with defined target populations. Researchers should consider their intended estimands, specify source and target populations, and weight EHR-linked biobank analyses accordingly.

Introduction to target trial emulation in rehabilitation: a systematic approach to emulate a randomized controlled trial using observational data

Rehabilitation providers and policymakers need valid evidence to make informed decisions about the healthcare needs of the population. Whenever possible, these decisions should be informed by randomized controlled trials (RCTs). However, there are circumstances when evidence needs to be generated rapidly, or when RCTs are not ethical or feasible. These situations apply to studying the effects of complex interventions, including rehabilitation as defined by Cochrane Rehabilitation. Therefore, we explore using the target trial emulation framework by Hernán and colleagues to obtain valid estimates of the causal effects of rehabilitation when RCTs cannot be conducted. Target trial emulation is a framework guiding the design and analysis of non-randomized comparative effectiveness studies using observational data, by emulating a hypothetical RCT. In the context of rehabilitation, we outline steps for applying the target trial emulation framework using real world data, highlighting methodological considerations, limitations, potential mitigating strategies, and causal inference and counterfactual theory as foundational principles to estimating causal effects. Overall, we aim to strengthen methodological approaches used to estimate causal effects of rehabilitation when RCTs cannot be conducted.

Prospective study design and data analysis in UK Biobank

Population-based prospective studies, such as UK Biobank, are valuable for generating and testing hypotheses about the potential causes of human disease. We describe how UK Biobank's study design, data access policies, and approaches to statistical analysis can help to minimize error and improve the interpretability of research findings, with implications for other population-based prospective studies being established worldwide.

Childhood body fatness and the risk of epithelial ovarian cancer: A population-based case-control study in Montreal, Canada

OBJECTIVE

To assess the association between childhood body fatness and epithelial ovarian cancer (EOC), and whether this association differs by type of EOC.

METHODS

Using data from a population-based case-control study (497 cases and 902 controls) in Montreal, Canada conducted 2011-2016, we examined the association between childhood body fatness and EOC, overall and separately for invasive vs. borderline EOCs. A figure rating scale was used to measure body fatness at ages 5 and 10. Multivariable logistic regression was used to estimate adjusted odds ratios (aORs) and 95% confidence intervals (95% CI). Quantitative bias analyses were conducted to assess the impact of exposure misclassification and non-participation.

RESULTS

The aOR (95% CI) of overall EOC for high vs. low body fatness was 1.07 (0.85-1.34) at age 5 and 1.28 (0.98-1.68) at age 10. The associations were stronger for invasive EOC, specifically the endometrioid histological type. For borderline cancers, the aORs were below the null value with wide confidence intervals. Bias analyses did not reveal a strong influence of non-participation. Non-differential exposure misclassification may have biased aORs towards the null for invasive cancers but did not appear to have an appreciable influence on the aORs for borderline cancers.

CONCLUSIONS

Childhood body fatness may be a risk factor for invasive EOC in later adult life. Our study highlights the potential importance of examining early life factors for a comprehensive understanding of EOC development.

Effectiveness of monovalent COVID-19 booster/additional vaccine doses in the United States

Background

Monovalent booster/additional doses of COVID-19 vaccines were first authorized in August 2021 in the United States. We evaluated the real-world effectiveness of receipt of a monovalent booster/additional dose of COVID-19 vaccine compared with receiving a primary vaccine series without a booster/additional dose.

Methods

Cohorts of individuals receiving a COVID-19 booster/additional dose after receipt of a complete primary vaccine series were identified in 2 administrative insurance claims databases (Optum, CVS Health) supplemented with state immunization information system data between August 2021 and March 2022. Individuals with a complete primary series but without a booster/additional dose were one-to-one matched to boosted individuals on calendar date, geography, and clinical factors. COVID-19 diagnoses were identified in any medical setting, or specifically in hospitals/emergency departments (EDs). Propensity score-weighted hazards ratios (HRs) and 95% confidence intervals (CI) were estimated with Cox proportional hazards models; vaccine effectiveness (VE) was estimated as 1 minus the HR by vaccine brand overall and within subgroups of variant-specific eras, immunocompromised status, and homologous/heterologous booster status.

