Peter B. Imrey

Nasopharyngeal carcinoma in Malaysian Chinese: Salted fish and other dietary exposures

We interviewed 282 histologically confirmed cases of nasopharyngeal carcinoma (NPC) in Chinese residents of Selangor and the Federal Territory, Malaysia, and an equal number of Chinese age‐, sex‐, and length‐of‐residence‐matched controls sampled from the general population. Consumption of 55 dietary items during childhood, and 5 years pre‐diagnosis of NPC, was analyzed by univariate and multivariate methods. Four salted preserved foods (fish, leafy vegetables, egg and root), fresh pork/beef organ meats and beer and liquor consumption exhibited strong positive associations, and 4 vegetable/fruit combinations strong negative associations with NPC. Factor analysis and multivariable modeling using estimated factor scores strongly supported separate effects on NPC of vegetables/fruits, salted preserved foods, pork/beef organ meats and beer/liquor consumption. Multivariable modeling associated NPC most clearly with high consumption of salted fish, salted eggs, pork/beef liver and beer and low consumption of Chinese flowering cabbage, oranges/tangerines and shrimp. A strong residual association of social class with NPC remained after adjustment for diet, which is consistent with a substantial role for non‐dietary environmental factors. Int. J. Cancer 77:228–235, 1998.© 1998 Wiley‐Liss, Inc.

Categorical Data Analysis: Some Reflections on the Log Linear Model and Logistic Regression. Part I: Historical and Methodological Overview*

Summary The literature of log linear models and logistic regression is surveyed from a contemporary point of view. A matrix formulation of the general log linear model for product-multinomial random counts is exploited to study the relationship between maximum likelihood and weighted least squares approaches to model fitting. Maximum likelihood fitted parameters and cell expectations are shown to be stationary solutions of a weighted least squares equation. Matrix expressions for asymptotic covariance matrices of efficient fitted parameters and cell counts are developed. Asymptotic covariance matrices of generalized raked contingency tables are obtained from the matrix formulation. Functional asymptotic regression methodology, an approach combining aspects of maximum likelihood and weighted least squares, is described and examined. In Part II, the several methods and relationships are illustrated by seven examples; extensions applicable to noncentrality problems and complex sample survey designs are also presented.

Infectious Disease Consultation and Microbiologic Surveillance for Intensive Care Unit Trauma Patients: A Pilot Study

Infection remains a major cause of posttrauma morbidity. We retrospectively reviewed 2 cohorts of trauma patients admitted to a regional trauma center before and after a policy change integrating prospective microbiologic surveillance and infectious disease (ID) consultation into management of trauma admissions. Primary interests were effects of this policy change on antimicrobial use and diagnostic precision (particularly differentiation of infection from colonization). Associated costs, microflora, survival, and disability were also compared. Patients were stratified for risk of infection. ID consultation was associated with a 49% increased odds that an infection diagnosis was microbiologically based (P=.006) and 57% reduction of antibiotics costs per hospitalized day (P=.0008). Costs of consultation and an 86% increase (P<10(-6)) in total cultures combined to minimally exceed that financial saving. The observed improvements in diagnostic precision and antimicrobial usage, however, suggest consideration of prospective microbiologic surveillance and multidisciplinary physician teams including ID physicians for high-risk trauma patients.

An Incomplete Contingency Table Approach to Paired-Comparison Experiments

Abstract A wide variety of paired-comparison experiments and surveys may be viewed within the framework of “incomplete” contingency table analysis. Bradley-Terry models for such data can be chosen and further examined by a noniterative logit analysis. Multivariate paired comparisons with factor structure can be analyzed through reparameterization of the underlying models for each combination of factor levels. The method is subject to sample-size restrictions but allows the experimenter to design his survey without complex symmetry conditions or independence assumptions.

Statistical Estimation Of Child Abuse Rates from Administrative Databases

Abstract Increasingly, child welfare agencies need to provide statistical summary reports on the safety of the children for whom they are responsible. Often these summaries include ordinary period rates of child abuse derived from administrative data. These indices fail to adjust for the length of exposure to risk, such as time in foster care during the year. Since duration of care often differs by living arrangement, race, gender, and other variables, the use of such period rates to measure child safety or abuse may bias comparisons over time or across groups. This bias may lead to misperceptions of trends in safety over time, and of the comparative safety of different modes of care. This article discusses fundamental issues in the extraction, from administrative data, of valid measures of child welfare outcomes targeted to specific populations. In addition, it provides an introduction to exposure adjustment of child welfare measures based on information that is generally readily available in administrative databases. Cohort-based incidence density rates are recommended in preference to period prevalence from cross-sectional data. Survival modeling/multiple event history analysis is described for more complex situations. The ready availability of such analytic tools suggests further directions for quantitative research in child welfare monitoring.

Minor ailments and illness behavior among physicians.

This study examined the illness behavior of a sample (n = 375) of physicians in regard to the treatment of their own minor ailments. The symptoms for each of 13 minor ailments were clearly worded to describe a common ailment which in the normal course of events should disappear if left untreated. For 3 of the ailments, however, there was a high frequency of physician-dependent action; 4 other ailments showed a moderate frequency of physician-dependent action. Assuming that physicians are role models for illness behavior, in the sense that they have the authority to define what illness “is” and that they will then do what they think “best,” this sample of physicians is implicitly demonstrating what lay persons should do. These data suggest that the normative expectation to seek professional assistance (a medical solution) is reinforced by physicians through their own behavior.

