Longitudinal depression trajectories: the persistence of depression symptomology and their genetic & environmental underpinnings from mid- to later-life
The work contained within this thesis primarily focused on estimating depression symptom score trajectories in mid- to late-life. There has been little consistency in the literature concerning such phenotypic trajectories, with reported differences by age, sex, birth cohort, and country; and different modelling approaches have often indicated different depression trajectories (i.e., reported depression scores following increasing, decreasing, or even U-shaped trajectories). Increased attention has been given to examining the genetic and environmental influences on depression symptom score development. In this literature, mostly based on twin samples, results have consistently indicated that the majority of variance in depression symptom score trajectories is non-shared environmental, and ~30% genetic, but specifics have varied. However, there have been few such well-powered studies in middle- to older-aged adults. Using a large population representative sample that included 16,881 twins aged 25- 102 years drawn from the Interplay of Genes and Environment across Multiple Studies (IGEMS) consortium representing nine studies from three countries (Denmark, Sweden, and the US), I attempted in this thesis to reconcile the discrepancies concerning phenotypic depression trajectories, and also focused on extending our understanding of the roles of genetic and environmental influences on individual differences in depression symptom scores in older adults. I hoped that, given the power of my large representative sample and long timespan of assessments and ages from multiple studies, I could pool the nine samples to offer generalizable depression trajectories. This thesis is presented in four parts. Part I provides a general introduction to the topic, drawing from the wider literature that included both cross-sectional and longitudinal analysis examining depression scores, as well as an outline of the methods and statistical approaches I used throughout the thesis. Part II concerns itself with applying two different statistical modelling approaches to estimate phenotypic depression trajectories. I hoped to produce convergent indications between the two approaches. Here, I estimated person- and age-based depression trajectories, and examined the associations with age, sex, birth year, and country. The results differed by modelling approach. There were variations in estimated trajectories by country too, and these differences themselves differed by approach, as well as whether the data were pooled or modelled individually by country. The only consistent findings across the two approaches were that quadratic models were generally best-fitting; females had consistently higher depression scores than males; and Swedes had higher scores than Danes and Americans. Thus, instead of furthering understanding of the development of depressive symptoms in middle- to older-aged adults, the contributions of Part II are more methodologically focused. However, this is at least as important given that I highlight potential pitfalls of pooling samples with different recruitment procedures, assessment intervals, birth cohorts, age ranges, and relative volumes of data at the various ages; and that most longitudinal studies only apply one estimation method but too often interpret results as if they had applied both. Part III focuses on estimating the degrees to which the magnitudes of genetic and environmental influences on depression symptom scores remained stable and/or changed. To do so, I exploited the genetically informative twin data available in the IGEMS sample. Although my original aim was to examine longitudinal change, based on the estimation difficulties and varying sample sizes described in Part II, I instead used Cholesky bivariate models to test the degree to which there were common genetic and environmental influences on depression symptom scores at two of the most consistently available assessment times, and their correlations over time. I fit models separately by sex and literature-relevant age groups. I also did country-specific analyses. My initial results suggested that genetic influences were generally stable across assessments, but that new environmental influences were present at each assessment – although the degree and extent of this varied by sex and age group. Although I was unable to capture true longitudinal change, Part III provides convergent evidence in line with the wider literature, where indications suggested that the variation in depression symptom scores resulted predominantly from unique non-shared environmental influences across assessments. Part IV concludes with a summary of the observations contextualised within the wider depression and ageing literature, acknowledging limitations of the current work were acknowledged. After evaluating the insights offered into the development of depression across the mid- to late-lifespan, as well as the role of genetic and environmental influences, suggestions for future research are outlined.