Results

Across both data sources, we identified 752,165 matched pairs for BNT162b2, 410,501 for mRNA-1273, and 11,398 for JNJ-7836735. For any medically diagnosed COVID-19, meta-analyzed VE estimates for BNT162b2, mRNA-1273, and JNJ-7836735, respectively, were: BNT162b2, 54% (95% CI, 53%-56%); mRNA-1273, 58% (95% CI, 56%-59%); JNJ-7836735, 34% (95% CI, 23%-44%). For hospital/ED-diagnosed COVID-19, VE estimates ranged from 70% to 76%. VE was generally lower during the Omicron era than the Delta era and for immunocompromised individuals. There was little difference observed by homologous or heterologous booster status.

Conclusion

The original, monovalent booster/additional doses were reasonably effective in real-world use among the populations for which they were indicated during the study period. Additional studies may be informative in the future as new variants emerge and new vaccines become available.Registration: The study protocol was publicly posted on the BEST Initiative website (https://bestinitiative.org/wp-content/uploads/2022/03/C19-VX-Effectiveness-Protocol_2022_508.pdf).

A scoping review of multigenerational impacts of grandparental exposures on mental health in grandchildren

PURPOSE OF REVIEW

The multigenerational effects of grandparental exposures on their grandchildren's mental health and neurodevelopment are gaining research attention. We conducted a scoping review to summarize the current epidemiological studies investigating pregnancy-related and environmental factors that affected grandparental pregnancies and mental health outcomes in their grandchildren. We also identified methodological challenges that affect these multigenerational health studies and discuss opportunities for future research.

RECENT FINDINGS

We performed a literature search using PubMed and Embase and included 18 articles for this review. The most investigated grandparental pregnancy-related factors were the grandparental age of pregnancy (N = 6), smoking during pregnancy (N = 4), and medication intake (N = 3). The most frequently examined grandchild outcomes were autism spectrum disorder (N = 6) and attention-deficit/hyperactivity disorder (N = 4). Among these studies, grandparental smoking and the use of diethylstilbestrol were more consistently reported to be associated with neurodevelopmental disorders, while the findings for grandparental age vary across the maternal or paternal line. Grandmaternal weight, adverse delivery outcomes, and other spatial-temporal markers of physical and social environmental stressors require further scrutiny. The current body of literature has suggested that mental and neurodevelopmental disorders may be outcomes of unfavorable exposures originating from the grandparental generation during their pregnancies. To advance the field, we recommend research efforts into setting up multigenerational studies with prospectively collected data that span through at least three generations, incorporating spatial, environmental, and biological markers for exposure assessment, expanding the outcome phenotypes evaluated, and developing a causal analytical framework including mediation analyses specific for multigenerational research.

Machine-Based Learning Hierarchical Cluster Analysis: Sex-Based Differences in Prognosis Following Resection of Hepatocellular Carcinoma

BACKGROUND

Patients with hepatocellular carcinoma (HCC) may have a heterogeneous presentation, as well as different long-term outcomes following surgical resection. We sought to use machine learning to cluster patients into different prognostic groups based on preoperative characteristics.

METHODS

Patients who underwent curative-intent liver resection for HCC between 2000 and 2020 were identified from a large international multi-institutional database. A hierarchical cluster analysis was performed based on preoperative factors to characterize patterns of presentation and define disease-free survival (DFS).

RESULTS

Among 966 with HCC, 3 distinct clusters were identified: Cluster 1 (n = 160, 16.5%), Cluster 2 (n = 537, 55.6%) and Cluster 3 (n = 269, 27.8%). Cluster 1 (n = 160, 16.5%) consisted of female patients (n = 160, 100%), low inflammation-based scores, intermediate tumor burden score (TBS) (median: 4.71) and high alpha-fetoprotein (AFP) levels (median 41.3 ng/mL); Cluster 2 consisted of male individuals (n = 537, 100%), mainly with a history of HBV infection (n = 429, 79.9%), low inflammation-based scores, intermediate AFP levels (median 26.0 ng/mL) and lower TBS (median 4.49); Cluster 3 was comprised of older patients (median age 68 years) predominantly male (n = 248, 92.2%) who had low incidence of HBV/HCV infection (7.1% and 8.2%, respectively), intermediate AFP levels (median 16.8 ng/mL), high inflammation-based scores and high TBS (median 6.58). Median DFS worsened incrementally among the three different clusters with Cluster 3 having the lowest DFS (Cluster 1: median not reached; Cluster 2: 34 months, 95% CI 23.0-48.0, Cluster 3: 19 months, 95% CI 15.0-29.0, p < 0.05).

CONCLUSION

Cluster analysis classified HCC patients into three distinct prognostic groups. Cluster assignment predicted DFS following resection of HCC with the female cluster having the most favorable prognosis following HCC resection.

