Search

Access to safe breast-feeding alternatives for HIV-infected mothers and their infants in many settings is limited. We compared the rates of early postpartum hospitalization of infants born to HIV-infected mothers using different infant-feeding practices in a large government hospital in Pune, India. From March 1, 2000 to November 30, 2001, infants born to HIV-infected mothers were followed in a postpartum clinic. All mothers had received a standard short course of antenatal zidovudine. Infant-feeding practices were assessed within 3 d of delivery, prior to postpartum hospital discharge. Sixty-two of 148 mothers (42%) were breast-feeding their infants. Eighty-six of the mothers (58%) were providing replacement feeding, primarily diluted cow, goat or buffalo milk (top feeding). Twenty-one of the 148 participating infants (14.2%) born during the study period required hospitalization within the 1st 6 mo of life and 6 infants required repeat hospitalization. All hospitalized infants were receiving replacement feeding with a rate of 0.093 hospitalizations per 100 person-days (95% CI, 0.062 to 0.136). The reasons for hospitalization included acute gastroenteritis (48.1%), pneumonia (18.5%), septicemia (11.1%) and jaundice (11.1%). A high risk for early postpartum hospitalization was seen in replacement-fed infants born to HIV-infected mothers in Pune, India. In settings such as India, where access to safe replacement feeding is limited, interventions making exclusive breast-feeding safer for HIV-infected mothers and infants are needed. Such interventions would be valuable additions to the very effective national prevention programs that currently rely on the provision of short-course zidovudine and nevirapine.

A panel study consists of individuals who have data collected at periodic follow-up visits or pre-specified time points following entry into the study. The objective of this paper is to consider design issues in a panel study when the response variable is the stage of disease, and with focus on the transition intensities. Important design issues include the choice of the time interval between follow-up visits and sample size considerations. We study the effects of time intervals between follow-up visits on the precision of the transition intensities estimators. We also consider the power of statistical tests on the ratio of transition intensities. Discussion is extended to incorporate heterogeneity in the population in which frailty is introduced to describe subject-specific transition intensities.

A critical consideration in effective and measured public health responses to an outbreak of inhalational anthrax is the optimum duration of antibiotic prophylaxis. We develop a competing-risks model to address the duration of antibiotic prophylaxis and the incubation period that accounts for the risks of spore germination and spore clearance. The model predicts the incubation period distribution, which is confirmed by empirical data. The optimum duration of antibiotic prophylaxis depends critically on the dose of inhaled spores. At high doses, we show that exposed persons would need to remain on antibiotic prophylaxis for at least 4 months, and considerable morbidity would likely occur before antibiotic prophylaxis could even be initiated. At very low doses, 60 days of antibiotic prophylaxis is adequate. Exposure doses can be estimated from the cumulative attack rate up to the point antibiotic prophylaxis begins. The model explains that whereas < or =60 days of antibiotics were enough to protect persons in the 2001 U.S. outbreak, because doses were very low, at moderate or high doses considerably longer durations would be necessary to adequately protect exposed populations.

Background: Recent studies have suggested that AD may reflect a chronic process that begins many years before the clinical expression of dementia. The current study examines premorbid Benton Visual Retention Test (BVRT) and Wechsler Adult Intelligence Scale-vocabulary (WAIS-voc) test scores in order to determine whether long-term deficits in these tests can predict the development of AD decades later in the Baltimore Longitudinal Study of Aging (BLSA)., Method: Participants are volunteers from the BLSA, a multidisciplinary study of normal aging conducted by the National Institute on Aging. A total of 1,425 BLSA participants who were older than 60 years were included in the analyses. Cox proportional hazards models were used to estimate the relative risk of developing AD associated with BVRT and WAIS-voc scores at different time periods up to 20 years before the diagnosis of AD., Results: The relative risks for 6 or more BVRT errors vs less than 6 errors at 1 to 3, 3 to 5, 5 to 10, and 10 to 15 years before the diagnosis of AD were 5.69, 2.11, 1.76, and 1.83 (p < 0.05). The relative risk for 15 or more years before diagnosis was not significant (p > 0.10). WAIS-voc scores were not significantly associated with the risk of AD in any time period., Conclusions: A greater number of errors on the BVRT is associated with an increased risk of AD up to 15 years later. Poor visual memory performance may represent an early expression of AD years before diagnosis. This result suggests the need to continue to revise views on the natural history of AD and the possibility of an increased window of opportunity for preventive treatment before definitive diagnosis.

Objectives: Estimates of the HIV/AIDS burden in Burma (Myanmar) are uncertain. Using data from the 1999 national HIV sentinel surveillance and available population data, we generated estimates of Burma's HIV burden in 1999., Methodology: The 1999 sentinel surveillance included women attending antenatal clinics, male military recruits, blood donors, injecting drug users, patients of sexually transmitted disease clinics, and sex workers. We used data for women attending antenatal clinics and male recruits aged 20-29 years to estimate HIV prevalence among women and men, respectively. Data points were merged to give five regional estimates of prevalence for men and women. Census figures were used to obtain national population estimates of the numbers of Burmese living with HIV infection, along with confidence intervals (CIs)., Results: HIV prevalence varied by region, with the lowest rates in the West, intermediate rates in the central region, and highest rates in the North, East, and South. The highest rates were in the East (Shan State), with female prevalence of 3.0% (95% CI, 1.9-4.5). The total number of infected women nationwide was 218,300 (95% CI, 159,400-277,100), and that of men was 468,700 (95% CI, 343,300-594,200). We estimated HIV prevalence of at least 3.46% (95% CI, 2.72-4.19) among adults aged 15-44 years; 5700 infants were born with HIV infection in 1999., Discussion: Burma has a generalized epidemic of HIV-1 in reproductive age adults. We estimated that there were 687,000 (95% CI, 541,100-832,900) Burmese adults living with HIV infection in 1999, or about one of every 29 adult citizens. This estimate is higher than the UNAIDS estimate for the same year of 530,000 adults and children living with AIDS, or a population prevalence of about one in 50 adults. HIV prevention and care programs are urgently needed in Burma.

Anti-HIV-1-specific T cell responses in early HIV-1 infection have been found to be important in deciding the course of disease progression. But there are few data concerning nonsubtype B HIV infection. HIV-1 subtype C is the most prevalent subtype in India. HIV-1 Gag-specific T cell responses in 12 Indian subjects with recent HIV-1 infection were characterized by an ELISpot assay at two consecutive visits and their correlation with plasma viral load and CD4(+) T lymphocyte counts was studied. Ten of the 12 subjects demonstrated T cell responses to either one or both Gag B and C peptides, on at least one visit. Five of 10 responders showed a consistent response (response at both visits): 4 exhibited a Gag C-specific consistent response and 1 showed a consistent response to Gag B. The remaining five patients, showing response at only one of the two visits, were considered inconsistent responders. None of the individuals showed a consistent response to both B and C Gag peptides. Marginally significant correlation was observed between consistency of the response and lower plasma viral load (p = 0.062). The subtype-specific Gag C response was also found to be correlated with lower viral load as compared with the response to Gag B (r = -0.336, p = 0.054 for subtype C and r = -0.234, p = 0.13 for subtype B). The data suggest that the patients exhibiting consistent subtype-specific responses to HIV-1 Gag might have better control of viral replication in early HIV infection.

