Measuring cognitive change in subjects with prodromal Alzheimer's disease.

OBJECTIVE: To investigate the sensitivity of a large set of neuropsychological tests to detect cognitive changes due to prodromal Alzheimer's disease(AD); to compare their metrological properties in order to select a restricted number of these tests for the longitudinal follow-up of subjects with prodromal AD. PARTICIPANTS: 212 patients with mild cognitive impairment were tested at baseline by a standardised neuropsychological battery, which included: the Free and Cued Selective Reminding test (FCSRT), the Benton Visual Retention test, the Deno100, verbal fluency, a serial digit learning test, the double task of Baddeley, the Wechsler Adult Intelligence Scale (WAIS) similarities, the Trail-Making Test and the WAIS digit symbol test. Patients were monitored every 6 months for up to 3 years in order to identify those who converted to AD (retrospectively classified as prodromal AD). Statistical analyses were performed using a nonlinear multivariate mixed model involving a latent process. This model assumes that the psychometric tests are nonlinear transformations of a common latent cognitive process, and it captures the metrological properties of tests. RESULTS: 57 patients converted to AD. The most sensitive tests in the detection of cognitive changes due to prodromal AD were the FCSRT, the semantic verbal fluency and the Deno100. Some tests exhibited a higher sensitivity to cognitive changes for subjects with high levels of cognition, such as the free recall, delayed free recall scores of the FCSRT and the semantic verbal fluency, whereas others showed a higher sensitivity at low levels of cognition, such as the total recall score of the FCSRT. CONCLUSIONS: Tests used for the follow-up of prodromal AD subjects should be chosen among those that actually decline in this stage of the disease and should be selected according to the subject's initial scores.

A simple score to predict survival with dementia in the general population.

BACKGROUND/AIMS: This study was designed to develop a practical risk score for predicting 5-year survival after the diagnosis of dementia. METHODS: Using the Paquid Study (prospective, population-based, long-term cohort study), we created a prognosis score with incident cases of dementia and validated it in another prospective, population-based, long-term cohort study, the Three City Study. - RESULTS: Among the 3,777 subjects enrolled in the Paquid Study, 454 incident cases of dementia were included in this study. After a 5-year follow-up period, 319 (70.3%) were deceased. The score was constructed from three independent prognostic variables (gender, age at diagnosis and number of ADL restricted). The discriminant ability of the score was good with a c index of 0.754. Sensitivity was 64.7% and specificity 76.3%. In the validation cohort, the discriminant ability of the prognostic score with c statistics was 0.700. Sensitivity was 26.3% and specificity 95.4%. CONCLUSIONS: The prognostic factors selected in the predictive model are easily assessable, so this simple score could provide helpful information for the management of dementia, particularly to identify patients with duration of the disease greater than 5 years.

Joint model with latent state for longitudinal and multistate data.

In many chronic diseases, the patient's health status is followed up by quantitative markers. The evolution is often characterized by a 2-phase degradation process, that is, a normal phase followed by a pathological degradation phase preceding the disease diagnosis. We propose a joint multistate model with latent state for the joint modeling of repeated measures of a quantitative marker, time-to-illness and time-to-death. Using data from the PAQUID cohort on cognitive aging, we jointly studied cognitive decline, dementia risk, and death risk. We estimated the mean evolution of cognitive scores given age at dementia for subjects alive and demented, the mean evolution of cognitive scores for subjects alive and nondemented, in addition to age at acceleration of cognitive decline and duration of the pre-dementia phase.

Ginkgo Biloba Extract Consumption and Long-Term Occurrence of Death and Dementia.