Dietary protein and chronic toxicity of 1,2‐dimethylhydrazine fed to mice

1,2-Dimethylhydrazine-HCl (DMH-2HCl) is derived from the natural toxin cycasin, and is extensively used to induce cancers in experiments with rodents. We examined the toxicity of DMH-2HCl, incorporated into purified diets varying in protein, to determine concentrations compatible with long-term survival in B6C3H1 mice. Initial studies showed single-dose oral LD50 values (95% confidence intervals) of 26 (18-32) mg DMH-2HCl/kg body weight for males, and 60 (53-65) for females. A 6-wk study was performed with diets containing 10 or 40% soybean protein with doses of 0, 11.25, 22.5, 45, 90, and 180 mg DMH-2HCl/kg diet. All mice fed the highest dose were removed from the study due to severe toxicity. Declines in food consumption and body weight occurred in both sexes, accelerated with increasing log(DMH) dose, and were substantially more severe in groups fed 10% protein. A 5-mo study was subsequently performed with male mice fed 10 or 40% protein diets containing doses of 0, 15, 30, or 45 mg DMH-2HCl/kg diet. In this longer study, dose-related declines of food intake and body weight were also more pronounced with 10% protein. Histopathologic examination of samples from 29 organs/tissues revealed hepatic changes most commonly, and these were more severe at higher DMH levels. Lesions ranged from focal centrilobular hepatocellular necrosis to severe toxic hepatitis, associated with lobular disorganization and hepatocellular hypertrophy. Frequent dose-dependent lesions were also found in kidneys, adrenals, and heart. Renal changes included focal subcapsular fibrosis with atrophy, and hyperplasia of the tubular epithelium. Adrenal cortical hypertrophy was noted at the two highest DMH doses. Focal cardiac myocytolysis was also noted at high DMH doses. Renal damage occurred only rarely in the absence of liver pathology, and adrenal hypertrophy only rarely without renal damage. Cardiac myocytolysis was found in 14% of mice without hepatic, renal, or adrenal damage, but in 62% of those with lesions in each of those organs. No evidence of gastrointestinal toxicity was observed. Hepatic, renal, and adrenal lesions were more frequent and severe in mice fed the low-protein diet. The protective effect of high protein was DMH-dose dependent. The lower doses in these studies could be used to investigate effects of diet, cocarcinogens, or chemopreventative agents on carcinogenesis resulting from chronic, low-level dietary exposure to DMH.

Poisson Regression, Logistic Regression, and Loglinear Models for Random Counts

Publisher Summary This chapter provides a unified discussion of Poisson regression, logistic regression, and loglinear modeling of contingency tables. These are three special cases of the general loglinear model, wherein expected category counts are products of effects of independent variables. This contrasts with the general linear model in which expected means of continuous measurements are sums of such effects. Poisson and logistic regression each provide regression, analysis of variance (ANOVA), and analysis of covariance (ANCOVA)-like analyses for response counts with, respectively, one and two levels. Poisson regression is most commonly used to analyze rates, whereas logistic regression is used to analyze proportions. The chapter considers statistical models for counts of independently occurring random events, and counts at different levels of one or more categorical outcomes. Furthermore, this chapter introduces and illustrates the methods for nonlinear modeling of counts that form the foundation of contemporary categorical data methodology.

DIETARY PROTEIN AND MIXED FUNCTION OXIDASE ACTIVITY

Publisher Summary Diet and nutrition are important environmental factors that may influence mixed function oxidase (MFO) activity and the biological response to xenobiotics. Studies in rats and man utilizing only 2 or 3 diets of different protein concentration have shown that dietary protein is one important nutritional variable controlling MFO activity. This chapter describes an experiment that was conducted to gain more information about the influence of dietary protein on MFO activity. In this study, rats were fed different concentrations of protein. In addition, the role of protein nutrition was assessed by a self-selection dietary regimen that simulates natural conditions. A group of rats was allowed to choose their own protein intake by eating from two feed jars; one containing 5% protein and the other 40% protein. This study indicates that protein excess or deficiency, relative to the 15% protein requirement, influences MFO activity. Similar conclusions are drawn when the data is expressed as total liver activity or activity per unit body weight. Each measure of MFO activity responded uniquely to dietary protein. Cyt. c reductase showed only a twofold change in activity from the lowest to highest observed means. However, cyt. P-450 and AHH showed fourfold and tenfold increases in activity respectively. This study suggests that the amount of dietary protein may have differing effects on the metabolism and biological action of various substrates.

Statistical Values, Quality, and Certification

Abstract Certification of statistical practitioners has been under consideration by the American Statistical Association. This article discusses, through a series of examples, some problematic aspects of statistical practice that certification can help to ameliorate. An optional, limited certification program is advocated as an indirect mechanism for communicating values, concerning the validity of information and responsible use of data, which are not widely understood or attributed to the profession by society at large. How society perceives the roles and professional values of statisticians may be more important to the future of statistics than specific technical issues of practice.