Context: Survival following a diagnosis of Alzheimer disease (AD) is important information for health planners, caregivers, patients, and their families., Objectives: To estimate the duration of survival following a diagnosis of AD and to determine the effect of AD on life span., Design, Setting, and Participants: Follow-up of participants of the Baltimore Longitudinal Study of Aging who were older than 55 years (January 1, 1985-September 30, 1999)., Main Outcome Measures: Survival duration., Results: The median survival time following a diagnosis of AD depended strongly on the patient's age at diagnosis. The median survival times ranged from 8.3 years for persons diagnosed as having AD at age 65 years to 3.4 years for persons diagnosed as having AD at age 90 years. There were no significant differences between men and women in survival after having a diagnosis of AD. Diagnoses of AD at ages 65 and 90 years were associated with approximately a 67% and 39% reduction in median life span, respectively., Conclusions: The effect of a diagnosis of AD on life span depends crucially on the age of the person when AD is diagnosed. Caregivers, patients, and their families could plan on a median life span as long as 7 to 10 years for patients whose conditions are diagnosed when they are in their 60s and early 70s, to only about 3 years or less for patients whose conditions are diagnosed when they are in their 90s.

The objective of this paper is to estimate survival curves for two different exposure groups when the exposure group is not known for all observations, and the data is subject to left truncation and right censoring. The situation we consider is when the probability that the exposure group is missing may depend on whether the observation is censored or uncensored, in which case the exposure is not missing at random. The problem was motivated by a study of Alzheimer's disease to estimate the distribution of ages at diagnosis for individuals with and without an apolipoprotein E4 allele (the exposure group). Genotyping for this risk factor was incomplete and performed more frequently on the cases of Alzheimer's disease (the uncensored observations) than the censored observations. The survival curves are estimated in discrete time using an EM algorithm. A bootstrapping procedure is proposed that guarantees each bootstrap sample has the same proportion of observations with missing exposure. A simulation is performed to evaluate the bias of the estimators and to investigate design and efficiency issues. The methods are applied to the Alzheimer's disease study., (Copyright 2002 John Wiley & Sons, Ltd.)

Background: Care for chronic renal failure involves management of complications and preparation for possible dialysis. Patients often are not evaluated by nephrologists in a timely manner., Objective: To identify factors associated with late evaluation by a nephrologist and to assess whether late evaluation is associated with worse survival once patients develop end-stage renal disease (ESRD)., Design: National prospective cohort study., Setting: 81 dialysis facilities throughout the United States., Patients: 828 patients with new-onset ESRD., Measurements: Time from first evaluation by a nephrologist to initiation of dialysis, classified as late (<4 months), intermediate (4 to 12 months), or early (>12 months); rate of death, from initiation of dialysis to an average of 2.2 years of follow-up; and demographic, clinical, and laboratory characteristics., Results: After adjustment for potential confounders, late evaluation was more common among black men than white men (44.8% vs. 24.5%; P < 0.05), uninsured patients than insured patients (56.7% vs. 29.0%; P < 0.05) and patients with severe comorbid disease than those with mild comorbid disease (35.0% vs. 23.0%; P < 0.05). Compared with patients who had early evaluation, the risk for death was greater among patients evaluated late and was graded (hazard ratio, 1.3 [95% CI, 0.87 to 2.06] for patients with intermediate evaluation and 1.8 [CI, 1.21 to 2.61] for those with late evaluation) after adjustment for dialysis method, demographic characteristics, and socioeconomic status in Cox proportional hazards regression analysis. After additional adjustment for such factors as the presence and severity of comorbid conditions, the association remained graded (hazard ratio, 1.2 [CI, 0.73 to 1.82] for patients evaluated at an intermediate point and 1.6 [CI, 1.04 to 2.39] for those evaluated late)., Conclusions: Late evaluation of patients with chronic renal failure by a nephrologist is associated with greater burden and severity of comorbid disease, black ethnicity, lack of health insurance, and shorter duration of survival.

Vitamin A deficiency is very prevalent and contributes substantially to morbidity and mortality among young children in developing countries. We identify and quantify three causes of vitamin A deficiency in young children based on data available in the literature: maternal vitamin A deficiency resulting in low concentrations of vitamin A in breast milk, inadequate dietary intake of vitamin A during and after weaning and prevalent illness. We developed a set of recursive equations to estimate the amount of vitamin A in the liver as a function of age over the first 2 y of life. To apply the equations, we selected a best estimate value for each input parameter as the most representative of a typical child in a developing country. Because of the great variability that exists for each variable, we also carried out sensitivity analyses, substituting more extreme values for input parameters. We then estimated stores, assuming a child in a developing country also receives the newly revised vitamin A supplementation regimen recommended by the World Health Organization. Without supplementation, a typical child in a developing country is not able to attain and maintain "minimally adequate" liver vitamin A stores. To overcome this deficit by eating fruits and vegetables alone, the child would need to increase portion sizes about 10-fold. If the child receives the new supplementation regimen, his or her liver stores will still be far short of the average American child (i.e., exceedingly far from toxic levels). However, our estimates indicate that the new supplementation regimen will permit a typical child in a developing country setting to attain minimally adequate vitamin A stores during the first 2 y of life.

Background: The epsilon 4 allele of the APOE gene (APOE) is more frequent in patients with AD than in the general population, but studies are inconclusive as to whether it affects rate of progression or survival. Because survival in AD is generally longer in women than in men, the authors investigated whether APOE affects 10-year survival equally in men and women., Methods: APOE testing was performed on 125 patients with probable AD enrolled in the Johns Hopkins AD Research Center between November 1984 and March 1987. The 39 men and 86 women were followed at 6-month intervals until censoring (by death or withdrawal from the study) or March 1997. Patients were dichotomized into those with and those without at least one epsilon 4 allele. For each sex, a Cox proportional hazards regression, allowing for delayed entry and covarying for age at onset, was used to examine the effect of epsilon 4 on survival., Results: All patients who died during the study period and had autopsy (n = 92) were found to have definite AD. Average survival from disease onset did not differ by sex (12.1 years in men; 12.3 years in women). In neither sex were differences found between epsilon 4-positive and epsilon 4-negative subgroups in education, duration of AD at entry, or severity of dementia. However, in both sexes the epsilon 4-positive subgroup was approximately 3 years older at onset of AD and at entry to the study than the epsilon 4-negative subgroup. Adjusting for age at onset, the presence of an epsilon 4 allele significantly increased the relative risk of death only for men (RR = 2.69; 95% CI = 1.23 to 5.87)., Conclusions: In this sample of mostly white, well-educated research participants with AD, the APOE epsilon 4 allele was associated with shorter survival in men but not in women.