OBJECTIVES: To study the benefit of Ginkgo Biloba Extract (GBe) consumption on the long term risk of dementia and death in elderly people. DESIGN: The Paquid study is a population-based cohort with regular follow-up screenings up to twenty-two years and systematic detection of incident cases of dementia. Statistical analysis was conducted with an illness-death model dealing with interval censoring of dementia and competing risk of death. SETTING: The sample was randomly selected from electoral rolls in two administrative areas of southwestern France in 1988-1989. PARTICIPANTS: 3,777 subjects aged 65 years or older at baseline who were living at home. MEASUREMENT: Participants were visited at home by a trained psychologist at baseline in 1988/1989, and then again approximately every two years. Drug consumption for the treatment of cognitive or neurosensory impairment was collected at baseline. Participants were classified as GBe consumers, other drug (OD) consumers and untreated controls (UC) for this motive. RESULTS: After adjustment for sociodemographic factors and cognitive measures at baseline the risk for dementia was not significantly different in GBe consumers and UC (Hazard Ratio (HR)=1.21, 95% Confidence Interval (95% CI)=0.95-1.55, p=0.42) and it was of the same magnitude but significantly increased in the OD group versus UC (HR=1.25, 95% CI=1.06-1.46, p=0.004). With the same adjustment, the risk of dying in non-demented subjects was reduced in GBe consumers versus UC (HR=0.67, 95% CI=0.49-0.93, p=0.02) while it was the same as the reference group in OD consumers. The mean lifetimes without dementia was of 11.2 years in the UC group (95% CI=10.9-11.5), 11.1 years in the GBe group (10.2-11.9) and 9.1 years for the OD group (8.7-9.6). CONCLUSION: GBe consumers have a lower risk of dying before dementia and a longer lifetime without dementia than participants taking other drugs for the same indication.

Clinical progression of HIV-1 infection according to the viral response during the first year of antiretroviral treatment. Groupe d'Epidémiologie du SIDA en Aquitaine (GECSA).

OBJECTIVE: To compare HIV-disease progression according to changes of plasma HIV RNA observed in the year following initiation of a new antiretroviral treatment. DESIGN: Prospective cohort treated with two nucleoside analogues or a triple combination including a protease inhibitor. METHODS: A Cox model was used to estimate the effect of viral response during the first year after initiation of treatment on the subsequent occurrence of new AIDS-defining events or death. Viral response was fitted either as HIV RNA reduction during the initial 4-12 months of treatment or reduction during the first month. RESULTS: Among 773 patients (47% with triple drug combination) followed for a median period of 27 months, 62 patients experienced a clinical event. Poor viral responders (at least two measurements > 3.7 log10 copies/ml during 4-12 months of treatment) had a higher risk of disease progression than good responders (RNA < 2.7 log10 copies/ml) after adjustment [hazard ratio (HR), 2.24; 95% confidence interval (CI), 1.1 7-4.29]. Intermediate responders (2.7 < or = RNA < or = 3.7 log10 copies/ml) had a risk of progression comparable with that of good responders (HR, 1.43; 95% CI, 0.64-3.22). A large initial viral reduction was also a protective factor for clinical progression (HR, 0.51 for 1 log10 copies/ml increase of the reduction; 95% CI, 0.31-0.85) and was associated with the viral response during the subsequent 4-12 month period. No patient with a reduction < 0.5 log10 copies/ml in the first month was classified as a good responder in the subsequent 4-12 month period (P < 0.01). CONCLUSIONS: A sustained HIV RNA > 3.7 log10 copies/ml should suggest a prompt change of treatment. When the reduction in HIV RNA is < 0.5 log10 after 1 month of treatment, this action should be anticipated. A sustained HIV RNA level between 2.7 and 3.7 log10 copies/ml may permit the deferral of a change of drug regimen according to the patient's history and therapeutic options.

A cognitive screening battery for dementia in the elderly.

The objective of this study is to propose a screening instrument for dementia based on a reduced number of neuropsychological tests. The sample consists in the pooled data of the five follow-up visits of the Paquid cohort study on cerebral aging: the estimation sample included 2792 subjects (8830 observations) and the validation sample included 985 subjects (2643 observations). Among scores significantly associated with dementia, we retained only those that increased the specificity of the model for a sensitivity of one. Seven neuropsychological tests and the MMSE subscores were considered. The most discriminant combination of tests included the MMSE and the subscores "orientation to time" and "recall three objects," the Benton Visual Retention Test, and Isaacs' Set Test of verbal fluency. The specificity of this screening instrument was 0. 77 for a sensitivity of 1.

[Joint modeling of quantitative longitudinal data and censored survival time]. / Modélisation conjointe de données longitudinales quantitatives et de délais censurés.

