Gary G. Koch

The measurement of observer agreement for categorical data.

This paper presents a general statistical methodology for the analysis of multivariate categorical data arising from observer reliability studies. The procedure essentially involves the construction of functions of the observed proportions which are directed at the extent to which the observers agree among themselves and the construction of test statistics for hypotheses involving these functions. Tests for interobserver bias are presented in terms of first-order marginal homogeneity and measures of interobserver agreement are developed as generalized kappa-type statistics. These procedures are illustrated with a clinical diagnosis example from the epidemiological literature.

U.S. householder survey of functional gastrointestinal disorders. Prevalence, sociodemography, and health impact

Our objective was to obtain national data of the estimated prevalence, sociodemographic relationships, and health impact of persons with functional gastrointestinal disorders. We surveyed a stratified probability random sample of U.S householders selected from a data base of a national market firm (National Family Opinion, Inc.). Questions were asked about bowel symptoms, sociodemographic associations, work absenteeism, and physician visits. The sampling frame was constructed to be demographically similar to the U.S. householder population based on geographic region, age of householder, population density, household income and household size. Of 8250 mailings, 5430 were returned suitable for analysis (66% response). The survey assessed the prevalence of 20 functional gastrointestinal syndromes based on fulfillment of multinational diagnostic (Rome) criteria. Additional variables studied included: demographic status, work absenteeism, health care use, employment status, family income, geographic area of residence, population density, and number of persons in household. For this sample, 69% reported having at least one of 20 functional gastrointestinal syndromes in the previous three months. The symptoms were attributed to four major anatomic regions: esophageal (42%), gastroduodenal (26%), bowel (44%), and anorectal (26%), with considerable overlap. Females reported greater frequencies of globus, functional dysphagia, irritable bowel syndrome, functional constipation, functional abdominal pain, functional biliary pain and dyschezia; males reported greater frequencies of aerophagia and functional bloating. Symptom reporting, except for incontinence, declines with age, and low income is associated with greater symptom reporting. The rate of work/school absenteeism and physician visits is increased for those having a functional gastrointestinal disorder. Furthermore, the greatest rates are associated with those having gross fecal incontinence and certain more painful functional gastrointestinal disorders such as chronic abdominal pain, biliary pain, functional dyspepsia and IBS. Preliminary information on the prevalence, socio-demographic features and health impact is provided for persons who fulfill diagnostic criteria for functional gastrointestinal disorders.

Analysis of categorical data by linear models.

Assume there are ni., i = 1, 2, *--, s, samples from s multinomial distributions each having r categories of response. Then define any u functions of the unknown true cell probabilities {7rij: i = 1, 2, * , s; j = 1, 2, * , r, where E jrij l 1 } that have derivatives up to the second order with respect to 7rij, and for which the matrix of first derivatives is of rank u. A general noniterative procedure is described for fitting these functions to a linear model, for testing the goodness-of-fit of the model, and for testing hypotheses about the parameters in the linear model. The special cases of linear functions and logarithmic functions of the 7rin are developed in detail, and some examples of how the general approach can be used to analyze various types of categorical data are presented.

Echocardiographic predictors of adverse outcomes in primary pulmonary hypertension

OBJECTIVES The aim of this study was to evaluate the relationships between echocardiographic findings and clinical outcomes in patients with severe primary pulmonary hypertension (PPH). BACKGROUND Primary pulmonary hypertension is associated with abnormalities of right heart structure and function that contribute to the poor prognosis of the disease. Echocardiographic abnormalities associated with PPH have been described, but the prognostic significance of these findings remains poorly characterized. METHODS Echocardiographic studies, invasive hemodynamic measurements and 6-min walk tests were performed and outcomes prospectively followed in 81 patients with severe PPH. Subjects were participants in a 12-week randomized trial examining the effects of prostacyclin plus conventional therapy compared with conventional therapy alone. RESULTS During the mean follow-up period of 36.9 +/- 15.4 months, 20 patients died and 21 patients underwent transplantation. Pericardial effusion (p = 0.003) and indexed right atrial area (p = 0.005) were predictors of mortality. Pericardial effusion (p = 0.017), indexed right atrial area (p = 0.012) and the degree of septal shift in diastole (p = 0.004) were predictors of a composite end point of death or transplantation. In multivariable analyses incorporating clinical, hemodynamic and echocardiographic variables, pericardial effusion and an enlarged right atrium remained predictors of adverse outcomes. Six-minute walk results, mixed venous oxygen saturation and initial treatment randomization were also independently associated with a poor prognosis. CONCLUSIONS Pericardial effusion, right atrial enlargement and septal displacement are echocardiographic abnormalities that reflect the severity of right heart failure and predict adverse outcomes in patients with severe PPH. These characteristics may help identify patients appropriate for more intensive medical therapy or earlier transplantation.