Some studies have indicated that needle exchange programs (NEPs) can be effective in reducing drug-related risks for human immunodeficiency virus (HIV) seroconversion; however, others have reported higher HIV incidence rates among NEP attendees. Since many studies rely on self-reports of NEP attendance, the authors investigated the extent to which differential misreporting of NEP attendance could bias risk estimates. Over a 3-year period from 1994 to 1997, self-reports of NEP attendance from participants in a prospective study in Baltimore, Maryland, were compared with NEP records. Of 1,315 participants, 459 (35%) had registered with the Baltimore NEP. There was 86.7% concordance between self-reported and actual NEP use; 11.0% reported NEP attendance but did not attend (overreported), and 2.2% reported not attending NEP but did attend (underreported). In multivariate analyses using generalized estimating equations, persons who overreported NEP attendance were more likely to have injected frequently (adjusted odds ratio (AOR) = 1.29, 95% confidence interval (CI): 1.04, 1.61), denied needle sharing (AOR = 0.69; 95% CI: 0.52, 0.89), and been an HIV seroconverter (AOR = 1.83, 95% CI: 1.11, 3.01). With Poisson regression to model predictors of HIV seroconversion, models that included measures of NEP attendance based on self-reports compared with actual program data underestimated a protective association by 18%. These findings have important implications for evaluations of NEPs.

An outbreak of anthrax occurred in the city of Sverdlovsk in Russia in the spring of 1979. The outbreak was due to the inhalation of spores that were accidentally released from a military microbiology facility. In response to the outbreak a public health intervention was mounted that included distribution of antibiotics and vaccine. The objective of this paper is to develop and apply statistical methodology to analyse the Sverdlovsk outbreak, and in particular to estimate the incubation period of inhalational anthrax and the number of deaths that may have been prevented by the public health intervention. The data available for analysis from this common source epidemic are the incubation periods of reported deaths. The statistical problem is that incubation periods are truncated because some individuals may have had their deaths prevented by the public health interventions and thus are not included in the data. However, it is not known how many persons received the intervention or how efficacious was the intervention. A likelihood function is formulated that accounts for the effects of truncation. The likelihood is decomposed into a binomial likelihood with unknown sample size and a conditional likelihood for the incubation periods. The methods are extended to allow for a phase-in of the intervention over time. Assuming a lognormal model for the incubation period distribution, the median and mean incubation periods were estimated to be 11.0 and 14.2 days respectively. These estimates are longer than have been previously reported in the literature. The death toll from the Sverdlovsk anthrax outbreak could have been about 14% larger had there not been a public health intervention; however, the confidence intervals are wide (95% CI 0-61%). The sensitivity of the results to model assumptions and the parametric model for the incubation period distribution are investigated. The results are useful for determining how long antibiotic therapy should be continued in suspected anthrax cases and also for estimating the ultimate number of deaths in a new outbreak in the absence of any public health interventions.

Background: The probability of HIV-1 transmission per coital act in representative African populations is unknown. We aimed to calculate this probability overall, and to estimate how it is affected by various factors thought to influence infectivity., Methods: 174 monogamous couples, in which one partner was HIV-1 positive, were retrospectively identified from a population cohort in Rakai, Uganda. Frequency of intercourse and reliability of reporting within couples was assessed prospectively. HIV-1 seroconversion was determined in the uninfected partners, and HIV-1 viral load was measured in the infected partners. Adjusted rate ratios of transmission per coital act were estimated by Poisson regression. Probabilities of transmission per act were estimated by log-log binomial regression for quartiles of age and HIV-1 viral load, and for symptoms or diagnoses of sexually transmitted diseases (STDs) in the HIV-1-infected partners., Results: The mean frequency of intercourse was 8.9 per month, which declined with age and HIV-1 viral load. Members of couples reported similar frequencies of intercourse. The overall unadjusted probability of HIV-1 transmission per coital act was 0.0011 (95% CI 0.0008-0.0015). Transmission probabilities increased from 0.0001 per act at viral loads of less than 1700 copies/mL to 0.0023 per act at 38 500 copies/mL or more (p=0.002), and were 0.0041 with genital ulceration versus 0.0011 without (p=0.02). Transmission probabilities per act did not differ significantly by HIV-1 subtypes A and D, sex, STDs, or symptoms of discharge or dysuria in the HIV-1-positive partner., Interpretation: Higher viral load and genital ulceration are the main determinants of HIV-1 transmission per coital act in this Ugandan population.

Background: Renal transplantation is the optimal treatment for persons with end-stage renal disease (ESRD). A shortage of kidneys in the U.S. has focused increasing attention on the process by which kidneys are allocated. A national survey was undertaken to determine the relative importance of both clinical and nonclinical factors in the recommendation for renal transplantation by U.S. nephrologists., Methods: We conducted a national random survey of 271 U.S. nephrologists using hypothetical patient scenarios to determine their recommendation for renal transplantation based on demographic, clinical, and social factors. Specifically, eight unique patient scenarios were randomly distributed to each survey respondent., Results: According to responding nephrologists (response rate 53%), females were less likely than males to be recommended for renal transplantation [adjusted odds ratio (OR)=0.41; confidence interval (CI) 0.21, 0.79; for whites]. Asian males were less likely than white males to be recommended for transplantation (OR=0.46, CI 0.24, 0.91). Black-white differences in rates of recommendation were not found. Other factors associated with low rates of recommendation for renal transplantation included history of noncompliance (OR=0.17, CI 0.13, 0.23), <25% cardiac ejection fraction (OR=0.15, CI 0.10, 0.21), HIV infection (OR=0.01, CI 0.00, 0.01), and being >200 lbs (OR=0.73, CI 0.56, 0.95)., Conclusions: Female gender, and Asian but not black race, were associated with a decreased likelihood that nephrologists would recommend renal transplantation for patients with end stage renal disease. The well-documented black-white disparities in use of renal transplantation may be due to unaccounted for factors or may arise at a subsequent step in the transplantation process.

