Potential sensitivity of bias analysis results to incorrect assumptions of nondifferential or differential binary exposure misclassification

Evaluation of quantitative bias analysis in epidemiological research: A systematic review from 2010 to mid-2023

OBJECTIVE

We aimed to demonstrate the use of quantitative bias analysis (QBA), which reveals the effects of systematic error, including confounding, misclassification and selection bias, on study results in epidemiological studies published in the period from 2010 to mid-23.

METHOD

The articles identified through a keyword search using Pubmed and Scopus were included in the study. The articles obtained from this search were eliminated according to the exclusion criteria, and the articles in which QBA analysis was applied were included in the detailed evaluation.

RESULTS

It can be said that the application of QBA analysis has gradually increased over the 13-year period. Accordingly, the number of articles in which simple is used as a method in QBA analysis is 9 (9.89%), the number of articles in which the multidimensional approach is used is 10 (10.99%), the number of articles in which the probabilistic approach is used is 60 (65.93%) and the number of articles in which the method is not specified is 12 (13.19%). The number of articles with misclassification bias model is 44 (48.35%), the number of articles with uncontrolled confounder(s) bias model is 32 (35.16%), the number of articles with selection bias model is 7 (7.69%) and the number of articles using more than one bias model is 8 (8.79%). Of the 49 (53.85%) articles in which the bias parameter source was specified, 19 (38.78%) used internal validation, 26 (53.06%) used external validation and 4 (8.16%) used educated guess, data constraints and hypothetical data. Probabilistic approach was used as a bias method in 60 (65.93%) of the articles, and mostly beta (8 [13.33%)], normal (9 [15.00%]) and uniform (8 [13.33%]) distributions were selected.

CONCLUSION

The application of QBA is rare in the literature but is increasing over time. Future researchers should include detailed analyzes such as QBA analysis to obtain inferences with higher evidence value, taking into account systematic errors.

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.

Evaluating surveillance definitions of HIV viral suppression 2015-2019: Which definition best detected barriers to care?

Background

People living with HIV (PLWH) who have not achieved or maintained viral suppression post-diagnosis likely face multiple barriers to HIV care. To identify these barriers a universally accepted definition of viral suppression is needed. The most common definition, the Center for Disease Control and Prevention (CDC) definition, contains simplifying assumptions that may misclassify individuals and attenuate associations. In this study, we evaluated alternative definitions of viral suppression on their ability to identify barriers to care.

Design and methods

We used HIV surveillance data to classify participants of the 2015-2019 Washington Medical Monitoring Project (MMP) as virally suppressed or not using the CDC definition and two definitions that assess viral suppression over a longer period ("Enriched" and "Durable"). We identified barriers to suppression from literature (unstable housing, illicit drug use, poor mental health, heavy drinking, recent incarceration, racism, and poverty) and measured them using interview questions from MMP. We compared the rate ratios (RR) of being not virally suppressed using each definition for each barrier.

Results

There were 858 PLWH in our study. All viral suppression definitions classified a similar proportion of people as suppressed (85%-89%). The durable viral suppression definition consistently yielded the largest rate ratios (e.g. unstable housing: CDC RR = 1.3, 95% CI 0.9-1.8; Enriched 1.5, 95% CI 1.0-2.2; Durable 2.2, 95% CI 1.6-3.1) and reclassified 10% of the population relative to the CDC definition.

Conclusions

Longitudinal definitions for viral suppression may yield less misclassification and serve as superior tools for identifying and curtailing barriers to HIV care.

e-Cigarette Use and Combustible Cigarette Smoking Initiation Among Youth: Accounting for Time-Varying Exposure and Time-Dependent Confounding

BACKGROUND

Youth e-cigarette use is associated with the initiation of combustible cigarette smoking, but prior studies have rarely accounted for time-varying measures of e-cigarette exposure or time-dependent confounding of e-cigarette use and smoking initiation.

