Subscribe: American Journal of Epidemiology - Advance Access
http://aje.oxfordjournals.org/rss/ahead.xml
Added By: Feedage Forager Feedage Grade A rated
Language: English
Tags:
abuse  birth  buprenorphine  data  days –  drug abuse  drug  missing data  missing  pregnancy  risk  studies  time     
Rate this Feed
Rate this feedRate this feedRate this feedRate this feedRate this feed
Rate this feed 1 starRate this feed 2 starRate this feed 3 starRate this feed 4 starRate this feed 5 star

Comments (0)

Feed Details and Statistics Feed Statistics
Preview: American Journal of Epidemiology - Advance Access

American Journal of Epidemiology Advance Access





Published: Mon, 20 Nov 2017 00:00:00 GMT

Last Build Date: Mon, 20 Nov 2017 07:47:20 GMT

 



Principled Approaches to Missing Data in Epidemiologic Studies

2017-11-20

Abstract
Principled methods to appropriately analyze missing data have long existed; however, broad implementation of these methods remains challenging. In this and companion papers, we discuss issues of missing data in the epidemiologic literature. We provide details regarding missing data mechanisms and nomenclature and motivate principled analyses through a detailed comparison of multiple imputation and inverse probability weighting. We do so in the setting of a masked data-analytic challenge with missing data induced by known mechanisms to data from the Collaborative Perinatal Project, a multisite US study conducted from 1959 to 1974. We illustrate the deleterious effects of missing data with naïve methods and show how principled methods can sometimes mitigate such effects. For example when data were missing at random, naïve methods showed a spurious protective effect of smoking on spontaneous abortion, odds ratio (OR) of 0.43 (95% confidence interval, CI: 0.19, 0.93) while implementing principled methods multiple imputation (OR = 1.30, CI: 0.95, 1.77) or augmented inverse probability weighting (OR = 1.40, CI: 1.00, 1.97) provided estimates closer to the “true” full data effect (OR = 1.31, CI: 1.05, 1.64). We call for greater acknowledgement of and attention to missing data and for the broad use of principled missing data methods in epidemiologic research.



Multiple Imputation for Incomplete Data in Epidemiologic Studies

2017-11-20

Abstract
Epidemiologic studies are frequently susceptible to missing information. Omitting observations with missing variables remains a common strategy in epidemiologic studies, yet this simple approach can often severely bias parameter estimates of interest if the values are not missing completely at random. Even when missingness is completely random, complete case analysis can reduce efficiency of estimated parameters, because large amounts of available data are simply tossed out with the incomplete observations. Alternative methods to mitigate missing information, such as multiple imputation, are becoming an increasing popular strategy to retain all available information, reduce potential bias, and improve efficiency in parameter estimation. In this paper, we describe the theoretical underpinnings of multiple imputation, and we illustrate application of this method in a collaborative challenge to assess the performance of various techniques to dealing with missing data. We detail the steps necessary to perform multiple imputation on a subset of the Collaborative Perinatal Project, where the goal is to estimate the odds of spontaneous abortion associated with smoking during pregnancy.



Inverse Probability Weighted Estimation for Monotone and Nonmonotone Missing Data

2017-11-20

Abstract
Missing data is of common occurrence in epidemiologic research. In this paper, three data sets with induced missing values from the Collaborative Perinatal Project, a multisite United States study conducted from 1959 to 1974, are provided as examples of prototypical epidemiologic studies with missing data. Our goal is to estimate the effect of maternal smoking behavior on spontaneous abortion while adjusting for numerous confounders. At the same time, we do not necessarily wish to evaluate the joint distribution among potentially unobserved covariates, which is seldom the subject of substantive scientific interest. The inverse probability weighting approach preserves the semiparametric structure of the underlying model of substantive interest, and clearly separates the model of substantive interest from the model used to account for the missing data. However, inverse probability weighting often will not result in valid inference if the missing data pattern is nonmonotone, even if the data are missing at random. We describe a recently proposed approach to model nonmonotone missing data mechanisms under missing at random for use to construct the weights in inverse probability weighted complete-case estimation, and we illustrate the approach in the three data sets described in the companion manuscript (Am J Epidemiol. 2017;000(0):000-000) of this issue of the journal.