Objective: To develop a pooling method for detection of viral RNA for diagnosis of acute HIV infection and estimation of HIV-1 incidence., Methods: Sera from 700 consecutive seronegative patients attending sexually transmitted disease clinics in Pune, India, were screened individually for p24 antigen, and pooled into seven pools of 100 for detection of HIV-1 RNA by reverse transcriptase-polymerase chain reaction. HIV-1 incidence was calculated by the traditional cohort method, the p24 antigen method, and a multistage pooling method in which RNA-positive pools were re-analyzed in smaller pools., Results: Sera from 700 individuals were grouped into seven pools of 100, of which four were positive. These four positive pools were subdivided into eight pools of 50, of which seven were positive. The seven positive pools were subdivided into 35 pools of 10, of which 10 were positive. Based on the 10 RNA-positive pools, the point estimate of HIV-1 incidence was 19.9% per year [95% confidence interval (CI), 7.3-31.8%]. Of the 700 samples analyzed for p24 antigen, eight were positive, resulting in a point estimate of incidence of 18.5%/year (8.0-36.5%). In contrast, the incidence rate based on the traditional cohort method of follow-up was lower at 9.4%/year (4.8-16.4%) due to a low follow-up rate. Testing of individual samples from the 10 RNA-positive pools identified 10 individuals with acute primary HIV-1., Conclusion: The multistage pooling method for detection of HIV-1 RNA was more sensitive than the p24 antigen method, and was five-fold less expensive than the p24 antigen assays. Pooling samples for RNA detection was effective in estimating current incidence rates with cost savings that would be practical for use in developing countries.

Selection of a dialysis modality for persons with end-stage renal disease (ESRD) has important lifestyle and occupational implications. The factors affecting modality choice remain unclear, resulting in a low rate of peritoneal dialysis (PD) in the United States compared with other countries. A national survey of 271 US nephrologists was conducted from June 1997 to June 1998 to assess the relative importance of nonclinical and clinical factors related to dialysis modality selection for patients with ESRD. Hypothetical patient scenarios were randomly assigned to nephrologists to determine their recommendation for dialytic therapy based on patient demographic, clinical, and social factors. US nephrologists were more likely to recommend PD for men with ESRD compared with women (39% versus 33%; P: < 0.05; adjusted odds ratio, 1.44; 95% confidence interval, 1.15 to 1.80), as well as for patients with good compliance (adjusted odds ratio, 11.80; 95% confidence interval, 9.29 to 15.01), weight less than 200 lb (adjusted odds ratio, 2.3; 95% confidence interval, 1.8 to 2.9), residual renal function (adjusted odds ratio, 2.14; 95% confidence interval, 1.71 to 2.70), absence of diabetes (adjusted odds ratio, 2.0; 95% confidence interval, 1.6 to 2.5), and living with family (adjusted odds ratio, 1.7; 95% confidence interval, 1.4 to 2.1). Nephrologists in practice for 11 or more years were less likely to recommend PD. The association of male sex with PD therapy suggests a potential bias or sensitivity to women's perception of body image. Race was not associated with PD recommendations after controlling for other demographic and clinical characteristics. Because the incident US ESRD population is increasingly characterized by factors associated with not selecting PD (diabetes, obesity, malnourishment, living alone, and substance abuse problems), our results suggest that PD use may decrease over time.

Projections of the incidence and prevalence of disease are important for public health planning. This paper describes methods for projecting the incidence and prevalence of a chronic disease in ageing populations. The approach uses age-specific disease incidence rates together with assumptions about survival to reconstruct disease prevalence. The methods can be used to evaluate the potential impact of public health interventions that may prevent disease or prolong survival. We used the methods to project the future prevalence of Alzheimer's disease in the United States. We found that the prevalence of Alzheimer's disease will nearly quadruple over the next 50 years. Although projections of the absolute prevalence are sensitive to assumptions about the age-specific incidence rates of disease, the proportionate growth is relatively insensitive. The increase in prevalence results from the ageing of the U.S. population. In order to perform the calculations, we have assembled U.S. Census population projections and U.S. mortality rates into computer software that is available from the authors at www.jhsph.edu/Departments/Biostats/software.h tml., (Copyright 2000 John Wiley & Sons, Ltd.)

Objective: To estimate age-specific incidence rates of AD in the Baltimore Longitudinal Study of Aging (BLSA)., Background: The BLSA is a volunteer cohort of normal subjects followed longitudinally with biennial evaluations at the Gerontology Research Center of the National Institute on Aging., Methods: Subjects are 1236 participants (802 men, 434 women) in the BLSA with longitudinal follow-up between January 1985 and May 1998. The average length of follow-up was 7.5 years, with participants evaluated every 2 years by physical, neurologic, and neuropsychological examinations. Using Diagnostic and Statistical Manual of Mental Disorders, 3rd ed., revised and National Institute of Neurological and Communicative Disorders and Stroke-Alzheimer's Disease and Related Disorders Association criteria, the authors diagnosed dementia and AD., Results: The authors diagnosed 155 cases of dementia, of which 114 (74%) were AD. Incidence rates of AD increased with age from an estimated 0.08% per year (95% CI 0.00 to 0.43) in the 60 to 65 age group to an estimated 6.48% per year (95% CI 5.01 to 8.38) in the 85+ age group for men and women combined. The doubling time of incidence rates was estimated to be approximately 4.4 years and the median time of conversion from mild cognitive impairment to diagnosis of AD was estimated to be 4.4 years. There was a trend for women to have higher incidence rates than men and for fewer years of education to be associated with higher incidence rates; however, these effects were not significant., Conclusion: Incidence rates for AD in the BLSA are consistent with published rates in other studies. The longitudinally followed subjects of the BLSA offer a unique opportunity to prospectively investigate the antecedents of AD.

The testing of pooled samples of biological specimens for the purpose of estimating disease prevalence may be more cost effective than testing individual samples, particularly if the prevalence of disease is low. Multistage pooling studies involve testing pools and then sequentially subdividing and testing the positive pools. A simple estimator of disease prevalence and its variance are derived for general multistage pooling studies and are shown to be natural generalizations of Thompson's (1962) original estimators for single-stage pooling studies. The reduction in variance associated with each additional stage is calibrated. The results are extended to estimating disease incidence rates. The methods are used to estimate HIV incidence rates from a prevalence study of early HIV infection using a PCR assay for HIV RNA.

Background: Current analytic methods applied to multidimensional health-related quality of life (HRQOL) data do not borrow strength across analyses and do not produce summary estimates of effect., Objectives: To compare a random effects modelling approach for the analysis of multidimensional HRQOL data to the following: (1) separate analyses for each dimension; (2) O'Brien's global test procedure; and (3) multivariate analysis of variance (MANOVA)., Research Design: Randomized clinical trial comparing 3 treatments (Trimethoprim-Sulfamethoxazole [TS], Dapsone-Trimethoprim [DT], and Clindamycin-Primaquine [CP] for Pneumocystis carinii pneumonia [PCP])., Subjects: Patients with PCP enrolled in AIDS Clinical Trials Group Protocol 108., Measures: A 33-item battery assessing 7 dimensions of HRQOL: physical functioning, pain, energy, general health perceptions, disability, pulmonary symptoms, and constitutional symptoms., Results: Analyses focused on changes in score from baseline to Day 7 (n = 145). Separate analyses for each dimension suggested a trend favoring CP versus TS, but using a Bonferroni correction no differences were statistically significant. O'Brien's global procedure for a test of no-treatment effect versus superiority of one treatment yielded P = 0.07. MANOVA did not reveal significant differences among treatment groups. A random effects model using fixed treatment and dimension effects and separate random effects for each person showed a significant overall treatment effect (P = 0.02); changes in scores for CP averaged 10 points greater than for TS., Conclusions: Random-effects models provide a flexible class of models for analyzing multidimensional quality of life data and estimating treatment effects because they borrow strength across dimensions.