BACKGROUND: In epidemiology, we are often interested in the association between the evolution of a quantitative variable and the onset of an event. The aim of this paper is to present a joint model for the analysis of Gaussian repeated data and survival time. Such models allow, for example, to perform survival analysis when a time-dependent explanatory variable is measured intermittently, or to study the evolution of a quantitative marker conditionally to an event. METHODS: They are constructed by combining a mixed model for repeated Gaussian variables and a survival model which can be parametric or semi-parametric (Cox model). RESULTS: We discuss the hypotheses underlying the different joint models proposed in the literature and the necessary assumptions for maximum likelihood estimation. The interest of these methods is illustrated with a study of the natural history of dementia in a cohort of elderly persons.

[Analysis of longitudinal Gaussian data with missing data on the response variable]. / Analyse de données longitudinales gaussiennes comportant des données manquantes sur la variable à expliquer.

BACKGROUND: Using an application and a simulation study we show the bias induced by missing data in the outcome in longitudinal studies and discuss suitable statistical methods according to the type of missing responses when the variable under study is gaussian. METHOD: The model used for the analysis of gaussian longitudinal data is the mixed effects linear model. When the probability of response does not depend on the missing values of the outcome and on the parameters of the linear model, missing data are ignorable, and parameters of the mixed effects linear model may be estimated by the maximum likelihood method with classical softwares. When the missing data are non ignorable, several methods have been proposed. We describe the method proposed by Diggle and Kenward (1994) (DK method) for which a software is available. This model consists in the combination of a linear mixed effects model for the outcome variable and a logistic model for the probability of response which depends on the outcome variable. RESULTS: A simulation study shows the efficacy of this method and its limits when the data are not normal. In this case, estimators obtained by the DK approach may be more biased than estimators obtained under the hypothesis of ignorable missing data even if the data are non ignorable. Data of the Paquid cohort about the evolution of the scores to a neuropsychological test among elderly subjects show the bias of a naive analysis using all available data. Although missing responses are not ignorable in this study, estimates of the linear mixed effects model are not very different using the DK approach and the hypothesis of ignorable missing data. CONCLUSION: Statistical methods for longitudinal data including non ignorable missing responses are sensitive to hypotheses difficult to verify. Thus, it will be better in practical applications to perform an analysis under the hypothesis of ignorable missing responses and compare the results obtained with several approaches for non ignorable missing data. However, such a strategy requires development of new softwares.

[Longitudinal study of cognitive aging in non-demented elderly subjects]. / Etude longitudinale du vieillissement cognitif de sujets âgés non déments.

BACKGROUND: The aim of this study is to describe and compare the course of the score of two psychometric tests during a five-year period in a cohort of non-demented elderly subjects. We studied a timed test of simple logical reasoning (the Digit Symbol Substitution Test of Wechsler) and a test of short-term visual memory (the Benton Visual Retention Test). Then we compared the course of these two tests with the course of the Mini-Mental State Examination (MMSE) which evaluates the general cognitive functioning and which has been studied previously. METHODS: The sample consisted of 2792 subjects aged 65 years and older living at home at the beginning of the study (the Paquid cohort in Gironde). Subjects were interviewed at home by a psychologist at the baseline visit and one year, 3 years and 5 years after the baseline visit. All subjects who were diagnosed as demented at any of the four examinations were excluded from this study (218 subjects). The statistical analysis was performed using a mixed effects linear model and an additional analysis was performed to study sensitivity of the results to hypotheses regarding missing data mechanism. RESULTS: Among the three tests studied, only the score of the Digit Symbol Substitution Test declined over the five years (by -1.15 points for subjects aged 65 years to -3.4 points for subjects aged 85 years). We observed an improvement of the score between the first two visits which is sharper for the test of Benton and the MMSE than for the Digit Symbol Substitution Test and which reflects a stress effect at the baseline visit or a learning effect at the second visit. The difference between the cross-sectional measure of age effect and the longitudinal measure of time effect suggests either a cohort effect or a practice effect which would persist during the five years. CONCLUSIONS: The cognitive change in a non-demented elderly population is different according to the psychometric test used. Among the three tests studied, only the score of the Digit Symbol Substitution Test which has a a speed component declined over the five years.

Determinants of response to first HAART regimen in antiretroviral-naïve patients with an estimated time since HIV seroconversion.

