Pairwise Odds Ratios Research Articles

Geographic clustering of underage drinking and the influence of community characteristics

The aim of this paper was to examine the extent to which underage drinking clusters geographically in a sample of communities, and to investigate the manner in which community-level contexts are related to this process. We used data from a randomized community trial of underage drinking to provide the first quantitative estimates of the magnitude of the geographic clustering of underage drinking based upon pairwise odds ratios (PWORs). The Enforcing Underage Drinking Laws Randomized Community Trial provided data from repeated cross-sectional samples of youth aged 14–20 from 68 communities surveyed in 2004, 2006, and 2007 ( n = 18,730). Past 30-day drinking, binge drinking, getting drunk, experiencing non-violent consequences as a result of drinking and making a purchase attempt all significantly clustered within-communities with PWORs ranging from 1.05 to 1.21. After adjustment for individual-level characteristics, results remained relatively unchanged. However, there was evidence that the magnitude of the clustering varied as a function of neighborhood disadvantage, neighborhood disorder, and family structure. Clustering of drunkenness and experiencing non-violent consequences as a result of drinking was greatest in the least economically disadvantaged and least disordered communities with the greatest percentage of married-couple families. The clustering of making a purchase attempt, however, was greatest in more disordered communities, specifically the largest communities with the highest degree of residential mobility and housing density. These findings that clustering of underage drinking behaviors varies by community context has the potential for identifying the types of communities to target for underage drinking behavior-specific preventive interventions.

A robust method for finely stratified familial studies with proband-based sampling.

This paper presents a robust method to conduct inference in finely stratified familial studies under proband-based sampling. We assume that the interest is in both the marginal effects of subject-specific covariates on a binary response and the familial aggregation of the response, as quantified by intrafamilial pairwise odds ratios. We adopt an estimating function for proband-based family studies originally developed by Zhao and others (1998) in the context of an unstratified design and treat the stratification effects as fixed nuisance parameters. Our method requires modeling only the first 2 joint moments of the observations and reduces by 2 orders of magnitude the bias induced by fitting the stratum-specific nuisance parameters. An analytical standard error estimator for the proposed estimator is also provided. The proposed approach is applied to a matched case-control familial study of sleep apnea. A simulation study confirms the usefulness of the approach.

Estimating co-occurring behavioral trajectories within a neighborhood context: a case study of multivariate transition models for clustered data.

Comorbidity is well-documented in psychiatric and risk behavior epidemiology. The authors present a novel application of clustered multivariate transition models to study comorbidity within a clustered context. The authors used data from the Project on Human Development in Chicago Neighborhoods (1995-2002) to assess trajectories in substance use, problems with police, and antisocial behavior among 1,517 participants in 80 neighborhoods followed from ages 12-15 years through ages 18-21 years. The authors used pairwise odds ratios to quantify behavior comorbidity at the individual and neighborhood levels. Risk behaviors co-occurred within individuals at specific points in time: antisocial behavior and substance use were 3.37 times more likely to co-occur within an individual at wave 1, as compared with the co-occurrence of any 2 behaviors from different individuals, while substance use and police problems were 2.94 times more likely to co-occur than substance use and antisocial behavior at wave 2. The authors also evaluated sequential comorbidity. Antisocial behavior was sequentially comorbid with substance use and police problems: 31% of youths who had reported antisocial behavior at baseline reported police problems or drug use at wave 2. These models can prove instrumental in answering the persistent questions about possible sequential relations among problem behaviors.

Therapeutic traditions, patient socioeconomic characteristics and physicians’ early new drug prescribing–a multilevel analysis of rosuvastatin prescription in south Sweden

To investigate the role that both patient and outpatient factors related to health care practice (HCP) play in physicians' early adoption of rosuvastatin. Generalized estimation equations (GEEs) and alternating logistic regression (ALR) with pair-wise odds ratios (PWORs) were used to measure similarities in rosuvastatin prescription within HCPs for all individuals with statin prescriptions in Skåne region, Sweden. After 12 months, 53% of the HCPs had adopted the new statin. Rosuvastatin prescriptions co-occurred within certain HCPs 3.56 times more often than one would have expected based on a random distribution. Private HCPs had four times higher probability of prescribing rosuvastatin than public HCPs. Contextual characteristics of the HCP seem to be relevant for understanding physicians' motivation to adopt rosuvastatin. Moreover, our study reveals inequity in health care as the socioeconomic status of the patients appears to influence the prescribing behavior of the physicians irrespective of medical reasons.