Community intervention trials involve randomization of communities to either an intervention or control arm. The objective of this paper is to evaluate person-time methods of analysis of paired community intervention trials when the number of community pairs is small. We consider several test procedures and evaluate their performance by simulation. Naive methods that ignore intracluster correlation, such as standard Mantel-Haenszel type statistics, can be misleading. The performance of the paired t-test depends on the distribution of the random community effects. Permutation tests perform well for the ranges of situations considered. However, there can be considerable loss of power with permutation methods compared to standard Mantel-Haenszel methods if in fact there is no intracluster correlation when the number of pairs is small. We consider methods to account for individual level covariates. Motivation for this work came from recent randomized community intervention trials in Africa to prevent transmission of the human immunodeficiency virus (HIV).

Objectives: The goal of this study was to project the future prevalence and incidence of Alzheimer's disease in the United States and the potential impact of interventions to delay disease onset., Methods: The numbers of individuals in the United States with Alzheimer's disease and the numbers of newly diagnosed cases that can be expected over the next 50 years were estimated from a model that used age-specific incidence rates summarized from several epidemiological studies, US mortality rates, and US Bureau of the Census projections., Results: in 1997, the prevalence of Alzheimer's disease in the United States was 2.32 million (range: 1.09 to 4.58 million); of these individuals, 68% were female. It is projected that the prevalence will nearly quadruple in the next 50 years, by which time approximately 1 in 45 Americans will be afflicted with the disease. Currently, the annual number of new incident cases in 360,000. If interventions could delay onset of the disease by 2 years, after 50 years there would be nearly 2 million fewer cases than projected; if onset could be delayed by 1 year, there would be nearly 800,000 fewer prevalent cases., Conclusions: As the US population ages, Alzheimer's disease will become an enormous public health problem. interventions that could delay disease onset even modestly would have a major public health impact.

Multidimensional longitudinal data result when researchers measure an outcome through time that is quantified by many different response variables. These response variables are often defined on different numerical scales. The objective of this paper is to present a method to summarize and estimate an overall treatment effect from this type of longitudinal data. A regression model is proposed that assumes the treatment effect can be parameterized as an acceleration or deceleration of the time scale of each response variable's trajectory. Generalized estimating equations are used to estimate the model parameters. Cognitive and functional ability data from Alzheimer's disease patients and quality of life data from an AIDS clinical trial are used to illustrate the model.

Objective: To describe the design and first-round survey results of a trial of intensive sexually transmitted disease (STD) control to reduce HIV-1 incidence., Study Design: Randomized, controlled, community-based trial in Rakai District, Uganda., Methods: In this ongoing study, 56 communities were grouped into 10 clusters designed to encompass social/sexual networks; clusters within blocks were randomly assigned to the intervention or control arm. Every 10 months, all consenting resident adults aged 15-59 years are visited in the home for interview and sample collection (serological sample, urine, and, in the case of women, self-administered vaginal swabs). Sera are tested for HIV-1, syphilis, gonorrhea, chlamydia, trichomonas and bacterial vaginosis. Following interview, all consenting adults are offered directly observed, single oral dose treatment (STD treatment in the intervention arm, anthelminthic and iron-folate in the control arm). Treatment is administered irrespective of symptoms or laboratory testing (mass treatment strategy). Both arms receive identical health education, condom and serological counseling services., Results: In the first home visit round, the study enrolled 5834 intervention and 5784 control arm subjects. Compliance with interview, sample collection and treatment was high in both arms (over 90%). Study arm populations were comparable with respect to sociodemographic and behavioral characteristics, and baseline HIV and STD rates. The latter were high: 16.9% of all subjects were HIV-positive, 10.0% had syphilis, and 23.8% of women had trichomonas and 50.9% had bacterial vaginosis., Conclusions: Testing the effects of STD control on AIDS prevention is feasible in this Ugandan setting.

While the worldwide AIDS epidemic continues to expand, directly measured incidence data are difficult to obtain. Methods to reliably estimate human immunodeficiency virus type 1 (HIV-1) incidence from more easily available data are particularly relevant in those parts of the world where prevalence is rising in heterosexually exposed populations. The authors set out to estimate HIV-1 incidence in a population of heterosexual sexually transmitted disease clinic attendees in Trinidad who had a known high prevalence of HIV-1 subtype B. Over the period 1987-1995, HIV-1 incidence estimates from serial cross-sectional studies of HIV-1 prevalence, passive follow-up of clinic recidivists, modeling of early markers of HIV-1 infection (p24 antigen screening), and a cohort study of seronegative genital ulcer disease cases were compared. Measuring incidence density in the genital ulcer disease cases directly gave the highest estimate, 6.9% per annum. Screening for the detection of early HIV-1 markers yielded an incidence of 5.0% per annum, while estimating incidence from serial cross-sectional prevalence data and clinic recidivists gave estimates of 3.5% and 4.5% per annum, respectively. These results were found to be internally consistent. Indirect estimates of incidence based on prevalence data can give accurate surrogates of true incidence. Within limitations, even crude measures of incidence are robust enough for health planning and evaluation purposes. For planning vaccine efficacy trials, consistent conservative estimates may be used to evaluate populations before targeting them for cohort studies.

Objectives: Despite concern about access to liver transplantation, there has been no nationally based analysis of patients waiting for cadaveric liver transplant. Using data from the United Network for Organ Sharing Organ Procurement and Transplantation Network database waiting and recipient lists, we examined the influence of medical and non-medical factors on the length of time patients waited before transplant and whether they survived the wait., Methods: The authors analyzed 7,422 entries to the waiting list from October 1, 1990 to December 31, 1992. Using Cox Proportional Hazard models, time to transplant was modelled by gender, nationality and ethnicity, age, blood type, medical status (critically ill versus non-critical), transplant number (first versus retransplant), United Network for Organ Sharing region of the country, and three measures of local demand and supply of organs. The risk of dying before being allocated an organ was compared with receiving an organ using multiple logistic regression models., Results: In addition to differences by medical status, blood type, geographic region, and organ supply and demand, it was found that women, Hispanic-Americans, Asian-Americans, and children waited longer for transplant, whereas foreign nationals and repeat transplant patients waited fewer days. The risk of dying before transplant was greater for critically ill and repeat transplant patients, as well as for women, older patients, Asian-Americans, and African-Americans. Children were less likely to die, as were patients from certain blood groups and geographic regions., Conclusions: Results confirm known patterns of waiting list experience for liver transplant patients, but also identify factors previously unrecognized as influencing waiting time and outcome. Potential explanatory factors and areas for further inquiry are discussed.