Can Weight of Evidence, Quantitative Bias, and Bounding Methods Evaluate Robustness of Real-world Evidence for Regulator and Health Technology Assessment Decisions on Medical Interventions?

PURPOSE

High-quality evidence is crucial for health care intervention decision-making. These decisions frequently use nonrandomized data, which can be more vulnerable to biases than randomized trials. Accordingly, methods to quantify biases and weigh available evidence could elucidate the robustness of findings, giving regulators more confidence in making approval and reimbursement decisions.

METHODS

We conducted an integrative literature review to identify methods for determining probability of causation, evaluating weight of evidence, and conducting quantitative bias analysis as related to health care interventions. Eligible studies were published from 2012 to 2021, applicable to pharmacoepidemiology, and presented a method that met our objective.

FINDINGS

Twenty-two eligible studies were classified into 4 categories: (1) quantitative bias analysis; (2) weight of evidence methods; (3) Bayesian networks; and (4) miscellaneous. All of the methods have strengths, limitations, and situations in which they are more well suited than others. Some methods seem to lend themselves more to applications of health care evidence on medical interventions than others.

IMPLICATIONS

To provide robust evidence for and improve confidence in regulatory or reimbursement decisions, we recommend applying multiple methods to triangulate associations of medical interventions, accounting for biases in different ways. This approach could lead to well-defined robustness assessments of study findings and appropriate science-driven decisions by regulators and payers for public health.

Chronic Hypertension and the Risk of Readmission for Postpartum Cardiovascular Complications

OBJECTIVE

Preeclampsia is an important risk factor for cardiovascular disease (CVD, including heart disease and stroke) along the life course. However, whether exposure to chronic hypertension in pregnancy, in the absence of preeclampsia, is implicated in CVD risk during the immediate postpartum period remains poorly understood. Our objective was to estimate the risk of readmission for CVD complications within the calendar year after delivery for people with chronic hypertension.

METHODS

The Healthcare Cost and Utilization Project's Nationwide Readmission Database (2010-2018) was used to conduct a retrospective cohort study of patients aged 15-54 years. International Classification of Diseases codes were used to identify patients with chronic hypertension and postpartum readmission for CVD complications within 1 year of delivery. People with CVD diagnosed during pregnancy or delivery admission, multiple births, or preeclampsia or eclampsia were excluded. Excess rates of CVD readmission among patients with and without chronic hypertension were estimated. Associations between chronic hypertension and CVD complications were determined from Cox proportional hazards regression models.

RESULTS

Of 27,395,346 delivery hospitalizations that resulted in singleton births, 2.0% of individuals had chronic hypertension (n=544,639). The CVD hospitalization rate among patients with chronic hypertension and normotensive patients was 645 (n=3,791) per 100,000 delivery hospitalizations and 136 (n=37,664) per 100,000 delivery hospitalizations, respectively (rate difference 508, 95% CI 467-549; adjusted hazard ratio 4.11, 95% CI 3.64-4.66). The risk of CVD readmission, in relation to chronic hypertension, persisted for 1 year after delivery.

CONCLUSION

The heightened CVD risk as early as 1 month postpartum in relation to chronic hypertension underscores the need for close monitoring and timely care after delivery to reduce blood pressure and related complications.

Concepts of Designing and Implementing Pharmacoepidemiology Studies on the Safety of Systemic Treatments in Dermatology Practice

The U.S. Food and Drug Administration and clinical guidelines use evidence from pharmacoepidemiology studies to inform prescribing decisions and fill evidence gaps left by randomized controlled trials (RCTs). The long-term safety and infrequent adverse reactions are not well-understood when RCTs are short and involve few patients, as is the case for most systemic immunomodulating drugs in dermatology. A better understanding of the design and implementation of pharmacoepidemiology studies will help practitioners assess the accuracy of etiologic findings and use them with confidence in clinical practice. Conducting pharmacoepidemiology studies follows a structured approach, which we discuss in this article: (i) a design layer connects the research question with the appropriate study design, and considering which hypothetical RCT one ideally would want to conduct reduces inadvertent investigator errors; (ii) a measurement layer transforms longitudinal patient-level data into variables that identify the study population, patient characteristics, treatment, and outcomes; and (iii) the analysis focuses on the causal treatment effect estimation. The review and interpretation of pharmacoepidemiology studies should consider issues beyond a typical review of RCTs, chiefly the lack of baseline randomization and the use of secondary data. Well-designed and well-conducted pharmacoepidemiologic studies complement dermatology practice with critical information on prescribing systemic medications.