METHODS

Using five waves of the Population Assessment of Tobacco and Health (2013-2019), we estimated marginal structural models with inverse probability of treatment and censoring weights to examine the association between time-varying e-cigarette initiation and subsequent cigarette smoking initiation among e-cigarette- and cigarette-naïve youth (12-17 years) at baseline. Time-dependent confounders used as predictors in inverse probability weights included tobacco-related attitudes or beliefs, mental health symptoms, substance use, and tobacco-marketing exposure.

RESULTS

Among 9,584 youth at baseline, those who initiated e-cigarettes were 2.4 times as likely to subsequently initiate cigarette smoking as youth who did not initiate e-cigarettes (risk ratio = 2.4, 95% confidence interval [CI] = 2.1, 2.7), after accounting for time-dependent confounding and selection bias. Among youth who initiated e-cigarettes, more frequent vaping was associated with greater risk of smoking initiation (risk ratio ≥3 days/month = 1.8, 95% CI = 1.4, 2.2; 1-2 days/month = 1.2; 95% CI = 0.93, 1.6 vs. 0 days/month). Weighted marginal structural model estimates were moderately attenuated compared with unweighted estimates adjusted for baseline-only confounders. At the US population level, we estimated over half a million youth initiated cigarette smoking because of prior e-cigarette use over follow-up.

CONCLUSIONS

The association between youth vaping and combustible cigarette smoking persisted after accounting for time-dependent confounding. We estimate that e-cigarette use accounts for a considerable share of cigarette initiation among US youth. See video abstract at, http://links.lww.com/EDE/B937.

A Bayesian approach for estimating the partial potential impact fraction with exposure measurement error under a main study/internal validation design

The partial potential impact fraction describes the proportion of disease cases that can be prevented if the distribution of modifiable continuous exposures is shifted in a population, while other risk factors are not modified. It is a useful quantity for evaluating the burden of disease in epidemiologic and public health studies. When exposures are measured with error, the partial potential impact fraction estimates may be biased, which necessitates methods to correct for the exposure measurement error. Motivated by the health professionals follow-up study, we develop a Bayesian approach to adjust for exposure measurement error when estimating the partial potential impact fraction under the main study/internal validation study design. We adopt the reclassification approach that leverages the strength of the main study/internal validation study design and clarifies transportability assumptions for valid inference. We assess the finite-sample performance of both the point and credible interval estimators via extensive simulations and apply the proposed approach in the health professionals follow-up study to estimate the partial potential impact fraction for colorectal cancer incidence under interventions exploring shifting the distributions of red meat, alcohol, and/or folate intake.

Estimation and inference for the population attributable risk in the presence of misclassification

Because it describes the proportion of disease cases that could be prevented if an exposure were entirely eliminated from a target population as a result of an intervention, estimation of the population attributable risk (PAR) has become an important goal of public health research. In epidemiologic studies, categorical covariates are often misclassified. We present methods for obtaining point and interval estimates of the PAR and the partial PAR (pPAR) in the presence of misclassification, filling an important existing gap in public health evaluation methods. We use a likelihood-based approach to estimate parameters in the models for the disease and for the misclassification process, under main study/internal validation study and main study/external validation study designs, and various plausible assumptions about transportability. We assessed the finite sample perf ormance of this method via a simulation study, and used it to obtain corrected point and interval estimates of the pPAR for high red meat intake and alcohol intake in relation to colorectal cancer incidence in the HPFS, where we found that the estimated pPAR for the two risk factors increased by up to 317% after correcting for bias due to misclassification.

Chilblains during lockdown are associated with household exposure to SARS-CoV-2: a multicentre case-control study

Objectives

During the COVID-19 pandemic, numerous cases of chilblains have been reported. However, in most cases, RT-PCR or serology did not confirm SARS-CoV-2 infection. Hypotheses have been raised about an interferon-mediated immunological response to SARS-CoV-2, leading to effective clearance of the SARS-CoV-2 without the involvement of humoral immunity. Our objective was to explore the association between chilblains and exposure to SARS-CoV-2.

