Loglinear models using capture-recapture methods to estimate the size of a measles epidemic.

Cases of measles reported separately by doctors, hospitals, laboratories, and others are treated as separate captures and releases of the population of children aged up to 10 years who have been infected by measles in a recent epidemic. Loglinear models are used to assess the adequacy of fit of various models for capture-recapture and the size of the epidemic in different age groups is estimated.

The inheritance of serum cholesterol: adjustment of observed cholesterol levels for age, sex and body weight using inverse-polynomial regression.

Adjustment of observed serum cholesterol levels for biological and environmental variables is an essential step in studying the mode of inheritance of serum cholesterol. Body weight, measured as weight/height2, is found to be an important variable not previously taken into account in regression equations. A method is given for adjusting cholesterol levels for age, sex and body weight, thus obtaining residuals which represent the deviations of cholesterol levels from the mean of the general population for each category of these three factors. The procedure for incorporating into the model other variables, that are found to have a significant effect on cholesterol level, is described. Polynomials and inverse polynomials of degree three fit the data equally well over a restricted range. However, inverse polynomials are chosen as they appear to have a more appropriate shape and are likely to agree better with observed values over a greater age range.

Trends in the prevalence of trachoma, South Australia, 1976 to 1990.

The eye health of rural Aboriginal Australians is known to be poor. Over the past 20 years, Aboriginal communities in remote parts of Australia have had increasing access to eye health services through the National Trachoma and Eye Health Program (NTEHP). Using published and unpublished data, we examined trends in the prevalence of inflammatory trachoma in the Anangu Pitjantjatjara of South Australia. Comparisons using a generalised linear model of surveys in 1976, 1985 and 1990 indicate that there has been a significant reduction in the age-standardised prevalence of inflammatory trachoma in 0- to 20-year-olds. When the 1990 survey was compared with 1976 interim report data from the NTEHP survey, the odds of inflammatory trachoma in 1990 were 0.25 (95 per cent confidence interval (CI) 0.18 to 0.35). When the comparison was with data from the NTEHP survey of the Red Centre, the odds of follicular trachoma in 1990 were 0.51 (CI 0.42 to 0.62), and in comparison with the 1985 NTEHP review data, the odds of inflammatory trachoma in 1990 were 0.28 (CI 0.20 to 0.39). In the older age groups (20 and over), an increase in the prevalence of inflammatory trachoma was found. Although significant, the increase affected a small proportion of the population and may have been because of difficulty in standardising the trachoma grading between surveys, or systematic grading error in the 1990 survey. This study therefore shows that the eye health of Aboriginal people in Central Australia may be improving. The decline in trachoma is welcome and may be caused by improvements in socioeconomic conditions, community development and increasing access to medical care.

Variance components for discordances.

Tests for biotyping isolates give a result that is classified as either positive or negative, indicative of growth or nongrowth of bacteria. The reproducibility of such tests is measured by the number of discordances in replicates of the same measurement. In this analysis the probability distribution of the number of discordances is estimated for each of several tests in the presence of possible random between-laboratory effects as well as random laboratory-test interactions, which are also estimated.

REML estimation for survival models with frailty.

A method of estimation for generalised mixed models is applied to the estimation of regression parameters in proportional hazards models for failure times when there are repeated observations of failure on each subject. The subject effect is incorporated into the model as a random frailty term. Best linear unbiased predictors are used as an initial step in the computation of maximum likelihood and residual maximum likelihood estimates.

Estimation of cumulative illness using cross-sectional data.

Using cross-sectional data, a general method is given for assessing cumulative illness due to a particular disease. An application is given to estimating cumulative illness due to otitis media in Australian aborigines and contrasting these results to the non-aboriginal population.

A model for the analysis of growth data from designed experiments.

A model for growth data from designed experiments is presented which extends the stochastic differential equation of Sandland and McGilchrist (1979, Biometrics 35, 255-272). Residual maximum likelihood (REML) is used to estimate the parameters of the model. The model is easily extended to incomplete data and is shown to overcome some of the practical difficulties encountered with the profile model. The procedure is applied to data from experiments on pigs and sheep.

Regression with frailty in survival analysis.

