Control-informant agreement on exposure history in case-control studies of Alzheimer's disease

Measuring agreement of multivariate discrete survival times using a modified weighted kappa coefficient

Assessing agreement is often of interest in clinical studies to evaluate the similarity of measurements produced by different raters or methods on the same subjects. We present a modified weighted kappa coefficient to measure agreement between bivariate discrete survival times. The proposed kappa coefficient accommodates censoring by redistributing the mass of censored observations within the grid where the unobserved events may potentially happen. A generalized modified weighted kappa is proposed for multivariate discrete survival times. We estimate the modified kappa coefficients nonparametrically through a multivariate survival function estimator. The asymptotic properties of the kappa estimators are established and the performance of the estimators are examined through simulation studies of bivariate and trivariate survival times. We illustrate the application of the modified kappa coefficient in the presence of censored observations with data from a prostate cancer study.

Modeling the agreement of discrete bivariate survival times using kappa coefficient

Using local kappa coefficients, we develop a method to assess the agreement between two discrete survival times that are measured on the same subject by different raters or methods. We model the marginal distributions for the two event times and local kappa coefficients in terms of covariates. An estimating equation is used for modeling the marginal distributions and a pseudo-likelihood procedure is used to estimate the parameters in the kappa model. The performance of the estimation procedure is examined through simulations. The proposed method can be extended to multivariate discrete survival distributions.

Is pharmacological prevention of Alzheimer's a realistic goal?

A growing body of evidence suggests that several classes of drugs marketed for other indications may be effective in the prevention of Alzheimer's disease. Among the most promising of these are nonsteroidal anti-inflammatory agents, oestrogens (oestrogen replacement therapy) and antioxidant vitamins. Other less well-established candidates include histamine H(2) receptor antagonists (H(2) blockers) and 3-hydroxy-3-methylglutaryl coenzyme A reductase inhibitors (statins). For each of these, we discuss possible mechanisms for their postulated neuroprotective effects and review the studies suggesting their benefits in Alzheimer's disease. We conclude that nonsteroidal anti-inflammatory drugs and oestrogen replacement therapies may be effective in preventing Alzheimer's disease only if taken during the latent phase of the disease several years prior to the appearance of disturbances. Antioxidants may also prevent Alzheimer's disease, but unlike nonsteroidal anti-inflammatory drugs and oestrogen replacement therapies, they may continue to have beneficial effects even after the clinical onset of the disease. The only way to demonstrate the efficacy of these agents will be through randomised, controlled prevention trials. Such trials are currently underway but conclusive results may not be available for several years. Although intriguing, more studies on the neuroprotective effects of statins and H(2) blockers are needed before trials of these agents are initiated. Finally, there are other classes of pharmacological compounds emerging on the horizon, including folic acid, anti-beta-sheet conformational agents, secretase inhibitors and vaccines, that may soon prove to be effective for the prevention of Alzheimer's disease.

Survey into health problems of elderly people: a comparison of self-report with proxy information

BACKGROUND

This study was conducted as part of the Medical Research Council Cognitive Function and Ageing Study.

OBJECTIVE

To compare information given by elderly people on their past and current health and family history of illness with similar information obtained from a relative, friend or carer.

DESIGN

Screening and assessment stages of a prevalence study.

SETTING

Three urban and two rural areas of England and Wales.

PARTICIPANTS

A subsample weighted by age and cognitive status of random population samples of people >/=65 years, living in their own home or in a residential or nursing home, interviewed between 1991 and 1994. A relative, friend or carer identified by each elderly person to provide proxy information. INTERVIEW: Computerized schedules including items on demographic details, cognitive function, lifetime illnesses, current health problems and family history of illness.

RESULTS

The rate of proxy 'don't know' responses and the agreement between the elderly person and their proxy were calculated for each item, both for the overall sample and for subgroups based on characteristics of the respondent and of the proxy and on the relationship between them. Higher 'don't know' rates were found to be particularly associated with more distant relationships, questions on family history, a shorter length of time known and a lack of co-residence. Agreement was strongly related to the nature of the question and less to co-residence, with other factors such as relationship having much smaller effects.

CONCLUSIONS

Proxy information on past and current health problems can be almost complete and in good agreement with self-report, particularly where the proxy lives with the respondent. On family history of illness, history of head injury or boxing and current sleep problems, proxy information is likely to be less complete and show poor agreement. A proxy who is not a close relative is likely to give less complete information but agreement will not be substantially lower.

Survey into health problems of elderly people: multivariate analysis of concordance between self-report and proxy information

BACKGROUND

In many cases probands are unable to give information about their own history. Our objective was to compare information given by elderly people on their past and current health and family history of illness with similar information given by a relative, friend or carer.

MATERIAL

In three urban and two rural areas of England and Wales a sub-sample, weighted by age and cognitive status, was drawn from random population samples of people aged >/=65, not permanently living in hospital, and interviewed between 1991 and 1994, plus for each of these a relative, friend or carer.

METHODS

The relation between the reports of probands and informants on cognitive function, lifetime illnesses, current health problems, and family history of illness was analysed using multiple regression of agreement, measured using Cohen's kappa, and logistic regression of bias and of proportion of 'not known' responses. Potential explanatory variables used were: centre, proband cognitive function, proband sex, informant sex, proband age, difference in age (generation), relationship, frequency of contact, period known, modality of informant interview.