Methods

In this multicentre case–control study, cases were the 102 individuals referred to five referral hospitals for chilblains occurring during the first lockdown (March to May 2020). Controls were recruited from healthy volunteers' files held by the same hospitals. All members of their households were included, resulting in 77 case households (262 individuals) and 74 control households (230 individuals). Household exposure to SARS-CoV-2 during the first lockdown was categorized as high, intermediate or low, using a pre-established algorithm based on individual data on symptoms, high-risk contacts, activities outside the home and RT-PCR testing. Participants were offered a SARS-CoV-2 serological test.

Results

After adjustment for age, the association between chilblains and viral exposure was estimated at OR 3.3, 95% CI (1.4–7.3) for an intermediate household exposure, and 6.9 (2.5–19.5) for a high household exposure to SARS-CoV-2. Out of 57 case households tested, six (11%) had positive serology for SARS-CoV-2, whereas all control households tested (n = 50) were seronegative (p = 0.03). The effect of potential misclassification on exposure has been assessed in a bias analysis.

Discussion

This case–control study demonstrates the association between chilblains occurring during the lockdown and household exposure to SARS-CoV-2.

Simulation of Random Differential Periodontitis Outcome Misclassification with Perfect Specificity

INTRODUCTION

Misclassification of clinical periodontitis can occur by partial-mouth protocols, particularly when tooth-based case definitions are applied. In these cases, the true prevalence of periodontal disease is underestimated, but specificity is perfect. In association studies of periodontal disease etiology, misclassification by this mechanism is independent of exposure status (i.e., nondifferential). Despite nondifferential mechanisms, differential misclassification may be realized by virtue of random errors.

OBJECTIVES

To gauge the amount of uncertainty around the expectation of differential periodontitis outcome misclassification due to random error only, we estimated the probability of differential outcome misclassification, its magnitude, and expected impacts via simulation methods using values from the periodontitis literature.

METHODS

We simulated data sets with a binary exposure and outcome that varied according to sample size (200, 1,000, 5,000, 10,000), exposure effect (risk ratio; 1.5, 2), exposure prevalence (0.1, 0.3), outcome incidence (0.1, 0.4), and outcome sensitivity (0.6, 0.8). Using a Bernoulli trial, we introduced misclassification by randomly sampling individuals with the outcome in each exposure group and repeated each scenario 10,000 times.

RESULTS

The probability of differential misclassification decreased as the simulation parameter values increased and occurred at least 37% of the time across the 10,000 repetitions. Across all scenarios, the risk ratio was biased, on average, toward the null when the sensitivity was higher among the unexposed and away from the null when it was higher among the exposed. The extent of bias for absolute sensitivity differences ≥0.04 ranged from 0.05 to 0.19 regardless of simulation parameters. However, similar trends were not observed for the odds ratio where the extent and direction of bias were dependent on the outcome incidence, sensitivity of classification, and effect size.

CONCLUSIONS

The results of this simulation provide helpful quantitative information to guide interpretation of findings in which nondifferential outcome misclassification mechanisms are known to be operational with perfect specificity.

KNOWLEDGE TRANSFER STATEMENT

Measurement of periodontitis can suffer from classification errors, such as when partial-mouth protocols are applied. In this case, specificity is perfect and sensitivity is expected to be nondifferential, leading to an expectation for no bias when studying periodontitis etiologies. Despite expectation, differential misclassification could occur from sources of random error, the effects of which are unknown. Proper scrutiny of research findings can occur when the probability and impact of random classification errors are known.

A systematic review of quantitative bias analysis applied to epidemiological research

BACKGROUND

Quantitative bias analysis (QBA) measures study errors in terms of direction, magnitude and uncertainty. This systematic review aimed to describe how QBA has been applied in epidemiological research in 2006-19.

METHODS

We searched PubMed for English peer-reviewed studies applying QBA to real-data applications. We also included studies citing selected sources or which were identified in a previous QBA review in pharmacoepidemiology. For each study, we extracted the rationale, methodology, bias-adjusted results and interpretation and assessed factors associated with reproducibility.