Cognitive-behavioral therapy versus education and desipramine versus placebo for moderate to severe functional bowel disorders

BACKGROUND & AIMS Studies of antidepressants and psychological treatments in functional bowel disorders (FBD) are methodologically limited. The aim of this study was to assess the clinical efficacy and safety of cognitive-behavioral therapy (CBT) against education (EDU) and desipramine (DES) against placebo (PLA) in female patients with moderate to severe FBD (irritable bowel syndrome, functional abdominal pain, painful constipation, and unspecified FBD). We also evaluated the amenability of clinically meaningful subgroups to these treatments. METHODS This randomized, comparator-controlled, multicenter trial enrolled 431 adults from the University of North Carolina and the University of Toronto with moderate to severe symptoms of FBD. Participants received psychological (CBT vs. EDU) or antidepressant (DES vs. PLA) treatment for 12 weeks. Clinical, physiologic, and psychosocial assessments were performed before and at the end of treatment. RESULTS The intention-to-treat analysis showed CBT as significantly more effective than EDU (P = 0.0001; responder rate, 70% CBT vs. 37% EDU; number needed to treat [NNT ], 3.1). DES did not show significant benefit over PLA in the intention-to-treat analysis (P = 0.16; responder rate, 60% DES vs. 47% PLA; NNT, 8.1) but did show a statistically significant benefit in the per-protocol analysis (P = 0.01; responder rate, 73% DES vs. 49% PLA; NNT, 5.2), especially when participants with nondetectable blood levels of DES were excluded (P = 0.002). Improvement was best gauged by satisfaction with treatment. Subgroup analyses showed that DES was beneficial over PLA for moderate more than severe symptoms, abuse history, no depression, and diarrhea-predominant symptoms; CBT was beneficial over EDU for all subgroups except for depression. CONCLUSIONS For female patients with moderate to severe FBD, CBT is effective and DES may be effective when taken adequately. Certain clinical subgroups are more or less amenable to these treatments.

Lifetime risk of symptomatic knee osteoarthritis

OBJECTIVE To estimate the lifetime risk of symptomatic knee osteoarthritis (OA), overall and stratified by sex, race, education, history of knee injury, and body mass index (BMI). METHODS The lifetime risk of symptomatic OA in at least 1 knee was estimated from logistic regression models with generalized estimating equations among 3,068 participants of the Johnston County Osteoarthritis Project, a longitudinal study of black and white women and men age >or=45 years living in rural North Carolina. Radiographic, sociodemographic, and symptomatic knee data measured at baseline (1990-1997) and first followup (1999-2003) were analyzed. RESULTS The lifetime risk of symptomatic knee OA was 44.7% (95% confidence interval [95% CI] 40.0-49.3%). Cohort members with history of a knee injury had a lifetime risk of 56.8% (95% CI 48.4-65.2%). Lifetime risk rose with increasing BMI, with a risk of 2 in 3 among those who were obese. CONCLUSION Nearly half of the adults in Johnston County will develop symptomatic knee OA by age 85 years, with lifetime risk highest among obese persons. These current high risks in Johnston County may suggest similar risks in the general US population, especially given the increase in 2 major risk factors for knee OA, aging, and obesity. This underscores the immediate need for greater use of clinical and public health interventions, especially those that address weight loss and self-management, to reduce the impact of having knee OA.

An evaluation of Safe Dates, an adolescent dating violence prevention program.

OBJECTIVES This study assessed the effects of the Safe Dates program on the primary and secondary prevention of adolescent dating violence. METHODS Fourteen schools were randomly allocated to treatment conditions. Eighty percent (n=1886) of the eighth and ninth graders in a rural county completed baseline questionnaires, and 1700 (90%) completed follow-up questionnaires. RESULTS Treatment and control groups were comparable at baseline. In the full sample at follow-up, less psychological abuse, sexual violence, and violence perpetrated against the current dating partner were reported in treatment than in control schools. In a subsample of adolescents reporting no dating violence at baseline (a primary prevention subsample), there was less initiation of psychological abuse in treatment than in control schools. In a subsample of adolescents reporting dating violence at baseline (a secondary prevention subsample), there was less psychological abuse and sexual violence perpetration reported at follow-up in treatment than in control schools. Most program effects were explained by changes in dating violence norms, gender stereotyping, and awareness of services. CONCLUSIONS The Safe Dates program shows promise for preventing dating violence among adolescents.

Average Partial Association in Three-way Contingency Tables: a Review and Discussion of Alternative Tests

Summary This paper reviews and discusses alternative methods for assessing average partial association in three-way contingency tables. Primary attention is directed at the relationship between two of the variables, while controlling for the effects of a set of covariables. One approach is a class of multivariate extensions of the Cochran-Mantel-Haenszel test to sets of (s x r) tables. These statistics are based on expected values and covariances from the multiple hypergeometric distribution for each table. As such they make no underlying assumption regarding second-order interaction, although the absence of such interaction is incorporated within the hypothesis being tested. Alternatively, a log-linear model can be used to investigate average partial association conditional on the assumption of no second-order interaction. If the fit of such a model is supported by the data, then likelihood ratio methods or functional asymptotic regression methods (FARM) can be used to test the significance of correspondingly appropriate parameters. These procedures are all illustrated with a data set relating atomic bomb radiation to the incidence of leukemia adjusted for age at exposure.

A one-way components of variance model for categorical data

A components of variance model for categorical data from unbalanced designs which is directly analogous to a one-way random effects ANO VA modelfor quantitative data is proposed. The variance components provide separate reliability measures for each of the response categories and disagreement measures between pairs of response categories in terms of (within subject) intraclass and interclass correlation coefficients. The estimation procedures involve usual MA NOVA calculations which can be expressed as compounded functions of the multinomial observations. Thus, the variances of these estimates can be obtainedfrom linearized Taylor series results. These procedures are illustrated with data from a psychiatric diagnosis study.

Comparison of phytotherapy (Permixon®) with finasteride in the treatment of benign prostate hyperplasia: A randomized international study of 1,098 patients

Controversy regarding the relative efficacy of treatments for the relief of the symptoms of benign prostatic hyperplasia (BPH).