Associations Between Divorce and Onset of Drug Abuse in a Swedish National Sample

2017-11-16

Abstract
Rates of drug abuse are higher among divorced than married individuals, but it is not clear whether divorce itself is a risk factor for drug abuse or if the observed association is confounded by other factors. We examined the association between divorce and onset of drug abuse in a population-based Swedish cohort born 1965–1975 (n = 651,092) using Cox proportional hazard methods, with marital status as a time-varying covariate. Potential confounders (e.g., demographics, adolescence deviance, family history of drug abuse) were included as covariates. Parallel analyses were conducted for widowhood and drug abuse onset. In adjusted models, divorce was associated with a substantial increase in risk of drug abuse onset in both sexes (hazard ratios >5). Co-relative analyses were consistent with a partially causal role of divorce on drug abuse onset. Widowhood also increased risk of drug abuse onset, though to a lesser extent. Divorce is a potent risk factor for onset of drug abuse, even after adjusting for deviant behavior in adolescence and family history of drug abuse. The somewhat less pronounced association with widowhood, particularly among men, suggests that the magnitude of association between divorce and drug abuse may not be generalizable to the end of a relationship.



Early Pregnancy Perfluoroalkyl Substance Plasma Concentrations and Birth Outcomes in Project Viva: Confounded by Pregnancy Hemodynamics?

2017-11-16

Abstract
Associations of prenatal exposure to perfluoroalkyl substances (PFASs), ubiquitous chemicals used in stain and water resistant products, with adverse birth outcomes may be confounded by pregnancy hemodynamics. We measured plasma concentrations of four PFASs in early pregnancy (median, 9 weeks) among 1,645 women in Project Viva, a Boston-area cohort recruited 1999–2002. We fit multivariable models to estimate PFAS associations with birth weight-for-gestational age z score and gestation length adjusting for sociodemographic confounders and two hemodynamic markers: 1) plasma albumin, a measure of plasma volume expansion, and 2) plasma creatinine, used to estimate glomerular filtration rate. Perfluorooctane sulfonate (PFOS) and perfluorononanoate (PFNA) were weakly inversely associated with birth weight-for-gestational age z scores (adjusted β = −0.04 (95% confidence interval (CI): −0.08, 0.1) and −0.06 (95% CI: −0.11, −0.01) per interquartile increase, respectively). PFOS and PFNA were also associated with higher odds of preterm birth (e.g., highest vs. lowest PFOS quartile adjusted odds ratio = 2.4 (95% CI: 1.3, 4.4)). Adjusting for markers of pregnancy hemodynamics (glomerular filtration rate and plasma albumin), to the extent that they accurately reflect underlying pregnancy physiology, did not materially impact associations. These results suggest that pregnancy hemodynamics may not confound associations with birth outcomes when PFASs are measured early in pregnancy.



Exposure Biomarkers Indicate More than Just Exposure

2017-11-16

Abstract
Biomarkers of environmental exposures have notable strengths in integrating across diverse sources and routes of exposure and providing a marker reflecting biological dose. However, the physiologic determinants of biomarker toxicokinetics and measured levels may also affect or be affected by disease determinants and thus introduce confounding. Sagiv et al. (1) sought empirical evidence on the role of renal clearance in biasing the association between perfluoroalkyl compounds measured in plasma during pregnancy and infant birth weight and found little empirical support for such bias. The risk of such bias is greater when the exposure and health outcome are assessed closely in time, when physiologic differences are large relative to variability in environmental levels, and when the physiologic determinant has diverse functions and implications. While empirical examination has value, the potential bias is difficult to measure and control when the underlying associations among exposure biomarker, health outcome, and physiologic determinant are weak.



Neonatal Outcomes in a Medicaid Population With Opioid Dependence

2017-11-16

Abstract
Confounding may account for the apparent improved infant outcomes after prenatal exposure to buprenorphine versus methadone. We used Massachusetts Medicaid Analytic eXtract (MAX) data to identify a cohort of opioid dependent mother-infant pairs (2006–2011) supplemented with confounder data from an external Boston cohort (2015–2016). Associations between prenatal buprenorphine versus methadone exposure and infant outcomes in the MAX Cohort were adjusted for measured MAX confounders, and unmeasured confounders with bias analysis using External Cohort data. 477 women in MAX were treated with methadone and 543 with buprenorphine. More buprenorphine users were white and used psychotropic medications. Adjusting for MAX confounders, risk ratios in buprenorphine versus methadone exposed infants were: preterm birth (0.45, 95% CI: 0.34, 0.61) and low birth weight for gestational age (0.75, 95% CI: 0.51, 1.11). The mean difference in infant hospitalization was −7.35 days (95% CI: −9.16, −5.55). After further adjustment with bias analysis estimates were: preterm birth (0.53, 95% CI: 0.39, 0.71), low birth weight for gestational age (1.14, 95% CI: 0.77, 1.69), and hospitalization (−3.66 days, 95% CI: −5.46, −1.87). External confounder data can be used to adjust for unmeasured confounding in studies of prenatal outcomes in women on opioid agonist therapy based on administrative databases.