RESULTS

Of the 238 studies, the majority were embedded within papers whose main inferences were drawn from conventional approaches as secondary (sensitivity) analyses to quantity-specific biases (52%) or to assess the extent of bias required to shift the point estimate to the null (25%); 10% were standalone papers. The most common approach was probabilistic (57%). Misclassification was modelled in 57%, uncontrolled confounder(s) in 40% and selection bias in 17%. Most did not consider multiple biases or correlations between errors. When specified, bias parameters came from the literature (48%) more often than internal validation studies (29%). The majority (60%) of analyses resulted in >10% change from the conventional point estimate; however, most investigators (63%) did not alter their original interpretation. Degree of reproducibility related to inclusion of code, formulas, sensitivity analyses and supplementary materials, as well as the QBA rationale.

CONCLUSIONS

QBA applications were rare though increased over time. Future investigators should reference good practices and include details to promote transparency and to serve as a reference for other researchers.

A primer on quantitative bias analysis with positive predictive values in research using electronic health data

OBJECTIVE

In health informatics, there have been concerns with reuse of electronic health data for research, including potential bias from incorrect or incomplete outcome ascertainment. In this tutorial, we provide a concise review of predictive value-based quantitative bias analysis (QBA), which comprises epidemiologic methods that use estimates of data quality accuracy to quantify the bias caused by outcome misclassification.

TARGET AUDIENCE

Health informaticians and investigators reusing large, electronic health data sources for research.

SCOPE

When electronic health data are reused for research, validation of outcome case definitions is recommended, and positive predictive values (PPVs) are the most commonly reported measure. Typically, case definitions with high PPVs are considered to be appropriate for use in research. However, in some studies, even small amounts of misclassification can cause bias. In this tutorial, we introduce methods for quantifying this bias that use predictive values as inputs. Using epidemiologic principles and examples, we first describe how multiple factors influence misclassification bias, including outcome misclassification levels, outcome prevalence, and whether outcome misclassification levels are the same or different by exposure. We then review 2 predictive value-based QBA methods and why outcome PPVs should be stratified by exposure for bias assessment. Using simulations, we apply and evaluate the methods in hypothetical electronic health record-based immunization schedule safety studies. By providing an overview of predictive value-based QBA, we hope to bridge the disciplines of health informatics and epidemiology to inform how the impact of data quality issues can be quantified in research using electronic health data sources.

Using datasets to ascertain the generalisability of clinical cohorts: the example of European QUALity Study on the treatment of advanced chronic kidney disease (EQUAL)

BACKGROUND

Cohort studies are among the most robust of observational studies but have issues with external validity. This study assesses threats to external validity (generalisability) in the European QUALity (EQUAL) study, a cohort study of people over 65 years with stage 4/5 chronic kidney disease.

METHODS

Patients meeting the EQUAL inclusion criteria were identified in The Health Improvement Network database and stratified into those attending renal units (secondary care cohort-SCC) and not (primary care cohort-PCC). Survival, progression to renal replacement therapy (RRT), and hospitalisation were compared.

RESULTS

The analysis included 250, 633, and 2,464 patients in EQUAL, PCC, and SCC. EQUAL had a higher proportion of men in comparison to PCC and SCC (60.0% vs. 34.8% vs. 51.4%). Increasing age (≥85 years odds ratio (OR) 0.25 (95% confidence interval (CI) 0.15-0.40)) and comorbidity (Charlson Comorbidity Index ≥ 4 OR 0.69 (CI 0.52-0.91)) were associated with non-participation in EQUAL. EQUAL had a higher proportion of patients starting RRT at 1 year compared to SCC (8.1% vs. 2.1%%, p < 0.001). Patients in the PCC and SCC had increased risk of Hospitalisation (incidence rate ratio=1.76 (95% CI 1.27-2.47) & 2.13 (95% CI 1.59-2.86)) and mortality at one year (hazard ratio=3.48 (95% CI 2.1-5.7) & 1.7 (95% CI 1.1-2.7)) compared to EQUAL.

CONCLUSIONS

This study provides evidence of how participants in a cohort study can differ from the broader population of patients, which is essential when considering external validity and applying to local practice.

