Selection of working correlation structure and best model in GEE analyses of longitudinal data

The Generalized Estimating Equations (GEE) method is one of the most commonly used statistical methods for the analysis of longitudinal data in epidemiological studies. A working correlation structure for the repeated measures of the outcome variable of a subject needs to be specified by this method. However, statistical criteria for selecting the best correlation structure and the best subset of explanatory variables in GEE are only available recently because the GEE method is developed on the basis of quasi-likelihood theory. Maximum likelihood based model selection methods, such as the widely used Akaike Information Criterion (AIC), are not applicable to GEE directly. Pan (2001) proposed a selection method called QIC which can be used to select the best correlation structure and the best subset of explanatory variables. Based on the QIC method, we developed a computing program to calculate the QIC value for a range of different distributions, link functions and correlation structures. This program was written in Stata software. In this article, we introduce this program and demonstrate how to use it to select the most parsimonious model in GEE analyses of longitudinal data through several representative examples.

Leptin in depressed women : Cross-sectional and longitudinal data from an epidemiologic study

Background There is conflicting evidence regarding levels of leptin in depression. In this study we aimed to investigate the relationship between serum leptin level and depression in a community sample of women using both cross-sectional and longitudinal data.

Methods From among 510 women aged 20–78 yr, 83 were identified with a lifetime history of major depressive disorder or dysthymia, ascertained using the Structured Clinical Interview for DSM-IV-TR Research Version, Non-patient edition (SCID-I/NP). Serum leptin levels were measured by radioimmunoassay. Medication use and lifestyle were self-reported and body mass index (BMI) determined from measures of height and weight.

Results Using multiple linear regression, serum leptin levels were greater among women with a lifetime history of depression compared to women without any history of depression, independent of BMI. Adjusted geometric mean values of serum leptin were 16.37 (95%CI 14.70–18.23) ng/mL for depressed and 14.46 (95%CI 13.79–15.16) ng/mL for non-depressed women (P = 0.039). The hazard ratio (HR) for a de novo depressive disorder over five years increased 2.56-fold for each standard deviation increase in log-transformed serum leptin among non-smokers and this was not explained by differences in BMI, medications or other lifestyle factors (HR = 2.56, 95%CI 1.52-4.30). No association was observed for smokers.

Limitations There is potential for unrecognised confounding, recall bias and transient changes in body composition.

Conclusion Women with a lifetime history of depression have elevated levels of serum leptin, and elevated serum leptin predicts subsequent development of a depressive disorder.

Modeling gene-environment interaction in longitudinal data : risk for neuroticism due to interaction between maternal care and the Dopamine 4 Receptor gene (DRD4)

The purpose of this study was to investigate risk for neuroticism due to the joint action of low maternal care and compromised mesocorticolimbic ‘reward’ system function linked to a variable number tandem repeat (VNTR) in the dopamine 4 receptor gene (DRD4). Data were drawn from the Victorian Adolescent Health Cohort Study, a longitudinal study of the health and well-being of 2,000 young Australians followed from adolescence to young adulthood across 8 waves from 14- to 28-years. Genetic risk was defined by carriage of at least one copy of the 7-repeat allele or derivative alleles 5, 6, and 8 (labeled 7R+). Neuroticism was assessed in adolescence and young adulthood. We observed an approximately fourfold increase in the odds of reporting neurotic symptoms in carriers of the 7R+ disposition who reported low maternal care compared with non-carriers who reported high maternal care. The percentage of risk attributable to mechanisms in which both factors played a role was 35%. Findings are discussed in terms of implications for prevention.

Contrasting longitudinal and cross-sectional relationships between insulin resistance and percentage of body fat, fitness, and physical activity in children—the LOOK study

Background: Knowledge of individual changes in insulin resistance (IR) and longitudinal relationships of IR with lifestyle-associated factors are of important practical significance, but little longitudinal data exist in asymptomatic children. We aimed to determine (a) changes in the homeostatic model of insulin resistance (HOMA-IR) over a 2-yr period and (b) comparisons of longitudinal and cross-sectional relationships between HOMA-IR and lifestyle-related risk factors.

Methods: Our subjects, 241 boys and 257 girls, were assessed at age 8.1 yr (SD 0.35) and again 2 yr later for fasting blood glucose and insulin, dual X-ray absorptiometry-assessed percentage of body fat (%BF), pedometer-assessed physical activity (PA), and cardio-respiratory fitness (CRF) by multistage running test.

Results: HOMA-IR was initially 9% greater in girls than boys and 27% greater 2 yr later. There was no evidence of longitudinal relationships between HOMA-IR and %BF in boys or girls, despite significant cross-sectional relationships (p < 0.001). In boys, there was evidence of a longitudinal relationship between HOMA-IR and both PA (p < 0.001) and CRF (p = 0.05). In girls, we found a cross-sectional relationship between HOMA-IR and CRF (p < 0.001).

