Proposal summaries
B2344 - Investigating the possible bidirectional association between smoking and caffeine using Mendelian randomization - 27/11/2014
Aims
To investigate the possible bidirectional association between smoking and caffeine consumption using Mendelian randomization.
Hypotheses
Coffee consumption and cigarette smoking are highly comorbid [1]. Observational studies suggest that coffee drinkers are more likely to be smokers than non-drinkers [1, 2], and also suggest a positive association between coffee consumption and cigarette consumption [3], and an inverse relationship between coffee consumption and smoking cessation [4].
However, the causal nature and direction of these effects are unknown and are difficult to infer from observational studies. Mendelian randomization methods minimise bias from confounding and remove the possibility of reverse causality [5].
Genetic variants for both smoking behaviour phenotypes and coffee consumption have been identified in genome-wide association studies [6, 7]. These can be used as instruments for measured coffee consumption and smoking behaviour in Mendelian randomization studies. Polygenic risk scores for these exposures can be generated from GWAS data to increase the power of Mendelian randomization analysis [8].
Polygenic risk scores for smoking behaviours and caffeine consumption will be generated from GWAS data for both the ALSPAC mothers and ALSPAC children. We will investigate the associations of genetic risk scores for smoking with caffeine consumption and the association of genetic risk scores for caffeine consumption with smoking. These analyses will be run in parallel with data from the Netherlands Twin Registry.
B2343 - The exploration of an association between previously reported orofacial cleft genes and lip trait phenotypes - 27/11/2014
The aim of this study it to explore a link between specific genes that have been reported to be associated with orofacial clefting and identified lip traits from 3D facial scans.
We hypothesise that SNP's associated with clefting, should have an effect on lip morphology. The exposure variable is the orofacial cleft gene carrier (see appendix for list of candidate genes) and the outcome variable is the characteristic lip trait (see categorisation table) - that may have increased/decreased occurrence of certain features. These features may in turn be considered risk factors in determining risk of orofacial clefting.
The GWAS for these lip traits has already been conducted (under B568) and so this proposal should just involve the look-up of the relevant variants in those results. Lavinia Paternoster (IEU) has access to these results. These look-ups and any subsequent analysis required will be carried out by Caryl Wilson-Nagrani as an IEU temporary visitor under the supervision of Lavinia Paternoster.
B2342 - Associations of Behaviour Disorders in Childhood and Psychotic-Like Symptoms in Adolescence - 27/11/2014
Associations of Behaviour Disorders in Childhood and Psychotic-Like Symptoms in Adolescence
The general aim of the project is to analyze the associations between childhood behaviour disorders and psychotic symptoms in adolescence. Many adults with psychotic symptoms and/or schizophreniform disorders received a psychiatric diagnosis in childhood. These diagnoses cover the full range of developmental, internalizing, externalizing, and behavioural disorders (Rubino et al., 2008). To date, it is not clear whether the presence of childhood mental symptoms and disorders as such represents a general risk factor for later psychopathology or whether there are specific associations indicated by stronger correlations for some disorders (research question 1) and how these associations are characterized (research question 2).
B2341 - Role of rare coding variants in speech and language disorder - 20/11/2014
AIM - to investigate the contribution of rare coding variants in selected candidate genes to speech and language disorders
BACKGROUND
Our lab investigates genetic contributions to speech and language disorders, primarily within a cohort of families in which at least one child is diagnosed with specific language impairment (The SLI Consortium (SLIC) cohort). Investigations in this cohort have highlighted specific genetic pathways and candidate genes that we believe might be linked to speech and language development. We would like to follow these genes up in more detail in a larger cohort. The effects we have identified are rare coding mutations. In order to follow these up, we therefore need access to a large cohort in which speech and language data and genetic sequencing are readily available.
B2339 - Investigation of the association between birth weight reproductive abnormalities and placental function - 20/11/2014
Background
Amongst disorders of sex development (DSD) cases, conditions associated with under virilisation of the presumed XY boys are the largest group of children. Our international group (I-DSD.org) recently reported in Pediatrics that this group of people will increase as more affected infants are raised as boys. XY DSD boys often have low birth wt and are also often small for gestational age (SGA), as confirmed by Hughes et al, Cambridge using the ALSPAC cohort. Altered exposure to androgens has been linked with DSDs, intra-uterine growth restriction (IUGR) and placental problems, possibly all involving the IGF signalling pathway. For example, exposure to exognous androgens in sheep advances placental differentiation and increases placental differentiation but this early compensation to maintain placental efficienc cannot be mainted long term leading to placental inefficiency and IUGR. Furthermore, altered steroid exposure in fetal rats through exposure to MPA, a synthetic progestogen, reduced external genitalia masculinisation in males and increased it in females, and reduced bodyweight and placental weight in both males and females.
Aims
We hypothesise that:
(1) Boys with DSDs (namely hypospadias and/or cryptorchidism) are more at risk of being SGA, possibly due to altered placental development and/or funtion.