Characterizing Bias Due to Differential Exposure Ascertainment in Electronic Health Record Data

Data derived from electronic health records (EHR) are heterogeneous with availability of specific measures dependent on the type and timing of patients' healthcare interactions. This creates a challenge for research using EHR-derived exposures because gold-standard exposure data, determined by a definitive assessment, may only be available for a subset of the population. Alternative approaches to exposure ascertainment in this case include restricting the analytic sample to only those patients with gold-standard exposure data available (exclusion); using gold-standard data, when available, and using a proxy exposure measure when the gold standard is unavailable (best available); or using a proxy exposure measure for everyone (common data). Exclusion may induce selection bias in outcome/exposure association estimates, while incorporating information from a proxy exposure via either the best available or common data approaches may result in information bias due to measurement error. The objective of this paper was to explore the bias and efficiency of these three analytic approaches across a broad range of scenarios motivated by a study of the association between chronic hyperglycemia and five-year mortality in an EHR-derived cohort of colon cancer survivors. We found that the best available approach tended to mitigate inefficiency and selection bias resulting from exclusion while suffering from less information bias than the common data approach. However, bias in all three approaches can be severe, particularly when both selection bias and information bias are present. When risk of either of these biases is judged to be more than moderate, EHR-based analyses may lead to erroneous conclusions.

Age and the association between apolipoprotein E genotype and Alzheimer disease: A cerebrospinal fluid biomarker-based case-control study

BACKGROUND

The ε4 allele of apolipoprotein E (APOE) gene and increasing age are two of the most important known risk factors for developing Alzheimer disease (AD). The diagnosis of AD based on clinical symptoms alone is known to have poor specificity; recently developed diagnostic criteria based on biomarkers that reflect underlying AD neuropathology allow better assessment of the strength of the associations of risk factors with AD. Accordingly, we examined the global and age-specific association between APOE genotype and AD by using the A/T/N classification, relying on the cerebrospinal fluid (CSF) levels of β-amyloid peptide (A, β-amyloid deposition), phosphorylated tau (T, pathologic tau), and total tau (N, neurodegeneration) to identify patients with AD.

METHODS AND FINDINGS

This case-control study included 1,593 white AD cases (55.4% women; mean age 72.8 [range = 44-96] years) with abnormal values of CSF biomarkers from nine European memory clinics and the American Alzheimer's Disease Neuroimaging Initiative (ADNI) study. A total of 11,723 dementia-free controls (47.1% women; mean age 65.6 [range = 44-94] years) were drawn from two longitudinal cohort studies (Whitehall II and Three-City), in which incident cases of dementia over the follow-up were excluded from the control population. Odds ratio (OR) and population attributable fraction (PAF) for AD associated with APOE genotypes were determined, overall and by 5-year age categories. In total, 63.4% of patients with AD and 22.6% of population controls carried at least one APOE ε4 allele. Compared with non-ε4 carriers, heterozygous ε4 carriers had a 4.6 (95% confidence interval 4.1-5.2; p < 0.001) and ε4/ε4 homozygotes a 25.4 (20.4-31.2; p < 0.001) higher OR of AD in unadjusted analysis. This association was modified by age (p for interaction < 0.001). The PAF associated with carrying at least one ε4 allele was greatest in the 65-70 age group (69.7%) and weaker before 55 years (14.2%) and after 85 years (22.6%). The protective effect of APOE ε2 allele for AD was unaffected by age. Main study limitations are that analyses were based on white individuals and AD cases were drawn from memory centers, which may not be representative of the general population of patients with AD.

CONCLUSIONS

In this study, we found that AD diagnosis based on biomarkers was associated with APOE ε4 carrier status, with a higher OR than previously reported from studies based on only clinical AD criteria. This association differs according to age, with the strongest effect at 65-70 years. These findings highlight the need for early interventions for dementia prevention to mitigate the effect of APOE ε4 at the population level.