Quantitative Assessment of Systematic Bias: A Guide for Researchers

Observational research provides valuable opportunities to advance oral health science but is limited by vulnerabilities to systematic bias, including unmeasured confounding, errors in variable measurement, or bias in the creation of study populations and/or analytic samples. The potential influence of systematic biases on observed results is often only briefly mentioned among the discussion of limitations of a given study, despite existing methods that support detailed assessments of their potential effects. Quantitative bias analysis is a set of methodological techniques that, when applied to observational data, can provide important context to aid in the interpretation and integration of observational research findings into the broader body of oral health research. Specifically, these methods were developed to provide quantitative estimates of the potential magnitude and direction of the influence of systematic biases on observed results. We aim to encourage and facilitate the broad adoption of quantitative bias analyses into observational oral health research. To this end, we provide an overview of quantitative bias analysis techniques, including a step-by-step implementation guide. We also provide a detailed appendix that guides readers through an applied example using real data obtained from a prospective observational cohort study of preconception periodontitis in relation to time to pregnancy. Quantitative bias analysis methods are available to all investigators. When appropriately applied to observational studies, findings from such studies can have a greater impact in the broader research context.

SAS and R code for probabilistic quantitative bias analysis for misclassified binary variables and binary unmeasured confounders

Systematic error from selection bias, uncontrolled confounding, and misclassification is ubiquitous in epidemiologic research but is rarely quantified using quantitative bias analysis (QBA). This gap may in part be due to the lack of readily modifiable software to implement these methods. Our objective is to provide computing code that can be tailored to an analyst's dataset. We briefly describe the methods for implementing QBA for misclassification and uncontrolled confounding and present the reader with example code for how such bias analyses, using both summary-level data and individual record-level data, can be implemented in both SAS and R. Our examples show how adjustment for uncontrolled confounding and misclassification can be implemented. Resulting bias-adjusted point estimates can then be compared to conventional results to see the impact of this bias in terms of its direction and magnitude. Further, we show how 95% simulation intervals can be generated that can be compared to conventional 95% confidence intervals to see the impact of the bias on uncertainty. Having easy to implement code that users can apply to their own datasets will hopefully help spur more frequent use of these methods and prevent poor inferences drawn from studies that do not quantify the impact of systematic error on their results.

Confounding adjustment in observational studies on cardiothoracic interventions: a systematic review of methodological practice

OBJECTIVES

It is unknown which confounding adjustment methods are currently used in the field of cardiothoracic surgery and whether these are appropriately applied. The aim of this study was to systematically evaluate the quality of conduct and reporting of confounding adjustment methods in observational studies on cardiothoracic interventions.

METHODS

A systematic review was performed, which included all observational studies that compared different interventions and were published between 1 January and 1 July 2022, in 3 European and American cardiothoracic surgery journals. Detailed information on confounding adjustment methods was extracted and subsequently described.

RESULTS

Ninety-two articles were included in the analysis. Outcome regression (n = 49, 53%) and propensity score (PS) matching (n = 44, 48%) were most popular (sometimes used in combination), whereas 11 (12%) studies applied no method at all. The way of selecting confounders was not reported in 42 (46%) of the studies, solely based on previous literature or clinical knowledge in 14 (16%), and (partly) data-driven in 25 (27%). For the studies that applied PS matching, the matched cohorts comprised on average 46% of the entire study population (range 9-82%).

CONCLUSIONS

Current reporting of confounding adjustment methods is insufficient in a large part of observational studies on cardiothoracic interventions, which makes quality judgement difficult. Appropriate application of confounding adjustment methods is crucial for causal inference on optimal treatment strategies for clinical practice. Reporting on these methods is an important aspect of this, which can be improved.

Misclassification of firearm-related violent crime in criminal legal system records: challenges and opportunities

BACKGROUND

Criminal legal system data are one source for measuring some types of firearm-related harms, including those that do not necessarily result in injury or death, but measurement can be hampered by imprecise criminal code statutes. We quantified the degree of misclassification in Washington state criminal codes for measuring firearm-related crime.