Objective: To determine whether enrollment in the Baltimore Needle Exchange Program (NEP) was associated with short-term reduction in risky injection practices., Methods: Demographic information was collected on NEP participants upon enrollment. A systematic sample of enrollees was interviewed at program entry, 2 weeks, and 6 months later on recent drug-related behaviors. Comparisons were performed using paired t-tests., Results: Among 221 NEP participants who completed baseline, 2-week and 6-month follow-up visits, significant reductions (p < .01) were reported in using a previously used syringe (21.6%, 11.0%, 7.8%, respectively), lending one's used syringe to a friend (26.7%, 18.4%, 12%, respectively), and several indirect sharing activities. Reductions were reported in the mean number of injections per syringe and the mean number of injections per day (p < .001)., Conclusions: These results show rapid and mostly large reductions in a variety of risky injection drug use behaviors. Study findings are consistent with earlier reports showing an association between behavioral risk reduction and participation in a needle exchange program.

The hypothesis that quiescent CD4+ T lymphocytes carrying proviral DNA provide a reservoir for human immunodeficiency virus-type 1 (HIV-1) in patients on highly active antiretroviral therapy (HAART) was examined. In a study of 22 patients successfully treated with HAART for up to 30 months, replication-competent virus was routinely recovered from resting CD4+ T lymphocytes. The frequency of resting CD4+ T cells harboring latent HIV-1 was low, 0.2 to 16.4 per 10(6) cells, and, in cross-sectional analysis, did not decrease with increasing time on therapy. The recovered viruses generally did not show mutations associated with resistance to the relevant antiretroviral drugs. This reservoir of nonevolving latent virus in resting CD4+ T cells should be considered in deciding whether to terminate treatment in patients who respond to HAART.

This paper concerns the regression analysis of discrete time survival data for heterogeneous populations by means of frailty models. We express the survival time for each individual as a sequence of binary variables that indicate if the individual survived at each time point. The main result is that the likelihood for these indicators can be factored into contributions that involve the conditional survival probabilities integrated over the frailty distribution of the risk set (population-averaged). We then model these population-averaged conditional probabilities as a function of covariates. The result justifies the practice of treating the failure indicators as independent Bernoulli trials and fitting binary regression models for the conditional failure probabilities at each time point. However, we must interpret the regression coefficients as population-averaged rather than subject-specific parameters. We apply the method to the Framingham Heart Study on risk factors for cardiovascular disease.

Objectives: Although genital ulcer disease (GUD) has been associated with human immunodeficiency virus (HIV) infection in a number of studies, definitions of genital ulceration have varied. The authors hypothesized that the association of GUD with prevalent HIV infection may vary according to the definition of GUD that is used., Methods: As part of a prospective cohort study, 863 patients were interviewed and examined who presented to a sexually transmitted disease (STD) clinic for new symptom evaluation and who agreed to HIV testing to determine demographic and behavioral risk associated with prevalent HIV infection. To determine the association between GUD and prevalent HIV, the following definitions of GUD were used: observed ulcers, history of syphilis, serologic evidence of syphilis, observed culture-proven genital herpes, and serologic evidence of herpes simplex virus type II (HSV-2) infection., Results: Of 481 men and 382 women enrolled, prevalent HIV infection was detected in 12.5% and 5.2%, respectively. In multivariate analyses controlling for known HIV risk behaviors, prevalent HIV infection was associated with observed GUD (odds ratio [OR] = 2.0, 95% confidence intervals (CI) = 1.0-3.9), a history of syphilis (OR = 6.0, CI = 2.8-12.7), and serologic evidence of syphilis (OR = 3.7, CI = 1.9-7.0), but not with serologic evidence of HSV-2 (OR = 1.2, CI = 0.7-2.1), nor with observed HSV-2 culture-positive genital ulcerations (OR = 1.0, CI = 0.4-4.2). Factors contributing to different strengths of association between HIV infection and a history of syphilis or serologic evidence of syphilis included the presence of underdiagnosed syphilis infection in people with reactive serologic tests and the absence of serologic reactivity in people with a positive history., Conclusions: Although GUD is strongly associated with prevalent HIV, the strength of the association depends on the definition of GUD used. For accurate evaluation of people at risk for HIV, clinicians and researchers should use multiple definitions of GUD.

Previous reports have suggested that estrogen replacement therapy (ERT) in women may exert a protective effect on their risk of developing Alzheimer's disease (AD). We investigated this relationship in the Baltimore Longitudinal Study of Aging (BLSA), a prospective multidisciplinary study of normal aging conducted by the National Institute on Aging. The sample consisted of 472 post- or perimenopausal women followed for up to 16 years in the BLSA. We documented ERT prospectively at each BLSA visit, and we categorized women who had used oral or transdermal estrogens at anytime as ERT users. We used Cox proportional hazards models with time-dependent covariates to estimate the relative risk of developing AD after ERT as compared with women who had not used estrogen replacement. Approximately 45% of the women in the cohort had used ERT, and we diagnosed 34 incident cases of AD (NINCDS/ADRDA criteria) during follow-up, including nine estrogen users. After adjusting for education, the relative risk for AD in ERT users as compared with nonusers was 0.46 (95% CI, 0.209-0.997), indicating a reduced risk of AD for women who had reported the use of estrogen. Our data did not show an effect for duration of ERT usage. Our finding offers additional support for a protective influence of estrogen in AD. Randomized clinical trials are necessary to confirm this association, which could have significant public health impact.

The capacity of HIV-1 to establish latent infection of CD4+ T cells may allow viral persistence despite immune responses and antiretroviral therapy. Measurements of infectious virus and viral RNA in plasma and of infectious virus, viral DNA and viral messenger RNA species in infected cells all suggest that HIV-1 replication continues throughout the course of infection. Uncertainty remains over what fraction of CD4+ T cells are infected and whether there are latent reservoirs for the virus. We show here that during the asymptomatic phase of infection there is an extremely low total body load of latently infected resting CD4+ T cells with replication-competent integrated provirus (<10(7) cells). The most prevalent form of HIV-1 DNA in resting and activated CD4+ T cells is a full-length, linear, unintegrated form that is not replication competent. The infection progresses even though at any given time in the lymphoid tissues integrated HIV-1 DNA is present in only a minute fraction of the susceptible populations, including resting and activated CD4+ T cells and macrophages.