Improving Efficiency of Inferences in Randomized Clinical Trials Using Auxiliary Covariates

The primary goal of a randomized clinical trial is to make comparisons among two or more treatments. For example, in a two-arm trial with continuous response, the focus may be on the difference in treatment means; with more than two treatments, the comparison may be based on pairwise differences. With binary outcomes, pairwise odds ratios or log odds ratios may be used. In general, comparisons may be based on meaningful parameters in a relevant statistical model. Standard analyses for estimation and testing in this context typically are based on the data collected on response and treatment assignment only. In many trials, auxiliary baseline covariate information may also be available, and it is of interest to exploit these data to improve the efficiency of inferences. Taking a semiparametric theory perspective, we propose a broadly applicable approach to adjustment for auxiliary covariates to achieve more efficient estimators and tests for treatment parameters in the analysis of randomized clinical trials. Simulations and applications demonstrate the performance of the methods.

Geographical clustering of eating disordered behaviors in U.S. high school students

The aim of this study was to determine if eating disorder behaviors geographically clustered among U.S high school students. Our sample consisted of 15,349 high school students who responded to the 1999 Youth Risk Behavior Study (YRBS). Weight control and eating disordered behaviors under investigation included dieting, exercising, fasting, using diet pills, and purging to control weight in the last 30 days. We calculated pairwise odds ratios (PWORs) to determine the degree of within-county clustering. Among all participants, adjusted analyses revealed that having any weight control or eating disorder symptom, severe restricting, dieting, exercising, and diet pill use each showed significant clustering (p < .05). Purging did not significantly cluster by county. The magnitude of clustering was stronger for female students than male students. The significant clustering of weight control and eating disorder behavior in U.S. high school students confirms evidence of a social contagion effect of eating disorders.

Modeling association in longitudinal binary outcomes: A brief review

In order to better understand aging, longitudinal studies are run in which participants are evaluated repeatedly and selected end-points (e.g., score on a cognitive screen, falls, occurrence/reoccurrence of a condition) are examined. The objective of the present paper is primarily to describe the methods available that take into account correlation between binary outcomes, and in particular to model the association of binary outcomes after controlling for covariates by using an implementation of generalized estimating equations (GEE) called ‘alternating logistic regression’ (ALR). In GEE, association within longitudinal outcomes is accounted for but not estimated. Alternating logistic regression, however, basically enables simultaneous estimation of pair-wise odds ratios of outcomes within a cluster, while accounting for the dependence of the outcome on covariates. A sub-sample (n = 2458) from a community-based sample of Duke Established Populations for Epidemiologic Studies of the Elderly is used. In the example used here, logistic regression using GEE and ALR is used to model binary outcomes at three time points (baseline, three and six years later) and to control for covariates in a representative community-based sample 65 years of age and older (n = 2458). The outcomes indicate any problem versus no problem on a five-item activities of daily living (ADL) scale in a community sample. The ALR model, however, provides insight into decline in ADL from baseline to each of the time-points whereas GEE does not. In both controlled and uncontrolled analyses, decline in ADL over three and six-year intervals (baseline to three years later, baseline to six years and three years post-baseline to six years post-baseline) is significant.

Composite Conditional Likelihood for Sparse Clustered Data

SummarySparse clustered data arise in finely stratified genetic and epidemiologic studies and pose at least two challenges to inference. First, it is difficult to model and interpret the full joint probability of dependent discrete data, which limits the utility of full likelihood methods. Second, standard methods for clustered data, such as pairwise likelihood and the generalized estimating function approach, are unsuitable when the data are sparse owing to the presence of many nuisance parameters. We present a composite conditional likelihood for use with sparse clustered data that provides valid inferences about covariate effects on both the marginal response probabilities and the intracluster pairwise association. Our primary focus is on sparse clustered binary data, in which case the method proposed utilizes doubly discordant quadruplets drawn from each stratum to conduct inference about the intracluster pairwise odds ratios.