FINDINGS

In this study of individuals aged 18 years and older who were convicted of a misdemeanor in Washington Superior Courts from 1/1/2015 through 12/31/2019, we compared firearm-related charges as measured with criminal codes and with manual review of probable cause documents, considered the gold standard. The sample included 5,390 criminal cases. Of these, 77 (1.4%) were firearm-related as measured with criminal codes and 437 (8.1%) were firearm-related as measured via manual record review. In the sample overall, the sensitivity of criminal codes was 17.6% (95% CI 14.2-21.5%), and negative predictive value (NPV) was 93.2% (95% CI 92.5-93.9%). Sensitivity and NPV were higher for cases with exclusively non-violent charges. For all cases and for cases with any violent crime charge, firearm-related crimes described in probable cause documents most often involved explicit verbal threats, firearm possession, and pointing a firearm at or touching a firearm to someone; almost 10% of all cases involved shooting/discharging a firearm. For cases with exclusively non-violent charges, the most common firearm-related crime was unlawful possession.

CONCLUSIONS

Criminal records can be used for large-scale policy-relevant studies of firearm-related harms, but this study suggests Washington state criminal codes substantially undercount firearm-related crime, especially firearm-related violent crime.

Addressing the unique challenges in studies of long-term medication use and dementia risk

This editorial comments on the article by Wu et al. in this issue.

A causal roadmap for generating high-quality real-world evidence

Increasing emphasis on the use of real-world evidence (RWE) to support clinical policy and regulatory decision-making has led to a proliferation of guidance, advice, and frameworks from regulatory agencies, academia, professional societies, and industry. A broad spectrum of studies use real-world data (RWD) to produce RWE, ranging from randomized trials with outcomes assessed using RWD to fully observational studies. Yet, many proposals for generating RWE lack sufficient detail, and many analyses of RWD suffer from implausible assumptions, other methodological flaws, or inappropriate interpretations. The Causal Roadmap is an explicit, itemized, iterative process that guides investigators to prespecify study design and analysis plans; it addresses a wide range of guidance within a single framework. By supporting the transparent evaluation of causal assumptions and facilitating objective comparisons of design and analysis choices based on prespecified criteria, the Roadmap can help investigators to evaluate the quality of evidence that a given study is likely to produce, specify a study to generate high-quality RWE, and communicate effectively with regulatory agencies and other stakeholders. This paper aims to disseminate and extend the Causal Roadmap framework for use by clinical and translational researchers; three companion papers demonstrate applications of the Causal Roadmap for specific use cases.

Reverse causation biases weighted cumulative exposure model estimates, but can be investigated in sensitivity analyses

OBJECTIVES

To examine the effects of reverse causation on estimates from the weighted cumulative exposure (WCE) model that is used in pharmacoepidemiology to explore drug-health outcome associations, and to identify sensitivity analyses for revealing such effects.

STUDY DESIGN AND SETTING

314,099 patients with diabetes under Clalit Health Services, Israel, were followed over 2002-2012. The association between metformin and pancreatic cancer (PC) was explored using a WCE model within the framework of discrete-time Cox regression. We used computer simulations to explore the effects of reverse causation on estimates of a WCE model and to examine sensitivity analyses for revealing and adjusting for reverse causation. We then applied those sensitivity analyses to our data.

RESULTS

Simulation demonstrated bias in the weighted cumulative exposure model and showed that sensitivity analysis could reveal and adjust for these biases. In our data, a positive association was observed (hazard ratio (HR) = 3.24, 95% confidence interval (CI): 2.24-4.73) with metformin exposure in the previous 2 years. After applying sensitivity analysis, assuming reverse causation operated up to 4 years before cancer diagnosis, the association between metformin and PC was no longer apparent.

CONCLUSION

Reverse causation can cause substantial bias in the WCE model. When suspected, sensitivity analyses based on causal analysis are advocated.

Epidemiologic trends and risk factors associated with the decline in mortality from coronary heart disease in the United States, 1990-2019

BACKGROUND

Despite the decline in the rate of coronary heart disease (CHD) mortality, it is unknown how the 3 strong and modifiable risk factors - alcohol, smoking, and obesity -have impacted these trends. We examine changes in CHD mortality rates in the United States and estimate the preventable fraction of CHD deaths by eliminating CHD risk factors.

METHODS

We performed a sequential time-series analysis to examine mortality trends among females and males aged 25 to 84 years in the United States, 1990-2019, with CHD recorded as the underlying cause of death. We also examined mortality rates from chronic ischemic heart disease (IHD), acute myocardial infarction (AMI), and atherosclerotic heart disease (AHD). All underlying causes of CHD deaths were classified based on the International Classification of Disease 9th and 10th revisions. We estimated the preventable fraction of CHD deaths attributable to alcohol, smoking, and high body-mass index (BMI) through the Global Burden of Disease.