A novel strategy to address unmeasured confounding when comparing opioid agonist therapies in pregnancy

2017-11-16

Abstract
Opioid addiction in pregnancy is a growing concern that has recently received a lot of attention. When comparing recommended opioid agonist therapies, many currently published studies guiding practice may be impacted by unmeasured confounding by indication. Populations of women who receive methadone are generally different from those treated with buprenorphine. Women treated with methadone frequently have more severe and uncontrolled addiction compared with buprenorphine treated patients; however, these factors are typically unmeasured or unavailable in large observational data sets. Consequently, findings of superior perinatal outcomes with buprenorphine may in truth be a result of an overall healthier profile of women taking this medication. In this issue of the American Journal of Epidemiology, Brogly et al. describes an approach utilizing detailed data from an External Cohort (n = 113) to account for confounding by indication in a larger Medicaid population (n = 1,020) to more accurately compare opioid agonist therapies in pregnancy. Authors found that the decreased risk of preterm birth and infant length of hospitalization associated with buprenorphine compared with methadone were attenuated after accounting for the additional confounding. These authors should be commended for providing a novel method to address this bias in future studies.



Long Term Risk of Cardiovascular Death With Use of Clarithromycin and Roxithromycin – a Nationwide Cohort Study

2017-11-16

Abstract
Recent studies have raised concern that macrolide antibiotics may be associated with a long-term increased risk of cardiovascular death. This study examines the one-year risk associated with treatment with clarithromycin (n = 751,543) and roxithromycin (n = 698,899) compared with penicillin V (n = 2,721,538), matched 1:4 on propensity score, in a nationwide registry-based cohort study in Danish outpatients, 1997–2011. Among clarithromycin courses, the rate ratio (RR) of cardiovascular death was 1.24; 95% CI: 0.96, 1.59). Among roxithromycin courses, the RR was 0.99 (0.86, 1.16). In analyses by time after treatment start, the RR associated with clarithromycin was 1.66 (0.98, 2.79) during days 0–7; this was attenuated in later time periods, RR days 8–89: 1.30 (0.88, 1.94) and RR days 90–364: 0.96 (0.63, 1.47). For roxithromycin, the RRs were 0.88 (0.59, 1.32) during days 0–7, 1.17 (0.92, 1.48) during days 8–89 and 0.88 (0.70, 1.10) during days 90–364. This study found no increased risk of cardiovascular death in a general outpatient population. With clarithromycin, we observed a transient increased risk during days 0–7 after treatment start, which corresponds to the period of active treatment. This association was absent in later time periods, which is consistent with no long-term toxicity resulting in cardiovascular death.



Accounting for Time-varying Confounding in the Relation between Obesity and Coronary Heart Disease: Analysis with G-estimation, the Atherosclerosis Risk in Communities (ARIC) study

2017-11-16

Abstract
In longitudinal studies, standard analysis may yield biased estimates of exposure effect in the presence of time-varying confounders that are also intermediate variables. We aimed to quantify the relationship between obesity and coronary heart disease (CHD) by appropriately adjusting for time-varying confounders. This study performed on a subset of the Atherosclerosis Risk in Communities Study (1987-2010). General obesity defined as body mass index ≥30 kg/m2, abdominal obesity (AOB) defined as waist circumference ≥102 cm in men and ≥88 cm in women, and waist to hip ratio categorized at ≥0.9 in men and ≥0.85 in women. The effect of obesity on CHD was estimated by G-estimation and compared with accelerated failure time models using three specifications. The first model adjusted for baseline covariates excluding metabolic mediators of obesity showed increased risk of CHD for all measures of obesity. Further adjustment for metabolic mediators in the second model and time-varying variables in the third model showed negligible hazard ratios. The hazard ratios estimated by G-estimation were 1.15 (95%CI: 0.83-1.47) for General obesity, 1.65 (95%CI: 1.35-1.92) and 1.38 (95%CI: 1.13-1.99) for AOB based on waist circumference and waist to hip ratio, respectively, suggesting that AOB increased the risk of CHD. The G-estimated hazard ratio of both measures was further from the null than standard models.