A gestational vulnerability window for smoking exposure and the increased risk of preterm birth: how timing and intensity of maternal smoking matter

BACKGROUND

Reducing the incidence of preterm birth is a national priority. Maternal cigarette smoking is strongly and consistently associated with preterm birth. The objective of this study was to examine prenatal exposure based on combined measures of timing (by trimester) and intensity level (the number of cigarettes smoked per day) of maternal smoking to identify a pregnancy period with the highest risk of preterm birth.

METHODS

A sample of 2,485,743 singleton births was drawn from the 2010 National Center of Health Statistics (NCHS) linked birth/infant death file of US residents in 33 states that implemented the revised 2003 birth certificate. Nine mutually exclusive smoking status categories were created to assess prenatal exposure across pregnancy in association with preterm birth. Gestational age was based on the obstetric estimate. Multiple logistic regression analyses were conducted to compare the odds of preterm birth among women who smoked at different intensity levels in the second or third trimester with those who smoked only in the first trimester.

RESULTS

Overall, 7.95% of women had a preterm birth; 8.90% of low intensity (less than a pack/day) smokers in the first trimester only, 12.99% of low and 15.38% of high intensity (pack a day or more) smokers in the first two trimesters, and 10.56% of low and 11.35% of high intensity smokers in all three trimesters delivered preterm. First and second trimester high (aOR 1.85, 95% CI: 1.66, 2.06) and low intensity smokers (aOR 1.51, 95% CI: 1.41, 1.61) had higher odds of preterm birth compared to those who smoked less than a pack a day only in the first trimester, but the odds did not increase for all three trimester smokers relative to the first and second trimester smokers. In sensitivity analysis, adjustment for exposure misclassification error corrected data and testing for effect modification by maternal race/ethnicity found no significant interaction.

CONCLUSIONS

This study documented a biologically plausible vulnerability window for smoking exposure and the increased risk of preterm birth. For women who do not modify their smoking behavior preconception, preterm birth risk of smoking remains low until late in the first trimester.

Multiple bias analysis using logistic regression: an example from the National Birth Defects Prevention Study

PURPOSE

Exposure misclassification, selection bias, and confounding are important biases in epidemiologic studies, yet only confounding is routinely addressed quantitatively. We describe how to combine two previously described methods and adjust for multiple biases using logistic regression.

METHODS

Weights were created from selection probabilities and predictive values for exposure classification and applied to multivariable logistic regression models in a case-control study of prepregnancy obesity (body mass index ≥30 vs. <30 kg/m²) and cleft lip with or without cleft palate (CL/P) using data from the National Birth Defects Prevention Study (2523 cases, 10,605 controls).

RESULTS

Adjusting for confounding by race/ethnicity, prepregnancy obesity, and CL/P were weakly associated (odds ratio [OR]: 1.10; 95% confidence interval: 0.98, 1.23). After weighting the data to account for exposure misclassification, missing exposure data, selection bias, and confounding, multiple bias-adjusted ORs ranged from 0.94 to 1.03 in nonprobabilistic bias analyses and median multiple bias-adjusted ORs ranged from 0.93 to 1.02 in probabilistic analyses.

CONCLUSIONS

This approach, adjusting for multiple biases using a logistic regression model, suggested that the observed association between obesity and CL/P could be due to the presence of bias.

Bayesian Correction for Misclassification in Multilevel Count Data Models

Covariate misclassification is well known to yield biased estimates in single level regression models. The impact on hierarchical count models has been less studied. A fully Bayesian approach to modeling both the misclassified covariate and the hierarchical response is proposed. Models with a single diagnostic test and with multiple diagnostic tests are considered. Simulation studies show the ability of the proposed model to appropriately account for the misclassification by reducing bias and improving performance of interval estimators. A real data example further demonstrated the consequences of ignoring the misclassification. Ignoring misclassification yielded a model that indicated there was a significant, positive impact on the number of children of females who observed spousal abuse between their parents. When the misclassification was accounted for, the relationship switched to negative, but not significant. Ignoring misclassification in standard linear and generalized linear models is well known to lead to biased results. We provide an approach to extend misclassification modeling to the important area of hierarchical generalized linear models.