RESULTS

Among females (3,452,043 CHD deaths; mean [standard deviation, SD] age 49.3 [15.7] years), the age-standardized CHD mortality rate declined from 210.5 in 1990 to 66.8 per 100,000 in 2019 (annual change -4.04%, 95% CI -4.05, -4.03; incidence rate ratio [IRR] 0.32, 95% CI, 0.41, 0.43). Among males (5,572,629 CHD deaths; mean [SD] age 47.9 [15.1] years), the age-standardized CHD mortality rate declined from 442.4 to 156.7 per 100,000 (annual change -3.74%, 95% CI, -3.75, -3.74; IRR 0.36, 95% CI, 0.35, 0.37). A slowing of the decline in CHD mortality rates among younger cohorts was evident. Correction for unmeasured confounders through a quantitative bias analysis slightly attenuated the decline. Half of all CHD deaths could have been prevented with the elimination of smoking, alcohol, and obesity, including 1,726,022 female and 2,897,767 male CHD deaths between 1990 and 2019.

CONCLUSIONS

The decline in CHD mortality is slowing among younger cohorts. The complex dynamics of risk factors appear to shape mortality rates, underscoring the importance of targeted strategies to reduce modifiable risk factors that contribute to CHD mortality.

Simultaneous adjustment of uncontrolled confounding, selection bias and misclassification in multiple-bias modelling

BACKGROUND

Adjusting for multiple biases usually involves adjusting for one bias at a time, with careful attention to the order in which these biases are adjusted. A novel, alternative approach to multiple-bias adjustment involves the simultaneous adjustment of all biases via imputation and/or regression weighting. The imputed value or weight corresponds to the probability of the missing data and serves to 'reconstruct' the unbiased data that would be observed based on the provided assumptions of the degree of bias.

METHODS

We motivate and describe the steps necessary to implement this method. We also demonstrate the validity of this method through a simulation study with an exposure-outcome relationship that is biased by uncontrolled confounding, exposure misclassification, and selection bias.

RESULTS

The study revealed that a non-biased effect estimate can be obtained when correct bias parameters are applied. It also found that incorrect specification of every bias parameter by +/-25% still produced an effect estimate with less bias than the observed, biased effect.

CONCLUSIONS

Simultaneous multi-bias analysis is a useful way of investigating and understanding how multiple sources of bias may affect naive effect estimates. This new method can be used to enhance the validity and transparency of real-world evidence obtained from observational, longitudinal studies.

Risk of Stroke Hospitalization After Infertility Treatment

Importance

Stroke accounts for 7% of pregnancy-related deaths in the US. As the use of infertility treatment is increasing, many studies have sought to characterize the association of infertility treatment with the risk of stroke with mixed results.

Objective

To evaluate the risk of hospitalization from hemorrhagic and ischemic strokes in patients who underwent infertility treatment.

Design, Setting, and Participants

This population-based, retrospective cohort study used data abstracted from the Nationwide Readmissions Database, which stores data from all-payer hospital inpatient stays from 28 states across the US, from 2010 and 2018. Eligible participants included individuals aged 15 to 54 who had a hospital delivery from January to November in a given calendar year, and any subsequent hospitalizations from January to December in the same calendar year of delivery during the study period. Statistical analysis was performed between November 2022 and April 2023.

Exposure

Hospital delivery after infertility treatment (ie, intrauterine insemination, assisted reproductive technology, fertility preservation procedures, or use of a gestational carrier) or after spontaneous conception.

Main Outcomes and Measures

The primary outcome was hospitalization for nonfatal stroke (either ischemic or hemorrhagic stroke) within the first calendar year after delivery. Secondary outcomes included risk of stroke hospitalization at less than 30 days, less than 60 days, less than 90 days, and less than 180 days post partum. Cox proportional hazards regression models were used to estimate associations, which were expressed as hazard ratios (HRs), adjusted for confounders. Effect size estimates were corrected for biases due to exposure misclassification, selection, and unmeasured confounding through a probabilistic bias analysis.