Conclusions: HOMA-IR increases between 8 and 10 yr of age and to a greater extent in girls. Longitudinal, unlike cross-sectional, relationships do not support the premise that body fat has any impact on HOMA-IR during this period or that PA or CRF changes affect HOMA-IR in girls. These data draw attention to difficulties in interpreting observational studies in young children.

Change in active travel and changes in recreational and total physical activity in adults : longitudinal findings from the iConnect study

Background : To better understand the health benefits of promoting active travel, it is important to understand the relationship between a change in active travel and changes in recreational and total physical activity.

Methods : These analyses, carried out in April 2012, use longitudinal data from 1628 adult respondents (mean age 54 years; 47% male) in the UK-based iConnect study. Travel and recreational physical activity were measured using detailed seven-day recall instruments. Adjusted linear regression models were fitted with change in active travel defined as ‘decreased’ (<−15 min/week), ‘maintained’ (±15 min/week) or ‘increased’ (>15 min/week) as the primary exposure variable and changes in (a) recreational and (b) total physical activity (min/week) as the primary outcome variables.

Results : Active travel increased in 32% (n=529), was maintained in 33% (n=534) and decreased in 35% (n=565) of respondents. Recreational physical activity decreased in all groups but this decrease was not greater in those whose active travel increased. Conversely, changes in active travel were associated with commensurate changes in total physical activity. Compared with those whose active travel remained unchanged, total physical activity decreased by 176.9 min/week in those whose active travel had decreased (adjusted regression coefficient −154.9, 95% CI −195.3 to −114.5) and was 112.2 min/week greater among those whose active travel had increased (adjusted regression coefficient 135.1, 95% CI 94.3 to 175.9).

Conclusion : An increase in active travel was associated with a commensurate increase in total physical activity and not a decrease in recreational physical activity.

A longitudinal analysis of self-prescribed complementary and alternative medicine use by a nationally representative sample of 19,783 Australian women, 2006-2010

BACKGROUND: Complementary and alternative medicine (CAM) use is increasingly popular amongst general populations around the world with women constituting substantial CAM users. However, self-prescribed CAM use does raise potential safety concerns and so it is important to identify those risk factors associated with self-prescribed CAM use. METHODS: Data was obtained from the Australian Longitudinal Study on Women's Health (ALSWH). Longitudinal data analyses were conducted on questionnaire data from the 1973-78 cohort (n=9,145) and the 1946-51 cohort (n=10,638), collected over the period 2006-2010. RESULTS: In the 1973-78 cohort, use of self-prescribed CAM was 73.2% in 2006 and 75.3% in 2009. For the 1946-51 cohort, use of self-prescribed CAM was 73.9% in 2007 and 74.7% in 2010. There were similar levels of use of individual self-prescribed CAM, with the exception that the use of herbal medicine was much higher among the 1946-51 cohort (20% vs. 27%). There was a substantial increase over three years in the use of vitamins/minerals in both cohorts (21% and 19%, respectively). In contrast, there was a considerable decline over three years in use of aromatherapy oils in both cohorts (34% and 28%, respectively). CONCLUSION: Self-prescribed CAM use is popular amongst women in Australia and it is important that conventional practitioners providing women's health care be cognizant of such use amongst their patients. In order to ensure effective practice, there is a need for further research to explore women's decision-making and experiences around self-prescribed CAM use.

Parenting style and behavior as longitudinal predictors of adolescent alcohol use

OBJECTIVE: Adolescent alcohol use is a serious problem in Australia and other nations. Longitudinal data on family predictors are valuable to guide parental education efforts. The present study tested Baumrind's proposal that parenting styles are direct predictors of adolescent alcohol use.

METHOD: Latent class modeling was used to investigate adolescent perceptions of parenting styles and multivariate regression to examine their predictive effect on the development of adolescent alcohol use. The data set comprised 2,081 secondary school students (55.9% female) from metropolitan Melbourne, Australia, who completed three waves of annual longitudinal data starting in 2004.

RESULTS: Baumrind's parenting styles were significant predictors in unadjusted analyses, but these effects were not maintained in multivariate models that also included parenting behavior dimensions.

CONCLUSIONS: Family influences on the development of adolescent alcohol use appear to operate more directly through specific family management behaviors rather than through more global parenting styles.

Growth of highly versus moderately trained competitive female artistic gymnasts

Purpose: To investigate whether growth was adversely affected in 137 young competitive female artistic gymnasts involved in different training volumes.

Methods: This was a 2-yr prospective cohort study in which height, sitting height, leg length, weight, skinfolds, and pubertal status were measured in competitive advanced (20–27 h·wk-1) and intermediate (7.5–22 h·wk-1) training level female gymnasts every 6–12 months. Biological parameters of the adolescent growth curve were estimated using the Preece–Baines growth model. Growth rates were estimated for both groups from the mixed-longitudinal data.