Background: Stroke occurs after coronary artery bypass grafting with an incidence ranging between 0.8% and 5.2%. To identify factors associated with stroke, we prospectively examined a study cohort and tested findings in an independent validation sample., Methods: The study cohort comprised 456 patients undergoing coronary artery bypass grafting only, and the validation sample comprised 1,298 patients. Stroke was detected postoperatively by the study team and confirmed by neurologic consultation and computed tomographic scanning., Results: Five factors taken together were correlated with stroke: previous stroke, presence of carotid bruit, history of hypertension, increasing age, and history of diabetes mellitus. The only significant intraoperative factor was cardiopulmonary bypass time. Probabilities were calculated, and patients were placed into low, medium, and high stroke-risk groups. In the validation sample, this model was able to rank the majority of patients with stroke into the high-risk group., Conclusions: These five factors taken together can identify the risk of stroke in patients having coronary artery bypass grafting. Recognition of the high-risk group will aid studies on the mechanism and prevention of stroke by modification of surgical procedures or pharmacologic intervention.

Statistical thinking has made significant contributions to our understanding of epidemics. Examples where statistics has played an important role in the Acquired Immunodeficiency Syndrome (AIDS) epidemic include estimating the number of individuals infected with the human immunodeficiency virus, estimating the incubation period of the disease, studying the etiology of the disease, and monitoring and forecasting the course of the epidemic. Some parallels with other epidemics in history are drawn. The AIDS epidemic has also raised important questions about the design of clinical studies and whether classical approaches are sufficiently flexible to provide timely answers to therapeutic questions in a growing epidemic. In a public crisis, there is a sense of urgency and data may be collected with unusual sampling schemes and inherent biases. Attention needs to be paid as much to sampling variation as to systematic sources of bias. Accurate disease surveillance data and methods for analyzing such data are crucial for detecting and monitoring future epidemics. There will almost certainly be new epidemics in the future, either of old diseases resurfacing or of new diseases, and statistical reasoning will continue to play a significant role in addressing the challenges of these public health crises.

Relatively little attention has been given to the proper design and statistical analysis of prevention and therapeutic trials of Alzheimer disease (AD). A number of important methodological issues arise in studies of AD that do not typically arise in studies of other chronic diseases. These issues include: the definition of the primary study endpoint that is simple to measure and easy to interpret; proper design of trials that take into account that the patient population is very old; and the proper statistical analysis that addresses the high drop out rate and the multiple ways of measuring disease severity. Addressing these issues will speed the progress of the evaluation of promising compounds to prevent and treat AD.

Measuring the incidence of the human immunodeficiency virus (HIV) is of vital importance but can be difficult and costly. We compared HIV-1 incidence measured directly from prospective cohort studies with rates derived from a method using the prevalence of HIV-1 antibody-negative, p24-antigen-positive individuals. Male and female commercial sex workers (CSWs) were enrolled and followed in separate cohort studies in northern Thailand between 1989 and 1994, and HIV incidence was measured by prospective follow-up of individuals seronegative at baseline. In 1991-1992 cross-sectional serosurveys were done among male and female CSWs in the same region; all HIV-1 antibody-negative subjects in these surveys were tested for p24 antigenemia. HIV incidence was estimated using the prevalence of p24 antigen and a model based on the mean duration of p24 antigenemia before HIV antibody detection. The cohort studies showed high initial incidence rates-23.8/100 person-years (PY) among female CSWs and 11.9/100 PY among male CSWs-but poor compliance with prospective follow-up. Subjects lost to follow-up appeared to be at greater risk of HIV seroconversion than those retained. The p24 antigen method estimate among female CSWs, 25.4%/year, was quite similar to the initial incidence rate found in the cohort. The estimate by the p24 antigen method was higher, 19.9%/year, among male CSWs than that measured prospectively. In populations with high rates of HIV transmission and in whom long-term follow-up is incomplete, estimates of incidence using p24 antigen prevalence among antibody-negative subjects can give useful and economical estimates of HIV incidence and allow for estimates of whether the incidence rates are similar in subjects successfully followed and those lost to follow-up.

Recent attempts to identify predictors of rate of decline in Alzheimer's disease (AD) have been extremely variable in choice of outcome variables, predictor variables tested, timing of assessments, and statistical approaches. In this study, a random effects regression model was applied to seek predictors of decline on the Mini-Mental State Exam in 132 patients with probable AD reassessed every 6 months for up to 7.5 years. Potential predictor variables at baseline were of three types: patients characteristics, clinical variables, and cognitive performances. The final multivariate analysis indicated that the following characteristics predicted more rapid cognitive decline: more education, history of dementia in a first degree relative, non-right handedness, better performances of Boston Naming Test, Gollin Incomplete Figures Test, and Benton Visual Retention Test-Delay, and worse performances on Responsive Naming Test, WAIS-R Block Design, and Benton Visual Retention Test-Copy.

Background: The risk of Alzheimer's disease (AD) appears to increase, and the age at onset to decrease, with the number of epsilon 4 alleles. If this relationship is due to increased rate of pathophysiological change, the presence of epsilon 4 would be expected to influence progression of disease, predicting a more rapid decline with increasing number of epsilon 4 alleles., Objective: To determine if the frequency of the epsilon 4 allele of the apolipoprotein E (ApoE) gene affects the rate of clinical progression in AD., Setting: Alzheimer's Disease Research Center., Subjects: One hundred one subjects meeting criteria of the National Institute of Neurological Disorders and Stroke for probable AD or of the Consortium to Establish a Registry for Alzheimer's Disease (CERAD) for definite AD; 78 of these subjects met the additional criterion of having a Mini-Mental State Examination score of at least 10 for analysis of rate of decline., Measurements: The subjects' characteristics and neuropsychological battery, including the Mini-Mental state Examination, Spatial Delayed Recognition Span, Boston Naming Test, Category Fluency Test, and the Physical Capacity Subscale of the Psychogeriatric Dependency Rating Scale., Design: The subjects were followed up longitudinally for approximately one decade. Medical histories were taken and physical and neurologic examinations and neuropsychological testing were performed every 6 months. Three and a half years of data were available for most tests and 5.5 for the Psychogeriatric Dependency Rating Scale; thereafter, patients were no longer testable. A general linear model analysis of variance was used to assess the influence of ApoE on demographic characteristics and baseline performances on neuropsychological measures. A random-effects regression model was used to predict change over time associated with presence of epsilon 4 on clinical and cognitive measures., Results: The age at onset was greatest for the epsilon 4-heterozygous subjects and least for the epsilon 4-negative subjects. The heterozygous subjects declined more rapidly on the Mini-Mental State Examination and the Category Fluency Test than the subjects without the epsilon 4 allele or with epsilon homozygosity. The homozygous subjects declined faster on only one subscale: the Physical Capacity subscale of the Psychogeriatric Dependency Rating Scale. Covarying for age at onset did not affect the results., Conclusions: The ApoE genotype does not strongly influence the rate of decline in AD, implying that epsilon 4 might predispose to the development of the disease without accelerating its pathogenesis or progression. The effects of epsilon 4 on both age at onset and rate of decline need to be further investigated.