Results

Of 31 339 991 patients, 287 813 (0.9%; median [IQR] age, 32.1 [28.5-35.8] years) underwent infertility treatment and 31 052 178 (99.1%; median [IQR] age, 27.7 [23.1-32.0] years) delivered after spontaneous conception. The rate of stroke hospitalization within 12 months of delivery was 37 hospitalizations per 100 000 people (105 patients) among those who received infertility treatment and 29 hospitalizations per 100 000 people (9027 patients) among those who delivered after spontaneous conception (rate difference, 8 hospitalizations per 100 000 people; 95% CI, -6 to 21 hospitalizations per 100 000 people; HR, 1.66; 95% CI, 1.17 to 2.35). The risk of hospitalization for hemorrhagic stroke (adjusted HR, 2.02; 95% CI, 1.13 to 3.61) was greater than that for ischemic stroke (adjusted HR, 1.55; 95% CI, 1.01 to 2.39). The risk of stroke hospitalization increased as the time between delivery and hospitalization for stroke increased, particularly for hemorrhagic strokes. In general, these associations became larger for hemorrhagic stroke and smaller for ischemic stroke following correction for biases.

Conclusions and Relevance

In this cohort study, infertility treatment was associated with an increased risk of stroke-related hospitalization within 12 months of delivery; this risk was evident as early as 30 days after delivery. Timely follow-up in the immediate days post partum and continued long-term follow-up should be considered to mitigate stroke risk.

Prevalent diabetes and long-term cardiovascular outcomes in adult sepsis survivors: a population-based cohort study

BACKGROUND

Sepsis survivors are at elevated risk for cardiovascular disease during long-term follow-up. Whether diabetes influences cardiovascular risk after sepsis survival remains unknown. We sought to describe the association of diabetes with long-term cardiovascular outcomes in adult sepsis survivors.

METHODS

Population-based cohort study in the province of Ontario, Canada (2008-2017). Adult survivors of a first sepsis-associated hospitalization, without pre-existing cardiovascular disease, were included. Main exposure was pre-existing diabetes (any type). The primary outcome was the composite of myocardial infarction, stroke, and cardiovascular death. Patients were followed up to 5 years from discharge date until outcome occurrence or end of study period (March 2018). We used propensity score matching (i.e., 1:1 to patients with sepsis but no pre-existing diabetes) to adjust for measured confounding at baseline. Cause-specific Cox proportional hazards models with robust standard errors were used to estimate hazard ratios (HR) alongside 95% confidence intervals (CI). A main secondary analysis evaluated the modification of the association between sepsis and cardiovascular disease by pre-existing diabetes.

RESULTS

78,638 patients with pre-existing diabetes who had a sepsis-associated hospitalization were matched to patients hospitalized for sepsis but without diabetes. Mean age of patients was 71 years, and 55% were female. Median duration from diabetes diagnosis was 9.8 years; mean HbA1c was 7.1%. Adult sepsis survivors with pre-existing diabetes experienced a higher hazard of major cardiovascular disease (HR 1.25; 95% CI 1.22-1.29)-including myocardial infarction (HR 1.40; 95% CI 1.34-1.47) and stroke (HR 1.24; 95% CI 1.18-1.29)-during long-term follow-up compared to sepsis survivors without diabetes. Pre-existing diabetes modified the association between sepsis and cardiovascular disease (risk difference: 2.3%; 95% CI 2.0-2.6 and risk difference: 1.8%; 95% CI 1.6-2.0 for the effect of sepsis-compared to no sepsis-among patients with and without diabetes, respectively).

CONCLUSIONS

Sepsis survivors with pre-existing diabetes experience a higher long-term hazard of major cardiovascular events when compared to sepsis survivors without diabetes. Compared to patients without sepsis, the absolute risk increase of cardiovascular events after sepsis is higher in patients with diabetes (i.e., diabetes intensified the higher cardiovascular risk induced by sepsis).

Effect of alcohol consumption on breast cancer: probabilistic bias analysis for adjustment of exposure misclassification bias and confounders

PURPOSE

This study was conducted to evaluate the effect of alcohol consumption on breast cancer, adjusting for alcohol consumption misclassification bias and confounders.

METHODS

This was a case-control study of 932 women with breast cancer and 1000 healthy control. Using probabilistic bias analysis method, the association between alcohol consumption and breast cancer was adjusted for the misclassification bias of alcohol consumption as well as a minimally sufficient set of adjustment of confounders derived from a causal directed acyclic graph. Population attributable fraction was estimated using the Miettinen's Formula.

RESULTS

Based on the conventional logistic regression model, the odds ratio estimate between alcohol consumption and breast cancer was 1.05 (95% CI: 0.57, 1.91). However, the adjusted estimates of odds ratio based on the probabilistic bias analysis ranged from 1.82 to 2.29 for non-differential and from 1.93 to 5.67 for differential misclassification. Population attributable fraction ranged from 1.51 to 2.57% using non-differential bias analysis and 1.54-3.56% based on differential bias analysis.

