Design considerations for characterizing psychiatric trajectories across the lifespan: application to effects of APOE-ε4 on cerebral cortical thickness in Alzheimer's disease - PubMed (original) (raw)

Review

Design considerations for characterizing psychiatric trajectories across the lifespan: application to effects of APOE-ε4 on cerebral cortical thickness in Alzheimer's disease

Wesley K Thompson et al. Am J Psychiatry. 2011 Sep.

Abstract

Characterization of developmental trajectories across the lifespan is integral to understanding the prodromal course of many neuropsychiatric illnesses and the significant risk factors for disease onset or unfavorable outcomes. However, the standard experimental designs used in psychiatric research are not ideal for this purpose. The authors review the limitations of the most commonly employed designs in studies that make developmental or lifespan inferences in psychiatry: cross-sectional, single-cohort longitudinal, and unstructured multicohort longitudinal designs. Cross-sectional studies completely confound within- and between-subject sources of variation and hence rely on the presence of parallel trajectories and negligible sampling and age cohort differences for making valid developmental inferences. Delineating trajectories of within-individual change over substantial periods of time requires data covering long age spans that often cannot be covered using single-cohort longitudinal designs. Unstructured multicohort longitudinal designs are a commonly used alternative that can cover a longer age span in a shorter interval than necessary for a single-cohort design. However, the impact of cohort and sampling effects is often minimized or ignored in unstructured multicohort longitudinal designs. The authors propose that structured multicohort longitudinal designs are a particularly viable and underutilized class of designs in psychiatry that represents a significant improvement over cross-sectional designs and unstructured multicohort longitudinal designs for making developmental inferences while being more practical to implement than single-cohort longitudinal designs. As an example of this approach, the authors analyze changes in entorhinal cortex thickness in Alzheimer's disease in relation to APOE-ε4 genotype.

PubMed Disclaimer

Conflict of interest statement

All authors report no financial relationships with commercial interests.

Figures

FIGURE 1. Sources of Confounding in Between-Patient and Within-Patient Variationa

aIn the left panel, simulated data show linear trajectories, censored when they fall below 0.1. The heavy red line is the average of censored trajectories; the “healthy survivor” effect biases the average upward. In the right panel, simulated data show nonparallel trajectories. The heavy red line is the average of population trajectories; nonparallel trajectories cause the population average to be unrepresentative of any one individual trajectory.

FIGURE 2. Entorhinal Cortex Thickness by Age and Apolipoprotein Eε4 Statusa

aThe left panel plots baseline entorhinal cortex thickness against baseline age for 157 patients with Alzheimer’s disease enrolled in the Alzheimer’s Disease Neuroimaging Initiative. Red dots indicate patients with ε4– status (N=50), and blue dots indicate patients with ε4+ status (N=107). The heavy red line is fitted from a linear model for ε4– patients, and the heavy blue line is fitted from the same model for ε4+ patients. The right panel plots longitudinal entorhinal cortex thickness for the same patients, using data from baseline and all follow-up assessments. The heavy red line is fitted from a linear mixed-effects model for ε4– patients, and the heavy blue line is fitted from the same model for ε4+ subjects. Entorhinal cortex thickness is standardized to have a mean of zero and a standard deviation of 1 at baseline.

FIGURE 3. Longitudinal Entorhinal Cortex Thickness, by Age and Apolipoprotein E ε4 (APOE-ε4) Status, With Cohort Effectsa

aLongitudinal entorhinal cortex thickness measures for patients with Alzheimer’s disease with ε4– (N=50) and ε4+ (N=107) status. Heavy lines and line segments indicate model fitted from linear mixed-effects model including age cohort, change in age from baseline, and APOE-ε4 status. Line segments extending from the longer lines indicate departures of within-subject slopes (change in age from baseline) from age cohort slopes, for seven age cohorts (55, 60, 65, 70, 75, 80, and 85 years of age at baseline).

FIGURE 4. Comparison of Unstructured and Structured Age Cohortsa

aThe left panel shows a frequency histogram of baseline ages of 157 patients with Alzheimer’s disease from the Alzheimer’s Disease Neuroimaging Initiative (ADNI), an example of an unstructured multicohort longitudinal design. The right panel shows a frequency histogram for a hypothetical structured multicohort longitudinal design with the same number of subjects and range of baseline ages as ADNI.

FIGURE 5. Hypothetical Accelerated Longitudinal Designa

aThis hypothetical accelerated longitudinal design covers ages 55–90 years. The design results in 17 age cohorts with each cohort assessed four times over the course of 3 years. Resulting trajectories for each age cohort overlap at two time points.

FIGURE 6. Comparison of Cross-Sectional and Structured Multicohort Longitudinal Design Trajectory Estimatesa

aThe left panel shows nonlinear, nonparallel trajectories from 100 subjects ages 55–90 years, generated from simulated data. The heavy dark line is a “typical” population trajectory; the heavy red line is the population average of trajectories. The middle panel plots data sampled from trajectories using an accelerated longitudinal design with four yearly measurements. The heavy red line is the data average, and the purple segments are mean within-subject trajectory estimates for evenly spaced age cohorts (baseline ages=55, 57.5, 60, and so on). The right panel plots data from the same accelerated longitudinal design. The heavy red line is again the data average, the heavy black line is a “typical” trajectory, and the heavy green line is the estimated typical trajectory constructed by minimizing age cohort effects.

References

1. Insel TR. Translating scientific opportunity into public health impact: a strategic plan for research on mental illness. Arch Gen Psychiatry. 2009;66:128–133. -PubMed
1. Bell RQ. Convergence: an accelerated longitudinal approach. Child Dev. 1953;24:145–152. -PubMed
1. Schaie KW. A general model for the study of developmental problems. Psychol Bull. 1965;64:92–107. -PubMed
1. Baltes PB. Longitudinal and cross-sectional sequences in the study of age and generation effects. Hum Dev. 1968;11:145–171. -PubMed
1. Botwinick J. Neuropsychology of aging. In: Filskov SB, Boll TJ, editors. Handbook of Clinical Neuropsychology. New York: Wiley-Interscience; 1981. pp. 135–171.

Design considerations for characterizing psychiatric trajectories across the lifespan: application to effects of APOE-ε4 on cerebral cortical thickness in Alzheimer's disease - PubMed (original) (raw)