In studies of survival, the hazard function for each individual may depend on observed risk variables but usually not all such variables are known or measurable. This unknown factor of the hazard function is usually termed the individual heterogeneity or frailty. When survival is time to the occurrence of a particular type of event and more than one such time may be obtained for each individual, frailty is a common factor among such recurrence times. A model including frailty is fitted to such repeated measures of recurrence times.

ML and REML estimation in survival analysis with time dependent correlated frailty.

In the study of multiple failure times for the same subjects, for example, recurrent infections for patients with a given disease, there are often subject effects, that is, subjects have different risks that cannot be explained by known covariates. Standard methods, which ignore subject effects, lead to overestimation of precision. The frailty model for subject effects is better, but can be insufficient, because it assumes that subject effects are constant over time. Experience has shown that the dependence between different time periods often decreases with distance in time. Such a model is presented here, assuming that the frailty is no longer constant, but time varying, with one value for each spell. The main example is a first-order autoregressive process. This is applied to a data set of 128 patients with chronic granulomatous disease (CGD), participating in a placebo controlled randomized trial of gamma interferon (gamma-IFN), suffering between 0 and 7 infections. It is shown that the time varying frailty model gives a significantly better fit than the constant frailty model.

Threshold models in a methadone programme evaluation.

The degree of non-compliance with a methadone programme is such that a zero is often recorded as the response variable. Such outcome variables cannot be transformed to normality in order to investigate a regression relationship between the response variable and various risk or treatment variables. The presence of a random effect as well in this regression causes added difficulty in the analysis. We show here that use of threshold models can provide a satisfactory solution to both these problems.

A Markov transition model in the analysis of the immune response.

A method for the analysis of repeated measures of immune response is given. Typically the reactions of response to stimulus do not occur at the same time for all experimental units so that regression models are inappropriate for this type of data. Instead Markov transition models are suggested and procedures given for testing differences among response patterns.

Environmental Factors Influencing Numbers of Rhizobium leguminosarum biovar trifolii and Its Bacteriophages in Two Field Soils.

Fluctuations in numbers of Rhizobium leguminosarum biovar trifolii and its bacteriophages in two fields with different soil types were followed during a 17-month period in 1981 and 1982. Mean levels of both phage and rhizobia varied significantly (P < 0.05) on different occasions, with rhizobial levels varying from 1.6 x 10 to 2.0 x 10 cell per g of soil and phage from 0 to 1.7 x 10 PFU/g of soil. Multivariate regression analysis showed rhizobial levels to be significantly and positively related to vegetation height and solar radiation, but not to mean temperature, precipitation, soil matric potential, or soil type. Rhizobiophage concentrations were significantly and positively related to soil matric potential and vegetation height. They were reduced in the silty clay loam soil, although the presence of 34% clay did not prevent phage multiplication and the occurrence of high phage levels.

The synergistic effect of weight loss and changes in dietary lipids on the serum cholesterol of obese men with hypercholesterolaemia: implications for prevention of coronary heart disease.

The hypolipidaemic effect of a low-fat, low-cholesterol diet, alone and in combination with weight reduction, has been evaluated in two groups of obese men with hypercholesterolaemia. In 41 men who lost 10.3 kg over 11 months and maintained their lower weight for 23.5 months serum cholesterol fell by 1.68 mmol/l and remained steady at lower weight. In 20 similar men, the controls, whose weight fell by 0.8 kg over 39.5 months, serum cholesterol fell by 0.80 mmol/l. There was a significant linear correlation between change in weight and change in serum cholesterol. The change in serum cholesterol in the weight losers was greater than could be accounted for by change in dietary lipids. Weight reduction and low-fat, low-cholesterol diets appear to have a synergistic effect in reducing serum cholesterol.

Survival analysis and prognosis for dementia patients.

How long a dementia patient is cared for in the home before admission to a nursing home depends on the state of the patient and the state of the caregiver. Using 5-year follow-up data, the times until entry to nursing home and until death are modeled using a Cox survival model in which patient and caregiver variables at entry to study as well as changes in these variables over the following 12 months are the regression variables. Treatment variables quantify the effects of a caregiver training program. Statistical methods used in fitting the Cox survival model and consequent predictions of survival rates are discussed.

Estimation of environmental and genetic components of quantitative traits with application to serum cholesterol levels.

