Analysis of cohort effects in mixed longitudinal data sets

Area of Science:

• Biostatistics
• Human Growth and Development
• Epidemiology

Background:

• Mixed longitudinal designs offer efficiency in studying growth and developmental processes.
• These designs involve studying multiple birth cohorts over shorter periods and linking their growth curves.
• Validating this approach requires addressing potential cohort effects (secular trends).

Purpose of the Study:

• To investigate methods for testing and correcting cohort effects in mixed longitudinal studies.
• To illustrate potential problems using height and weight data from the Nijmegen Growth Study.
• To demonstrate the application of these methods with cleft lip and palate patient data and the National Dutch Growth Study 1980.

Main Methods:

• Utilizing mixed longitudinal designs to collect data from multiple birth cohorts.
• Analyzing height and weight data to identify cohort effects at specific ages (e.g., 9.25 years).
• Comparing raw data with adjusted values estimated at target ages to assess cohort differences.

Main Results:

• Raw height and weight data, particularly for girls, showed significant cohort effects at 9.25 years in the Nijmegen Growth Study.
• After adjusting data to estimate values at the target age, cohort differences became non-significant.
• Illustrative examples confirmed the presence and potential correction of cohort effects in other datasets.

Conclusions:

• Mixed longitudinal designs are efficient but require careful consideration of cohort effects.
• Adjusting for cohort effects by estimating values at target ages can yield more accurate growth curves.
• The methods discussed are applicable to various growth studies, including clinical and national datasets.