CONCLUSION

A marked measurement error was in self-reported alcohol consumption so after correcting misclassification bias, no evidence against independence between alcohol consumption and breast cancer changed to a substantial positive association.

Outcomes of patients with respiratory failure declined for extracorporeal membrane oxygenation: a prospective observational study

PURPOSE

Descriptive information on referral patterns and short-term outcomes of patients with respiratory failure declined for extracorporeal membrane oxygenation (ECMO) is lacking.

METHODS

We conducted a prospective single-centre observational cohort study of ECMO referrals to Toronto General Hospital (receiving hospital) for severe respiratory failure (COVID-19 and non-COVID-19), between 1 December 2019 and 30 November 2020. Data related to the referral, the referral decision, and reasons for refusal were collected. Reasons for refusal were grouped into three mutually exclusive categories selected a priori: "too sick now," "too sick before," and "not sick enough." In declined referrals, referring physicians were surveyed to collect patient outcome on day 7 after the referral. The primary study endpoints were referral outcome (accepted/declined) and patient outcome (alive/deceased).

RESULTS

A total of 193 referrals were included; 73% were declined for transfer. Referral outcome was influenced by age (odds ratio [OR], 0.97; 95% confidence interval [CI], 0.95 to 0.96; P < 0.01) and involvement of other members of the ECMO team in the discussion (OR, 4.42; 95% CI, 1.28 to 15.2; P < 0.01). Patient outcomes were missing in 46 (24%) referrals (inability to locate the referring physician or the referring physician being unable to recall the outcome). Using available data (95 declined and 52 accepted referrals; n = 147), survival to day 7 was 49% for declined referrals (35% for patients deemed "too sick now," 53% for "too sick before," 100% for "not sick enough," and 50% for reason for refusal not reported) and 98% for transferred patients. Sensitivity analysis setting missing outcomes to directional extreme values retained robustness of survival probabilities.

CONCLUSION

Nearly half of the patients declined for ECMO consideration were alive on day 7. More information on patient trajectory and long-term outcomes in declined referrals is needed to refine selection criteria.

Exploring causality from observational data: An example assessing whether religiosity promotes cooperation

Causal inference from observational data is notoriously difficult, and relies upon many unverifiable assumptions, including no confounding or selection bias. Here, we demonstrate how to apply a range of sensitivity analyses to examine whether a causal interpretation from observational data may be justified. These methods include: testing different confounding structures (as the assumed confounding model may be incorrect), exploring potential residual confounding and assessing the impact of selection bias due to missing data. We aim to answer the causal question 'Does religiosity promote cooperative behaviour?' as a motivating example of how these methods can be applied. We use data from the parental generation of a large-scale (n = approximately 14,000) prospective UK birth cohort (the Avon Longitudinal Study of Parents and Children), which has detailed information on religiosity and potential confounding variables, while cooperation was measured via self-reported history of blood donation. In this study, there was no association between religious belief or affiliation and blood donation. Religious attendance was positively associated with blood donation, but could plausibly be explained by unmeasured confounding. In this population, evidence that religiosity causes blood donation is suggestive, but rather weak. These analyses illustrate how sensitivity analyses can aid causal inference from observational research.

There is a need for more precise models to assess the determinants of health crises like COVID-19

The COVID-19 pandemic has had a significant impact on global mortality. While the causal relationship between SARS-CoV-2 and the anomalous increase in deaths is established, more precise and complex models are needed to determine the exact weight of epidemiological factors involved. Indeed, COVID-19 behavior is influenced by a wide range of variables, including demographic characteristics, population habits and behavior, healthcare performance, and environmental and seasonal risk factors. The bidirectional causality between impacted and impacting aspects, as well as confounding variables, complicates efforts to draw clear, generalizable conclusions regarding the effectiveness and cost-benefit ratio of non-pharmaceutical health countermeasures. Thus, it is imperative that the scientific community and health authorities worldwide develop comprehensive models not only for the current pandemic but also for future health crises. These models should be implemented locally to account for micro-differences in epidemiological characteristics that may have relevant effects. It is important to note that the lack of a universal model does not imply that local decisions have been unjustified, and the request to decrease scientific uncertainty does not mean denying the evidence of the effectiveness of the countermeasures adopted. Therefore, this paper must not be exploited to denigrate either the scientific community or the health authorities.