A mixed model of environmental, polygenic, and major locus effects is developed, allowing for environmental correlations between first-degree relatives and spouses. Maximum-likelihood techniques are used to determine the relative contributions of each of these effects to a quantitative trait. Inclusion of a nuclear family in the sample is assumed to depend on the value of the quantitative trait of one member of the family, so conditional distributions are used. Application of the method to serum cholesterol data from the general population shows that the addition of a polygenic effect to a model that assumes only an environmental effect makes a significant improvement. A completely dominant single gene is also found to be influencing serum cholesterol levels. Although cholesterol levels have been adjusted for a range of factors, such as age, sex, weight/height, and marital status, environmental factors still account for about half the variability in the residual values.

The effects of body weight on serum cholesterol, serum triglycerides, serum urate and systolic blood pressure.

The effects of body weight and age on serum cholesterol, serum triglycerides, serum urate and systolic blood pressure were examined in 600 male and 400 female blood donors aged 20 to 49 years. In the men significant correlations with body mass index were found for all four variables in each decade. In the women below 40 only the correlation with blood pressure was significant. In the fifth decade the correlations resembled those in the men, save for triglycerides. After adjusting for weight, age had no independent influence on the prevalence of hypertriglyceridaemia or hypertension in either sex. In men the effect of body weight on the prevalence of hypercholesterolaemia was age dependent. Age influenced hyperuricaemia independently of weight. In women only serum cholesterol was affected by age after allowing for weight. There was marked clustering of high values of the four variables in the heavier men and women and this increased with age. The leanest men and women were remarkably free of high values.

Smoking: a major predictor of left ventricular function after occlusion of the left anterior descending coronary artery.

The major predictors of left ventricular function after coronary artery occlusion were assessed in 108 consecutive patients who had complete occlusion of the left anterior descending artery as the only important lesion demonstrated at angiography between June 1978 and June 1983. A scoring system was used to identify regional damage on left ventriculograms. Forty two patients were classified as having good left ventricular function and 66 as having varying degrees of impairment. Apart from a history of myocardial infarction, the only variables discriminating between those with good and those with impaired left ventricular function were the area of distribution of the artery beyond the occlusion and cigarette smoking. Hypertension, hypercholesterolaemia, family history of vascular disease, diabetes, obesity, duration of angina, age, and presence of identifiable collaterals were not discriminators. Smoking was itself significantly associated with a history of infarction; but after controlling for this, smoking exerted a significant additional effect on the amount of left ventricular damage. It is concluded that smoking is not only a risk factor for myocardial infarction in patients with single left anterior descending artery occlusion, but that it is also a major factor in determining the extent of associated left ventricular damage.

Percutaneous aortic valvuloplasty.

OBJECTIVE: To determine the short-term and long-term results of percutaneous aortic valvuloplasty. DESIGN: A retrospective follow-up study. SETTING: The Cardiac Catheterisation Laboratory and Cardiovascular Medicine Unit of a teaching hospital. PATIENTS: Eighteen patients with severe aortic stenosis who were not accepted for surgery; there were 11 men and 7 women, mean age 79 +/- 4.5 years. INTERVENTIONS: Percutaneous transluminal aortic valvuloplasty (PTAV) by means of balloon catheter techniques. MAIN OUTCOME MEASURES: Patient survival and symptomatic status. RESULTS: PTAV resulted in a significant decrease in the aortic valve pressure gradient from 64.8 +/- 23.2 mmHg to 38 +/- 14.7 mmHg (mean +/- SD) (P less than 0.00001) in the 18 patients and a significant increase in the mean aortic valve area from 0.4 +/- 0.16 cm2 to 0.6 +/- 0.18 cm2 (P less than 0.0001) in 14 patients. Complications occurred in seven patients; two of them, who had been in terminal heart failure with a low output state before PTAV, died. Fifteen patients improved in at least one New York Heart Association functional class early after PTAV and one patient had an aortic valve replacement. In follow-up of between 5 and 32 months (mean, 13.3 +/- 7.7 months) six patients are in a better functional class than before PTAV, one patient is in Class IV, one patient had an aortic valve replacement and seven patients died (three died of cardiac failure and four of non-cardiac causes). CONCLUSIONS: PTAV has a place as a palliative procedure in selected patients with aortic stenosis in whom another condition precludes aortic valve replacement.