Sibling Aggregation for Psychopathology in Offspring of Opiate Addicts: Effects of Parental Comorbidity

Applied a new analytic approach within the high-risk paradigm, the analysis of sibling aggregation, to identify homogeneous subtypes of familial risk for psychopathology and addiction. All sibling pairs participating in a study of offspring of opiate-addicted parents were identified and their aggregation for psychiatric disorders was determined using pairwise odds ratios, an analytic technique used in genetic epidemiology. Sibling aggregation was most notable for depressive and anxiety disorders but only in the presence of comorbid depressive disorders in the parents. Parental comorbid alcoholism did not impact sibling aggregation. We emphasize methodological implications of this approach for addressing issues of phenotypic and etiologic heterogeneity in the study of developmental risk for substance abuse.

Clusters of marijuana use in the United States.

In this study, the authors test for and estimate the clustering of marijuana use within United States neighborhoods, making use of data from annual nationally representative household sample surveys conducted during the period 1990-1995. A recently developed statistical method, alternating logistic regression, was used to quantify the clustering of marijuana users in neighborhoods. The resulting estimates of pairwise odds ratios ranged from 1.3 (95% confidence interval 1.22-1.42) for the lifetime history of marijuana use to 2.0 (95% confidence interval 1.6-2.6) for recent sharing of marijuana from one person to another. Exploratory analysis showed a slight decrease of clustering effects after adjustment for individual-level covariates: age, sex, race, education, annual family income, and history of tobacco use. Nevertheless, the main factors that account for clustering remain to be determined. Alternating logistic regression provided useful estimates of marijuana use clustering and can be used to estimate clustering of the other drug-related behavior, including sharing of needle injection equipment and other human immunodeficiency virus risk behaviors. As a form of multilevel analysis, the alternating logistic regression can accommodate shared, community-level characteristics that might influence drug taking (e.g., collective efficacy), as well as individual-level covariates, such as age and sex.

Modeling Multivariate Discrete Failure Time Data

A bivariate discrete survival distribution that allows flexible modeling of the marginal distributions and yields a constant odds ratio at any grid point is proposed. The distribution can be extended to a multivariate distribution and is readily generalized to accommodate covariates in the marginal distributions and pairwise odds ratios. In addition, a pseudo-likelihood estimation procedure for estimating the regression coefficients in the marginal models and the association parameters in the pairwise odds ratios is presented. We evaluate the performance of the proposed estimation procedure through simulations. For bivariate data, pseudo-likelihood estimation of the association parameter has high efficiency. Loss of efficiency in the marginal regression coefficient estimates is small when the association is not strong. For both the marginal regression coefficients and the association parameter, coverage probabilities are close to the 95% nominal level. For multivariate data, the simulation results show that the parameter estimates are consistent. Coverage probability for the regression coefficient in the marginal model is close to the 95% nominal level but is slightly less than the nominal level for the association parameter. We illustrate the proposed methods using a subset of the Framingham Heart Study data where a significant positive association was found between the failure times of siblings.

Familial aggregation of oesophageal cancer in a high incidence area in China.

The high incidence of oesophageal cancer in northern China is attributed predominantly to environmental factors. The role of genetic factors has not been extensively studied. Our aim was to study familial aggregation of oesophageal cancer in pedigrees from a defined population base in a high incidence area in China and to quantify the risk associated with different first degree relatives using different analytical approaches. Detailed data on family members of three successive generations and the occurrence of oesophageal and other cancers in family members were collected from a population-based series of 244 oesophageal cancer cases which occurred between 1987 and mid-1992 in Huixian County, Henan. Compared to expected rates, the standardized mortality ratio (SMR) of oesophageal cancer among first degree relatives of oesophageal cancer patients was 2.4 (2.2 in male and 2.7 in female relatives). The corresponding SMR for first and second degree relatives were 1.6 and 2.2. The null hypothesis of 'no familial aggregation' was rejected using Tarone's one-sided score test for binomial distributions indicating some evidence for clustering within families. To account for variance due to between-pairs correlation and family and/or individual specific variables, we fitted a series of regression models using a Generalized Estimation Equations (GEE) approach. The pairwise odds ratios were 2.3 for parent-parent, 1.9 for sib-parent and 1.1 for sib-sib, adjusted for sex, age and sex of index case. The existence of familial aggregation of oesophageal cancer in the study population was confirmed using different analyses and a two- to threefold increased risk was found for first degree relatives. The clear association of disease between parent and sib provides some indication of a genetic component. The pairwise association between parents but not between sibs suggests that environmental factors have a stronger action after childhood.