RESULTS

Overall agreement per interview measured by kappa between informant and proband was lower for older probands, probands who were cognitively impaired, and those who saw one another less often; each effect adjusted for all the others using multiple regression. Informants tended to over-report relative to probands if the proband was cognitively impaired, male, they themselves were female, of a younger generation, and co-resident; each effect adjusted for all the others using logistic regression. Interviews were more likely to be missing at least one item if the informant was not the spouse, had known the proband for less time, and was not co-resident; each effect adjusted for all the others using logistic regression.

CONCLUSION

There is reasonable accord between probands and informants which is enhanced by close contact, long period of knowledge, and a female informant. Degree of kinship is not critical. Informant interviews may be conducted by telephone rather than face-to-face without serious loss of quality.

Testing the equality of two dependent kappa statistics

Procedures are developed and compared for testing the equality of two dependent kappa statistics in the case of two raters and a dichotomous outcome variable. Such problems may arise when each of a sample of subjects are rated under two distinct settings, and it is of interest to compare the observed levels of inter-observer and intra-observer agreement. The procedures compared are extensions of previously developed procedures for comparing kappa statistics computed from independent samples. The results of a Monte Carlo simulation show that adjusting for the dependency between samples tends to be worthwhile only if the between-setting correlation is comparable in magnitude to the within-setting correlations. In this case, a goodness-of-fit procedure that takes into account the dependency between samples is recommended.

Some contributions to the epidemiology of dementia and depression

OBJECTIVE

To give an account of how a research programme in epidemiology at the NHMRC Social Psychiatry Research Unit developed, and an overview of work completed.

DATA SOURCES

General population samples of elderly persons; and persons clinically diagnosed as having Alzheimer's disease, or as being free of dementia.

STUDY SELECTION

The studies undertaken include integrative analyses of published work; instrument development; cross-sectional and prospective longitudinal surveys of cognitive decline, dementia and depression in general population samples; and a case-control study of Alzheimer's disease.

DATA SYNTHESIS

The data are of two types: continuous and categorical measures of morbidity; and measures of personal, social and biological attributes of the sample, including assessment of disablement and use of services. The findings have been both descriptive and analytic. New instruments developed are the Canberra Interview for the Elderly (CIE), the IQCODE and the Psychogeriatric Assessment Scales (PAS).

CONCLUSION

Beyond simple estimates of the prevalence of dementia, cognitive impairment and depression, we have focused mainly on aetiology and outcome in community samples. Unexpected opportunities have had a substantial influence on the direction of our work.

Reporting source bias in estimating prevalence of cognitive impairment

Reporting source bias was examined using cognitive data from a probability sample of adult day health care (ADHC) clients. Data were obtained from four sources: client, family, staff, and chart. These data suggest that prevalence estimates are influenced by the reporting source and method of assessment, with chart diagnoses yielding the least convergent estimates. Examination of agreement summary statistics and bias indices show that both staff and family underreport prevalence of cognitive impairment, but that more bias is associated with staff than is with family sources. Such bias should be considered in studies of prevalence estimation of cognitive impairment.

Use of surrogate respondents in studies of stroke and dementia

A study was performed to determine the misclassification of information on exposure when surrogate informants are interviewed. Information from nondementia patients admitted to a geriatric ward was compared with that obtained from the patients' family members or friends (informants). Thirty-seven patients were recruited (17 with current or previous stroke, and 20 no history of stroke). Information on past medical and family history, cigarette smoking and alcohol consumption was collected using a questionnaire that was administered independently to patients and informants. For nonstroke patients, agreement on smoking and alcohol consumption at different ages was generally good. Informants reported slightly lower smoking and alcohol intake than patients: 1.3 fewer total pack-years (95% CI, -3.8, +6.5) and 2.7 fewer units/week (95% CI, -1.5, +6.9), respectively. For stroke patients, informants tended to report a greater smoking consumption at all ages, and a slightly higher alcohol consumption. Informants reported an excess of 7.4 total pack-years (95% CI, -1.2, + 16.0) and 2.19 units/week (95% CI, -1.0, +5.3). Agreement for medical history was good, except for hypertension. There was poor agreement for family medical history. The impact of misclassification is discussed.

Risk factors in clinically diagnosed presenile dementia of the Alzheimer type: a case-control study in northern England

STUDY OBJECTIVE

To investigate the relationship between presenile dementia of the Alzheimer type (PDAT) and family history, medical history, cigarette smoking, and exposure to aluminum.

DESIGN

A case-control study in which 109 cases of clinically diagnosed PDAT and 109 controls matched for age and sex were compared for exposure to the risk factors. Odds ratios (ORs) were calculated using McNemar's test.

SETTING

The northern health region of England.

PATIENTS

Cases comprised those under 65 years diagnosed as having dementia by specialist services, who met clinical algorithm criteria for Alzheimer's disease (AD). Cases were confirmed at interview.

MAIN RESULTS

Comparing cases with controls, (ORs) significantly greater than unity were obtained when there was a first degree relative with dementia (OR 2.5, 95% confidence interval 1.05, 6.56), any relative with dementia (OR 2.1, 95% CI 1.01, 4.55), and any relative aged less than 65 with dementia (OR 8.0, 95% CI 1.07, 348). Exposure to moderate levels of cigarette smoking (cumulative) was not significant; nor was exposure to aluminum in drinking water, diet, and medicinal sources.

CONCLUSION

In this study of modest statistical power, a family history of dementia was confirmed as a risk factor in PDAT. No significant relationship between exposure to aluminium in water supplies, tea, and antacids was found. What is important, however, is the bioavailability of all dietary aluminium, determined by the concentrations of dissolved silicon in water: this requires further investigation.