OBJECTIVE: To study the determinants of immunological and virological response to highly active antiretroviral therapy (HAART) in naïve patients, adjusting for time since HIV-1 seroconversion. DESIGN: Data from HIV-cohort studies where dates of seroconversion have been reliably estimated. Methods In previously untreated patients, short- and long-term marker responses from HAART initiation (three or more antiretroviral drugs) to the end of follow-up or any treatment modification were considered using mixed effects models accounting for undetectable HIV viral load and informative dropout. RESULTS: In total, 943 patients were treated with a first HAART regimen for a median of 29 months. In adjusted analyses, compared with a reference group of homosexual men without AIDS initiating treatment 4 years after seroconversion, injecting drug users (IDUs) were treated at similar CD4 and HIV RNA levels but had poorer short-term virological response (2.54 vs 2.13 log(10) HIV-1 RNA copies/mL at 1.5 months, P=0.03) and poorer long-term immunological response (522 vs 631 cells/microL at 24 months, P<0.0001). Although individuals with AIDS at HAART initiation had lower CD4 counts (206 vs 382 cells/microL, P<0.0001), their immunological responses were similar to those of individuals without AIDS. Similarly, individuals further from seroconversion started HAART at lower CD4 counts (e.g. 311 vs 382 cells/microL at vs before 9 years from seroconversion, P<0.0001), but had similar CD4 responses. However, they experienced poorer long-term virological response (0.67 log(10) copies/mL/year smaller decline, P<0.0001) compared to those treated before 9 years from seroconversion. CONCLUSION: Taking into account the time elapsed since seroconversion, this study suggests that careful choices of initial treatment should be made and intensive follow-up carried out in high-risk subgroups such as IDUs who have poorer responses.

A latent process model for joint modeling of events and marker.

The paper formulates joint modeling of a counting process and a sequence of longitudinal measurements, governed by a common latent stochastic process. The latent process is modeled as a function of explanatory variables and a Brownian motion process. The conditional likelihood given values of the latent process at the measurement times, has been drawn using Brownian bridge properties; then integrating over all possible values of the latent process at the measurement times leads to the desired joint likelihood. An estimation procedure using joint likelihood and a numerical optimization is described. The method is applied to the study of cognitive decline and Alzheimer's disease.

Analysis of left-censored longitudinal data with application to viral load in HIV infection.

The classical model for the analysis of progression of markers in HIV-infected patients is the mixed effects linear model. However, longitudinal studies of viral load are complicated by left censoring of the measures due to a lower quantification limit. We propose a full likelihood approach to estimate parameters from the linear mixed effects model for left-censored Gaussian data. For each subject, the contribution to the likelihood is the product of the density for the vector of the completely observed outcome and of the conditional distribution function of the vector of the censored outcome, given the observed outcomes. Values of the distribution function were computed by numerical integration. The maximization is performed by a combination of the Simplex algorithm and the Marquardt algorithm. Subject-specific deviations and random effects are estimated by modified empirical Bayes replacing censored measures by their conditional expectations given the data. A simulation study showed that the proposed estimators are less biased than those obtained by imputing the quantification limit to censored data. Moreover, for models with complex covariance structures, they are less biased than Monte Carlo expectation maximization (MCEM) estimators developed by Hughes (1999) Mixed effects models with censored data with application to HIV RNA Levels. Biometrics 55, 625-629. The method was then applied to the data of the ALBI-ANRS 070 clinical trial for which HIV-1 RNA levels were measured with an ultrasensitive assay (quantification limit 50 copies/ml). Using the proposed method, estimates obtained with data artificially censored at 500 copies/ml were close to those obtained with the real data set.

Serum triglycerides, HIV infection, and highly active antiretroviral therapy, Aquitaine Cohort, France, 1996 to 1998. Groupe d'Epidémiologie Clinique du Sida en Aquitaine (GECSA).