Effects of randomizing second eyes in a trial to evaluate preoperative medical testing for cataract surgery

The statistical and practical implications of including second eye surgeries were examined in a clinical trial designed to evaluate the impact of routine preoperative testing prior to cataract surgery on major medical events occurring within seven days following surgery. In order to detect a 0.8% difference in the rates of rare major medical events between the tested and untested groups, 20,000 surgeries must be randomized. About 30% of cataract operations were estimated to be done on second eyes of patients already included in the cohort. Different options for dealing with second eye surgeries were: (1) exclusion of all second eye surgeries, (2) inclusion of second eye surgeries only if the first eye is not enrolled, (3) inclusion of first and second eyes but randomization of patients rather than eyes, and (4) inclusion of first and second eyes but randomization of surgeries rather than patients. The final decision was to exclude second eye surgeries done within 28 days of first eye surgeries, but to rerandomize all other second eye surgeries. Differences in event rates between treatment groups can be estimated using Generalized Estimating Equations, and the association between outcomes of first and second eye operations estimated with pairwise odds ratios. An anticipated small positive correlation is likely to have minimal impact on statistical power and effective sample size.

A meta-analysis of the association of the deletion allele of the angiotensin-converting enzyme gene with myocardial infarction.

The ACE gene is characterized by a polymorphism based on the presence (insertion [I]) or absence (deletion [D]) within intron 16 of a 287-basepair alu repeat sequence, resulting in three genotypes (DD and II homozygotes and ID heterozygotes). In 1992, the DD genotype was reported to be associated with an increased risk of myocardial infarction (MI). Subsequent studies have produced conflicting findings. To further evaluate the association of the ACE I/D genotype with MI risk, we carried out a meta-analysis of all the published studies. In total, 15 studies containing 3394 MI cases and 5479 control subjects were analyzed. The overall distribution of genotypes in the control subjects was 22.7% II, 49.0% ID, and 28.3% DD. The mean odds ratio for MI for DD versus ID/II genotypes across all studies was 1.26 (95% CI, 1.15, 1.39; P < .0001). Pairwise odds ratios were 1.36 (95% CI, 1.19, 1.55) for DD and II, 1.24 (95% CI, 1.11, 1.38) for DD and ID, and 1.09 (95% CI, 0.96, 1.23) for ID and II. The relative risk appeared to be increased in Japanese populations (2.55; 95% CI, 1.75, 3.70). Within the limitations of the available data, the meta-analysis therefore supports an association of the ACE D allele with MI risk and strengthens the justification for further evaluation in appropriately powered studies.

The Familial Aggregation of Obstructive Sleep Apnea

An inherited basis for sleep-disordered breathing (SDB) has been suggested by reports of families with multiple affected members and by a previous study of the familial aggregation of symptoms of SDB. In this study, we quantify and characterize the aggregation of SDB and assess the degree to which familial similarities may be independent of obesity. This was a genetic-epidemiologic study that assessed the distribution of SDB in families identified through a proband with diagnosed sleep apnea and among families in the same community with no relative with known sleep apnea. SDB was assessed with overnight in-home monitoring of airflow, oxygen saturation, chest wall impedance, heart rate, and body movement. Standardized questionnaires were used to assess symptoms, and weight, height, and neck circumference were measured directly. Intergenerational and intragenerational correlation coefficients and pairwise odds ratios (ORs) were calculated with adjustment for proband sampling. In toto, 561 members of 91 families were studied: (1) 47 subjects with laboratory-confirmed SDB (index probands), (2) 44 community control subjects, and (3) the spouses and relatives of 1 and 2. Of all 91 families, 32 (35%) had two or more members with SDB, 30 (33%) had one affected member, and 29 had no affected members. SDB was more prevalent in the relatives of index probands (21%) than among neighborhood control subjects (12%) (p = 0.02).(ABSTRACT TRUNCATED AT 250 WORDS)

Estimation of design effects in cluster surveys.