Short communication: Model for metritis severity predicts that disease misclassification underestimates projected milk production losses

The objective of this research was to determine the effect of disease misclassification on the estimated effect of metritis on milk production. Misclassification introduces bias that usually results in an underestimation of the association between exposure (disease) and the outcome of interest (milk production). This distorted measure of association results from the comparison of an affected population (some of which may not truly be affected) to a nonaffected population (which often includes affected subjects that are unidentified). A convenience sample of DairyComp305 (Valley Agricultural Software, Tulare, CA) data representing 1 yr of calvings (n = 3,277) from 1 Midwestern Holstein herd was used. This herd was chosen because of its ongoing efforts to consistently and completely record all clinical diseases, including the incidence of both mild and severe metritis cases. Metritis was defined as the presence of a flaccid uterus containing fetid fluids or a foul watery discharge within 14 d of calving. Cows that appeared clinically normal other than the discharge were considered mild and those with systemic signs of disease were classified as severe. The original data set included metritis recorded as mild, severe, or not recorded (NR), where no metritis was observed, and was considered to contain the metritis true severity (TrS). First, to evaluate the effect of misclassification bias, we retrospectively randomized 45% of mild metritis to be classified as NR to simulate inconsistent disease recording (IR); then, in a separate model, all mild metritis cases were changed to NR to simulate a situation of very poor disease recording (PR), where only the most severe cases are recorded. The TrS, IR, and PR data sets were analyzed separately in JMP (SAS Institute Inc., Cary, NC). An ANOVA was conducted for second test 305-d mature-equivalent milk projection (2nd305ME), and nonsignificant variables were removed, but the variable metritis was forced into all models. Based upon the TrS model, adjusting for effects of lactation group, month of calving, dystocia, twins, retained placenta, early-lactation mastitis, displaced abomasum, and significant interactions, a case of mild metritis was associated with 384 kg less 2nd305ME and a case of severe metritis was associated with 847 kg less 2nd305ME compared with no metritis. For the IR model, a case of mild metritis was associated with 315 kg less 2nd305ME and a case of severe metritis was associated with 758 kg less 2nd305ME compared with no metritis. For the PR model, severe metritis was associated with 680 kg less 2nd305ME compared with NR. The IR and PR models underestimated 2nd305ME loss for severe metritis cases by 89 and 166 kg/cow, and resulted in 180,441 and 330,256 kg of total milk loss unaccounted for at the herd level, respectively, compared with TrS. Overall, misclassification of metritis cases results in greater bias and largely underestimates the true association between metritis and the consequence costs of the disease.

Bias from outcome misclassification in immunization schedule safety research

PURPOSE

The Institute of Medicine recommended conducting observational studies of childhood immunization schedule safety. Such studies could be biased by outcome misclassification, leading to incorrect inferences. Using simulations, we evaluated (1) outcome positive predictive values (PPVs) as indicators of bias of an exposure-outcome association, and (2) quantitative bias analyses (QBA) for bias correction.

METHODS

Simulations were conducted based on proposed or ongoing Vaccine Safety Datalink studies. We simulated 4 studies of 2 exposure groups (children with no vaccines or on alternative schedules) and 2 baseline outcome levels (100 and 1000/100 000 person-years), with 3 relative risk (RR) levels (RR = 0.50, 1.00, and 2.00), across 1000 replications using probabilistic modeling. We quantified bias from non-differential and differential outcome misclassification, based on levels previously measured in database research (sensitivity > 95%; specificity > 99%). We calculated median outcome PPVs, median observed RRs, Type 1 error, and bias-corrected RRs following QBA.

RESULTS

We observed PPVs from 34% to 98%. With non-differential misclassification and true RR = 2.00, median bias was toward the null, with severe bias (median observed RR = 1.33) with PPV = 34% and modest bias (median observed RR = 1.83) with PPV = 83%. With differential misclassification, PPVs did not reflect median bias, and there was Type 1 error of 100% with PPV = 90%. QBA was generally effective in correcting misclassification bias.