The aim of this study was to identify factors associated with serum triglyceride (TG) evolution in HIV-1-infected patients when highly active antiretroviral treatment (HAART) with or without protease inhibitors (PI) was introduced. Among 3191 patients of the Aquitaine Cohort (multirisk, both genders, multiple treatment patterns) observed during 1996 through 1998, 1429 had at least two measurements of TG, viral load, and CD4 cell count. Median follow-up was 21 months (interquartile range [IQR], 11-26) and median number of TG measures was 6 (IQR, 3-10). Median TG at baseline was 1.32 mmol/L (IQR, 0.91-2.05) and increased significantly over time (+2.5% for 100 days; 95% confidence interval [CI], 1.9-3.1). Longitudinal analysis of variations of TG was performed using mixed models. In crude analysis, baseline TG was higher in men, in those aged over 36 years, and in homosexuals. The following time-dependent variables were associated with an increase of TG: body weight increasing to >65 kg, diagnosis of AIDS, CD4 cell count falling to <50 cells/mm3, viral load falling to <500 cp/ml, and introduction of nucleoside analogues and PIs. In multivariate analysis, age >36 years (change of +17% of the TG level; 95% CI, 11-24), homosexuals (+13%; 95% CI, 4-23), AIDS stage (+12%; 95% CI, 5-19), weight >65 kg (+7%; 95% CI, 2-12) and PI (+21%; 95% CI, 17-27) remained significant. Factors identified before the availability of PI remain important but HAART with PI is a new major contributing factor to increased TG levels.

Longitudinal analysis of the effect of apolipoprotein E epsilon4 and education on cognitive performance in elderly subjects: the PAQUID study.

BACKGROUND: The apolipoprotein E (apoE) epsilon4 allele has been shown to be a risk factor for dementia, but it is not clear to what extent apoE affects overall cognitive function in non-demented elderly subjects, or how this risk may be modified by gene-environment interactions. OBJECTIVE: To examine changes in cognitive function in elderly people as a function of the apoE epsilon4 phenotype. METHODS: A community based prospective cohort study of 600 non-demented subjects aged over 65 years living in Gironde (France) was analysed to evaluate change over time (seven years) in scores on the mini-mental state examination (MMSE). RESULTS: Age at cohort inception was negatively associated with cognitive performance for both epsilon4 carriers and non-carriers (p < 0.001). The evolution of MMSE scores differed as a function of age: scores remained stable among younger subjects but decreased over time in older subjects. The epsilon4 allele was shown to be significantly associated with lower cognitive performance at baseline (p = 0.02). The course of cognitive performance during the follow up was the same for both epsilon4 carriers and non-carriers. Lower educational level was associated with lower cognitive performance at baseline (p < 0.001) and the effect of an epsilon4 allele on cognitive performance disappeared after adjustment for education. When incident cases of dementia were excluded, the results were unchanged except for the course of the MMSE scores, which now remained stable over time in the older subjects. CONCLUSIONS: apoE epsilon4 carriers show decreased MMSE scores compared with epsilon4 non-carriers, but the effect of apoE on cognition disappears after adjustment for education. Non-demented elderly people maintain a stable cognitive performance regardless of their apoE phenotype.

Prevalence of asthma and mean levels of air pollution: results from the French PAARC survey. Pollution Atomosphérique et Affections Respiratoires Chroniques.

Among the possible explanations for the recent increase in the prevalence of asthma in several countries, air pollution is one of the foremost public health concerns. Data from the "Pollution Atmosphérique et Affections Respiratoires Chroniques" (PAARC) survey collected in 24 areas of seven French towns during 1974-1976 were reanalysed to assess the relationship between the prevalence of asthma and the following air pollutants: sulphur dioxide (specific (SO2) and acidimetric methods), total suspended particles (TSP), black smoke (BS), nitrogen dioxide and nitric oxide. Correlation coefficients between annual mean levels of pollution and prevalence of asthma in the different areas were first calculated. Random-effects models were then estimated. Of the 20,310 adults aged 25-59 yrs, 1,291 (6.4%) were found to be asthmatics as well as 195 (6.1%) of the 3,193 children aged 5-9 yrs. A geographical correlation between asthma and annual mean level of SO2 (ranging 17-85 microg x m(-3)) was found (r=0.45, p=0.01) in adults. No relationship was found in children. After controlling for age, educational level, smoking, and geographical clustering with a multivariate random-effects model, the relationship remained significant in adults for SO2 (odds ratio for a 50 microg x m(-3) increase=1.24, confidence interval 1.08-1.44, p=0.0035). It also remained significant when taking into account only the people reporting their last asthma attack occurring after settling in the study area. These results are consistent with the known short-term effects of SO2 in asthma and demonstrate the necessity for further studies on delayed effects of air pollution in respiratory diseases.