Cluster sampling can produce estimates of disease prevalence that are more variable than those from simple random sampling. This variance inflation or "design effect" depends on the prevalence of disease, the cluster sizes, and the magnitude of disease association within clusters. Design effects from prior surveys may not be appropriate for a planned survey if these components differ. We estimated within-cluster associations using pairwise odds ratios, which are more portable than design effects because they do not depend on the cluster sizes. Within-village pairwise odds ratios and design effects were estimated for fever and cough from four studies in Africa and Asia. Odds ratio ranged from 1.04 to 1.34 and 1.03 to 1.24, respectively. Design effects ranged from 2.35 to 6.80 for fever and 1.99 to 7.39 for cough. The design effect was more affected by cluster size and odds ratio than by variation in cluster size for a given sample size.

Estimation of design effects and diarrhea clustering within households and villages.

The degree to which diarrheal disease clustered within households and within villages among preschool age children was examined using data from four population-based prevalence surveys undertaken in Malawi, Zambia, Indonesia, and Nepal over the past decade. The design effect for each cluster survey was calculated using the diarrhea prevalence, the cluster sizes, and the magnitude of diarrhea clustering within the sampling unit (villages). A recently developed statistical method, alternating logistic regression, was used to estimate disease associations within households of up to nine preschool age children residing within villages of up to 589 such children. Pairwise odds ratios estimating diarrhea clustering within villages ranged from 1.03 (95% confidence interval (CI) 1.01-1.07) in Zambia to 2.19 (95% CI 1.73-2.78) in Indonesia. The design effects ranged from 2.07 (95% CI 1.26-3.19) in Zambia to 7.93 (95% CI 5.16-11.52) in Indonesia. Design effects were strongly dependent on cluster size. The design effects for clusters of size 50 would have ranged from 1.38 to 4.73. Pairwise odds ratios for diarrhea clustering within households ranged from 1.88 (95% CI 1.61-2.19) in Nepal to 10.05 (95% CI 8.46-11.94) in Indonesia. Household odds ratios were always larger than village odds ratios. The village and household pairwise odds ratios adjusted for age, the type of latrine used by the household, and presence of a market in the village were slightly higher than the unadjusted odds ratios. Alternating logistic regression provided useful estimates of disease clustering within villages and household while allowing for covariate adjustment.

Clustering of xerophthalmia within households and villages.

The clustering of xerophthalmia within households and villages was estimated among preschool age children using data from studies conducted in Malawi, Zambia, Indonesia and Nepal over the past decade. Pairwise odds ratios (OR) were used to measure the magnitude of clustering. This OR measures the risk of xerophthalmia for a preschool child given that another randomly chosen preschool child from the same household (or same village but different household) had xerophthalmia, relative to the risk if that randomly chosen preschool child did not have xerophthalmia. Village pairwise OR ranged from 1.2 in Malawi to 2.2 in Nepal. Household pairwise OR ranged from 4.4 in Malawi to 9.7 in Indonesia, indicating that xerophthalmia clustering was much greater within households than villages. The magnitude of this clustering was as large, or larger than, infectious outcomes such as diarrhoea, fever and cough. Although xerophthalmia was associated with a weekly history of infectious morbidity, the clustering of diarrhoea, fever and cough explained very little of the xerophthalmia clustering observed in each of these studies, Hence, other household factors such as food availability and dietary practices should be examined for their role in the clustering of xerophthalmia within certain households.

Modelling multivariate binary data with alternating logistic regressions

Marginal models for multivariate binary data permit separate modelling of the relationship of the response with explanatory variables, and the association between pairs of responses. When the former is the scientific focus, a first-order generalized estimating equation method (Liang & Zeger, 1986) is easy to implement and gives efficient estimates of regression coefficients, although estimates of the association among the binary outcomes can be inefficient. When the association model is a focus, simultaneous modelling of the responses and all pairwise products (Prentice, 1988) using second-order estimating equations gives more efficient estimates of association parameters as well. However, this procedure can become computationally infeasible as the cluster size gets large. This paper proposes an alternative approach, alternating logistic regressions, for simultaneously regressing the response on explanatory variables as well as modelling the association among responses in terms of pairwise odds ratios. This algorithm iterates between a logistic regression using first-order generalized estimating equations to estimate regression coefficients and a logistic regression of each response on others from the same cluster using an appropriate offset to update the odds ratio parameters. For clusters of size n, alternating logistic regression involves evaluation and inversion of matrices of order n2 rather than n4 as required for second-order generalized estimating equations. The alternating logistic regression estimates are shown to be reasonably efficient relative to solutions of second-order equations in a few problems. The new method is illustrated with an analysis of neuropsychological tests on patients with epileptic seizures.