Results: Estimated ages at peak height velocity (PHV) (13–13.5 yr) and mean PHV (6.2–6.4 cm·yr-1) for the advanced- and intermediate-level gymnasts suggest that these gymnasts were later maturing and experienced a blunting of the growth spurt relative to reference values for U.S. youth. Comparison of growth velocities by pubertal status revealed that height velocity was lower in the advanced- versus the intermediate-level peripubertal gymnasts, which was due to a significant reduction in sitting height velocity (2.3 vs 3.1 cm·yr-1, P. < 0.05). No marked acceleration in height or sitting height velocity was detected in the advanced-level gymnasts from pre- to peripuberty. Inspection of individual growth rates revealed that over 35% of the pre- and peripubertal gymnasts experienced growth faltering (height velocity less than 4.5 cm·yr-1) during follow-up.

Conclusion: Advanced–and intermediate–training level competitive female gymnasts tend to exhibit an adolescent growth spurt that is similar in timing and tempo to short, normal, slowly maturing girls, but the high frequency of growth faltering suggests that training may alter the tempo of growth and maturation in some, but not all, female gymnasts.

Older Australians : involvement in civil society in the 1990s

As Australia’s population continues to age, questions about how older individuals use their time holds increasing interest and significance for scholars and policy makers. As individuals near the end of their paid working life, and family roles and responsibilities diminish, the type of activities that will fill this time void have important implications for the health and wellbeing of older Australians and for the strength of civil society. In Australia, there have been sustained moves at all levels of government to encourage the more active engagement in community services of this group of citizens, given the size and significant amount of human capital of this cohort. However, international research suggests that this enthusiasm has not translated into increased volunteer activity for seniors, and that older citizens tend to spend their expanding discretionary time pursuing leisure activities, such as watching television or listening to the radio (Robinson & Godbey 1997; Wilson & Musick 1997; Thoits & Hewitt 2001). This study builds on a broader interest in how people choose to utilise time across the life course and how the experience of ageing shapes such decisions. This aim of this paper is twofold – first, to investigate how older Australians allocated their time in the 1990s, and how these time use patterns changed over a 5-year period, using nationally representative, longitudinal data from two waves of the Australian Time Use Survey. Second, the time use characteristics of those individuals who devote more time to social participation activities are examined, to investigate trends in volunteering across age cohorts, with a focus on those above the age of fifty.

The relationship between quality of life and perceived body weight an dieting history in Dutch men and women

OBJECTIVES: (1) To study the relationship between quality of life (QoL) and measured and perceived weight and dieting history in Dutch men and women; (2) to assess the effect of weight loss over a 5 y period on QoL.

DESIGN: A cross-sectional study, in a sub-sample longitudinal over 5 y.

SUBJECTS: A total of 2155 men and 2446 women, aged 20-59 and recruited from the general population from three towns in The Netherlands.

MEASUREMENTS: Body weight, height, self-administered questionnaire including questions concerning demographic variables and weight loss practices as part of the Dutch Monitoring project on Risk Factors for Chronic Disease (MORGEN). The Rand-36 questionnaire was used as the QoL measure.

RESULTS: In men, measured overweight (body mass index, BMI>25 kg=m2) was not associated with any dimension of QoL after adjustment for age, educational level and perceived overweight. Perceived overweight was related to reduced scores for general health and vitality. This relationship was independent of measured obesity. A history of repeated weight loss was associated with reduced scores for role functioning due to both physical and emotional problems. In women, measured overweight was significantly associated with lower scores for five out of eight QoL dimensions and perceived overweight with three: general health, vitality and physical functioning. A history of frequent weight loss was related to significantly reduced scores in six dimensions. However, only with history of frequent weight loss, and uniquely in women, was there a significant reduction in
scores on mental health and limited emotional role functioning. Measured and perceived overweight and frequent weight loss were all related to reduced scores for physical functioning. Longitudinal data indicate that in older women weight gain of 10% body weight or more was associated with a significant deterioration in QoL.

CONCLUSIONS: When looking at measures of QoL in relation to overweight it is important to separate the effects of perception of weight status and history of weight loss. We observed that the latter two factors were associated with reduced scores on several dimensions of QoL, particularly in women. These associations were observed to be independent of body weight. International Journal of Obesity (2001) 25, 1386 – 1392

Mothering in the freelance film industry: case studies in the postproduction film industry in Melbourne, Australia

This paper examines the dynamic relationship between work and life through the experiences of three mothers seeking to realise careers in the freelance film industry in Melbourne, Australia. The women were part of a larger sample of postproduction workers who were surveyed on their responses to technological change in the postproduction freeelance film industry but were out of the industry having children during the first wave of technological change. The paper examines qualitative data from these women and later longitudinal data in order to understand the limits of equal opportunity in an uneven playing field.