Tests of geographical correlation with adjustment for explanatory variables: an application to dyspnoea in the elderly.

We propose a test of correlation of the residuals in generalized linear models which is a generalization of the spatial autocorrelation test based on Moran's I. It allows adjustment for sizes of geographical areas and for explanatory variables. A formula is given to compute the weights according to the alternative hypothesis. We compare inference using the distribution in the model and using the permutation distribution. A simulation study showed that the model-based test may be very conservative and this leads to a loss of power compared to the permutation test or to the model-based test with correction for estimated parameters. As this latter is intractable for very large samples when the model includes explanatory variables, we recommend the use of the permutation test. The permutation test is used to study geographical correlation of dyspnoea in the elderly.

A 5-year longitudinal study of the Mini-Mental State Examination in normal aging.

The Mini-Mental State Examination (MMSE) measures global cognitive performance and is often used as a screening test for dementia. This paper presents a 5-year longitudinal study of the MMSE score in a sample of 2,537 non-demented French community residents aged 65 years and older who were participants in the Paquid Study in 1988-1992. Subjects were evaluated at the baseline visit (T0) and 1 year (T1), 3 years (T3), and 5 years (T5) later. Analyses performed with a random effects linear model showed that the score rose between T0 and T1 (by 0.60 points for subjects aged 65 years at T0 to 0.83 points for subjects aged 85 years at T0), then it decreased very slightly between T1 and T5 (by 0.02 points for subjects aged 65 years to 0.57 points for subjects aged 85 years). The improvement during the first year, which was larger for less educated subjects, may be explained by the stress due to the test situation at T0 or by a learning effect at T1. The decline during the last 4 years was more pronounced for older and less well educated subjects. The cross-sectional measure of age effect was larger than the longitudinal measure of time effect. This difference may be explained by a cohort effect or by a practice effect induced by repetition of the test. The authors conclude that the MMSE score declines very slightly in non-demented subjects, thus suggesting that the cognitive processes involved are spared by the aging process. These results may have implications for dementia screening.

Silica and aluminum in drinking water and cognitive impairment in the elderly.

We studied the relation between silica and aluminum levels in drinking water and the risk of cognitive impairment using data from a population-based survey of 3,777 French subjects age 65 years and older. We also studied the effect of pH and the concentrations of calcium, magnesium, fluorine, zinc, copper, and iron. We used a mixed effects logistic regression adjusting for age, sex, educational level, and occupation of the subjects. We confirmed the inverse relation previously found between calcium level and cognitive impairment. We found no important association between cognitive impairment and fluorine, magnesium, iron, copper, or zinc. The association between cognitive impairment and aluminum depended on the pH and the concentration of silica: high levels of aluminum appeared to have a deleterious effect when the silica concentration was low, but there was a protective effect when the pH and the silica level were high. The threshold for an aluminum effect, however, was very low (3.5 micrograms per liter) and did not support the hypothesis of a deleterious effect for only high levels of aluminum.

Risk factors for fractures in the elderly.

We report the results of a 5-year prospective cohort study of risk factors for fractures, including drinking fluoridated water, in a cohort of 3,216 men and women aged 65 years and older. We studied risk factors for hip fracture and fractures at other locations separately. We found a higher risk of hip fractures for subjects exposed to fluorine concentrations over 0.11 mg per liter but without a dose-effect relation (odds ratio (OR) = 3.25 for a concentration of 0.11-0.25 mg per liter; OR = 2.43 for > or = 0.25 mg per liter]. For higher thresholds (0.7 and 1 mg per liter), however, the OR was less than 1. We found no association between fluorine and non-hip fractures. Non-hip fractures were associated with polymedication rather than with specific drug use, whereas fracture was associated with polymedication and use of anxiolytic and antidepressive drugs. Subjects drinking spirits every day were more likely to have hip fractures. Tobacco consumption increased the risk for non-hip fractures.