Both hepatitis B virus (HBV) and human immunodeficiency virus (HIV) type 1 seroconversions have been considered as outcome measures to evaluate the effectiveness of needle exchange programs. To assess the relationships between incident HBV and HIV infections among injecting drug users (IDUs), seroconversions were prospectively studied among a cohort of 240 HBV- and HIV- seronegative IDUs. The incidence of HBV seroconversion declined from 24.41/100 person-years in 1988 to 0 seroconversions in 1992. In contrast, HIV seroconversion rates varied little from the overall rate of 3.29/100 person-years. HBV seroconversion predicted subsequent HIV seroconversion among male IDUs (relative incidence [RI]= 4.23) but not among female IDUs (RI=0.86). Because of different transmission dynamics, HBV seroconversion probably has limited utility as a surrogate outcome measure for incident HIV. However, HBV seroconversion itself is an appropriate and important outcome measure for evaluation of prevention programs among IDUs.

Because of limitations of the univariate frailty model in analysis of multivariate survival data, a bivariate frailty model is introduced for the analysis of bivariate survival data. This provides tremendous flexibility especially in allowing negative associations between subjects within the same cluster. The approach involves incorporating into the model two possibly correlated frailties for each cluster. The bivariate lognormal distribution is used as the frailty distribution. The model is then generalized to multivariate survival data with two distinguished groups and also to alternating process data. A modified EM algorithm is developed with no requirement of specification of the baseline hazards. The estimators are generalized maximum likelihood estimators with subject-specific interpretation. The model is applied to a mental health study on evaluation of health policy effects for inpatient psychiatric care.

Human immunodeficiency virus (HIV) incidence rates in India were estimated using a new method that accounts for follow-up bias. Follow-up bias arises in epidemiologic cohort studies when the incidence rate among individuals who do and do not return for follow-up are different. The new method combines data on the prevalence of p24 antigenemia among all those initially screened together with the longitudinal follow-up data on the subset of patients who returned for follow-up. Using these methods, the current HIV incidence rate among patients attending sexually transmitted disease clinics in Pune, India, was 18.6% per year. It was found that follow-up bias can cause significant underestimation in HIV incidence rates, perhaps by as much as 60%. These incidence estimates, together with other HIV seroprevalence studies, suggest the HIV epidemic in India is growing rapidly.

In-person cognitive evaluations can be costly and labor intensive in geographically widespread populations. Reliable telephone instruments that screen for cognitive status would greatly facilitate epidemiologic and other longitudinal studies. We evaluated the reliability of the Blessed Information-Memory-Concentration (IMC) test when administered by telephone. Eighty-four subjects with a wide range of cognitive abilities were administered the Blessed IMC twice over a 3-week interval. Forty-nine of the subjects were administered the test both by telephone and in-person, and 35 of the subjects were tested twice by telephone. Spearman's rank correlation was used to compare scores of the different administrations (.96; P < .001) and to examine test-retest reliability (.96; P < .001). The Blessed Telephone IMC (TIMC) test exhibits excellent reliability both when compared to in-person administration as well as in test-retest results. The Blessed TIMC appears to be a practical instrument for population and longitudinal studies when in-person assessment is not feasible.

Objective: To investigate potential methodological reasons for the differences in published Alzheimer's disease (AD) prevalence rates., Background: Studies reporting prevalence rates of AD have been published worldwide. These rates differ considerably, but may greatly reflect methodological differences., Methods: All studies published between 1984 and 1993 that reported age-specific AD rates and sample sizes were included. Logistic regression identified variables that contribute to the variation in rates. Estimates of extrabinomial variation were also calculated., Results: Studies characterized by the following features yielded significantly higher rates: inclusion of mild cases, use of laboratory studies, ascertainment of a sample rather than the total population, inclusion of both urban and rural populations, non-use of computerized tomography (CT) scans, non-use of the Hachinski Ischemic Score, and no adjustment for false negatives. The odds of having AD increased 18% for every year of age. The variation in the age-specific prevalence rates of AD was approximately 15 times that expected by sampling variation. However, approximately 76% of this excess variation in rates could be accounted for by methodological differences., Conclusions: After accounting for age, much of the variability in prevalence rates of AD in the published literature may be explained by differences in methodology. Some unexplained variation in prevalence rates, however, still remains.

Passive surveillance disease data involve a registry of individuals who are at risk of disease and who are not under active follow up. The most serious limitations with such data are incomplete ascertainment of cases of disease and little or no follow-up information on patient vital status. This paper considers whether it is possible to estimate disease risk from such data and, if not, what additional information is required. In general, relative risks based on passive surveillance data will be biased even under the assumption that the probability of disease reporting and the hazard of death from other causes are the same for all individuals in the registry. However if the disease is rare, this bias is negligible. Methods are developed for estimating absolute disease incidence rates by combining passive surveillance data with a cohort study. Analytic approaches are proposed for the situations when death rates from all other causes are known and also unknown, and it is found that there is little loss in efficiency even if death rates are not known. There are considerable gains in efficiency for estimating absolute disease incidence rates by supplementing a cohort study with passive surveillance registry data compared to using the cohort study alone, especially if the exposure is rare and the cohort study is small relative to the size of the registry. Intuitively, the cohort data provides information about absolute rates of disease, while the passive surveillance data provides information about relative risks. The methods are applied to a registry of patients with an artificial heart valve that is at risk of breaking.

In this analysis we made use of a large data base of individuals insured by large American corporations to estimate the impact of managed care provision on hospital care for depression. Data on 6,348 individuals hospitalized for depression were examined to assess the effect of managed care techniques on the cost per episode and the likelihood of rehospitalization. Preadmission certification programs were found to lead to significant long- and short-run savings for payers.

Backcalculation is a methodology to reconstruct the past human immunodeficiency virus (HIV) infection rates from the AIDS incidence data and incubation distribution by deconvolution. Smoothing has proved important in backcalculation, and a key question is how to choose the amount of smoothing. This paper proposes an empirical Bayes approach in which the smoothing parameter is estimated from the data. We introduce a family of priors that reflect the notion of closeness of neighboring infection rates. The variance parameter in the prior family plays the role of the smoothing parameter and is estimated by a method similar to the residual maximum likelihood in linear random effects model through an efficient EM (expectation/maximization) algorithm. A number of penalized likelihood functions that have been used in backcalculation have an empirical Bayes formulation. A bootstrap confidence interval for the infection rates is proposed. The methodology is illustrated with United States AIDS incidence data.