CONCLUSIONS

In immunization schedule studies, outcome misclassification may be non-differential or differential to exposure. Overall outcome PPVs do not reflect the distribution of false positives by exposure and are poor indicators of bias in individual studies. Our results support QBA for immunization schedule safety research.

A probabilistic bias analysis for misclassified categorical exposures, with application to oral anti-hyperglycaemic drugs

PURPOSE

The effect of drug exposure misclassification generally receives little attention in pharmacoepidemiological research. In this paper, we illustrate a probabilistic bias analysis approach for misclassified categorical exposures and apply it in a database study of oral anti-hyperglycaemic drugs (OADs).

METHODS

A cohort study based on the Health Search Database general-practice database was carried out by including 12 640 adult (≥40 years) patients newly treated with OADs during 2003-2010. The proportion of days covered by OADs prescriptions during the first year of follow-up was evaluated for each individual, either by means of the prescribed daily dose or the defined daily dose. The effect of misclassification on hypothetical OAD-outcome association profiles was assessed through the proposed probabilistic bias analysis approach, taking advantage of available exposure validation data.

RESULTS

During the first year of follow-up, the average (SD) number of months with OADs available was 7 (4) months and 5 (3) months according to the prescribed daily dose and defined daily dose metrics, respectively. Probabilistic bias analysis results based on validation data suggest that the effect of misclassification is complex, as conventional exposure-outcome association estimates may be of greater or lower magnitude than their misclassification-adjusted values.

CONCLUSIONS

Misclassification should be taken into account in database studies on the safety of prescribed medications. To this aim, investigators should take advantage of external exposure validation data in sensitivity analysis approaches such as ours. Copyright © 2016 John Wiley & Sons, Ltd.

Caffeine: Friend or Foe?

The debate on the safety of and regulatory approaches for caffeine continues among various stakeholders and regulatory authorities. This decision-making process comes with significant challenges, particularly when considering the complexities of the available scientific data, making the formulation of clear science-based regulatory guidance more difficult. To allow for discussions of a number of key issues, the North American Branch of the International Life Sciences Institute (ILSI) convened a panel of subject matter experts for a caffeine-focused session entitled "Caffeine: Friend or Foe?," which was held during the 2015 ILSI Annual Meeting. The panelists' expertise covered topics ranging from the natural occurrence of caffeine in plants and interindividual metabolism of caffeine in humans to specific behavioral, reproductive, and cardiovascular effects related to caffeine consumption. Each presentation highlighted the potential risks, benefits, and challenges that inform whether caffeine exposure warrants concern. This paper aims to summarize the key topics discussed during the session.

Association of miR-146a rs2910164 polymorphism with cardio-cerebrovascular diseases: A systematic review and meta-analysis

The microRNA146a rs2910164 polymorphism has been associated with the development of cardio-cerebrovascular diseases (CCDs); however, the results were inconsistent among different studies. The present report was aimed to investigate the association between rs2910164 G/C polymorphism and the risk of CCDs. Based on the data extracted from 12 eligible studies with a total of 5433 CCD cases and 6278 controls, we performed a meta-analysis to assess the diseases risk of rs2910164 G/C polymorphism under allelic contrast (C vs. G), homozygote comparisons (CC vs. GG), heterozygote comparisons (GC vs. GG), dominant model (CC+GC vs. GG) and recessive models (CC vs. GC+GG) in fixed or random effects models. We also conducted pathway enrichment analyses using the putative and validated miR-146a interacting targets to explore the functional impacts of rs2910164. The current meta-analysis results showed that rs2910164 CC genotype has a decreased risk with overall cardiovascular diseases and the specific coronary artery disease. Stratified analysis based on ethnicity showed that the CC genotype has a decreased risk with CCDs in Chinese population, but has an increased risk with CCDs in Korean and Indian populations. The results from pathway enrichment analysis also revealed the association of rs2910164 G allele with heart function and disease related pathways. Our findings suggested that miR-146a CC genotype might be a protective factor for cardiovascular diseases in Chinese population, but a risk factor in Korean and Indian populations.