(2) The SGA phenotype associated with DSDs, such as hypospadias, is different to the SGA phenotype without hypospadias.
(3) Boys with hypospadias may have been denied sufficient androgen exposure prenatally and/or experience an ongoing lack of androgens, which may have a negative impact on their future health and well being.
We therefore propose to analyse the ALSPAC data for variables such as birth weight, gestational age at birth, placenta weight, congenital abnormalities at birth such as hypospadias, anogenital distance in the child, any fertility treatment prior to pregnancy, as well as moderately long-term outcome that relates to height, cardiometabolic and body composition including skeletal health.
We wish to compare the following groups - hypospadias and SGA, SGA and no hypospadias, average GA and hypospadias and AGA and no hypospadias. We may also be interested in the same analysis for individuals with cryptorchidism
B2338 - Ascertaining direction of causality in associations between smoking cognition and schizophrenia - 20/11/2014
Aims:
Associations are consistently seen between cigarette use and psychosis, although direction of causation is not known. There is also a large body of literature showing cognitive impairment in patients with schizophrenia. There has been some suggestion that high rates of smoking in schizophrenia could be to alleviate cognitive impairment from both the disorder and anti-psychotic medication side-effects, as nicotine has been posited as a cognitive enhancer. We aim to investigate whether associations between cigarette use, cognition and psychotic experiences are causal, using Mendelian Randomisation.
Hypotheses:
As these are exploratory analyses, we do not have hypotheses as to what we will find in terms of causality.
Exposure variables:
Genetic instruments as proxy variables for cigarette use and for risk of schizophrenia.
Outcome variables:
Cognitive tasks at age 18, smoking at age 18, psychotic experiences at age 18 (all measured at clinic).
Confounding variables:
As this is a Mendelian randomisation design, the analyses should not be subject to the usual confounders that impact upon observational studies. However, gender and ethnicity will be accounted for in our analyses. Data on confounders has been requested to check for pleiotropy.
B2337 - The association between caffeine consumption during pregnancy and birth weight a Mendelian randomization study - 20/11/2014
Aim
To investigate if maternal caffeine intake during pregnancy causally influences offspring birth weight using a Mendelian randomization approach.
Hypotheses
A recent meta-analysis of observational studies found evidence that caffeine consumption during pregnancy is associated with increased risk of having a low birth weight baby [1]. However, it is not clear whether this association is causal or whether this may be due to confounding by other lifestyle factors. For example, coffee consumption is positively correlated with cigarette consumption, a known risk factor for low birth weight [2].
This problem of confounding can be minimised using Mendelian randomisation analyses, whereby genetic variants associated with caffeine are used as proxies for measured caffeine consumption. Genome-wide association studies have identified 8 independent genetic loci associated with coffee consumption [3, 4]. Some of these variants have been shown to associate specifically with caffeinated beverage intake in ALSPAC [5].
This project will investigate associations between maternal caffeine-related genetic variants and offspring birth weight using a Mendelian randomisation approach.
B2336 - Enhancing data collection in ALSPAC-G2 using novel methods A feasibility study - 20/11/2014
Background:
A high proportion of women start pregnancy overweight/obese. According to the developmental overnutrition hypothesis, this could cause offspring to be fatter throughout their lives, and therefore, this may perpetuate the obesity epidemic across generations.1, 2 Concerns about this hypothesis are influencing antenatal care. Two key changes to antentatal care which have occurred in relation to concerns about developmental overnutrition are:
1.Monitoring weight throughout pregnancy. This was previously done with the aim of identifying those at risk of SGA, but was stopped in the mid-1990s because of evidence of its poor predictive value.3 It is now being done with the intent of limiting gestational weight gain (GWG) in order to prevent women having infants who are LGA, since they may go on to be more adipose throughout their lives.4
2.The new International Association of Diabetes and Pregnancy Study Groups (IADPSG) criteria for defining gestational diabetes (GDM), which have been adopted by the WHO, aim to identify those at risk of LGA and greater offspring adiposity.5
However, whether the associations of maternal adiposity with offspring outcomes are causal is unknown, and if they are indeed causal, then the mechanisms are unclear.1, 2 Being able to accurately quantify maternal and fetal fat mass, glucose variability, diet, physical activity and a wide range of potential circulating nutrients in women during pregnancy could clarify the effects of maternal overweight/adiposity on her future health and that of her offspring. Having tools for obtaining these data that are acceptable, feasible, cost-effective and safe in pregnancy would support translation of research findings into clinical practice as appropriate.
B2335 - 10000 hours Children Childcare and Developmental Opportunity - 20/11/2014
Analysis of the ALSPAC data forms part of a larger grant application examining the genetic markers of phenotypic measures of emotional reactivity in early childcare settings. More specifically, the ALSPAC data will be used for the validation of a GWAS study examining associations with individual temprements and will form a cross-test of methodological reliability.
AIMS. The aims of the broader research program are to:
1.Characterise individual difference in child emotional reactivity through assessment of the the statistical association of genotype with physiological, observational, physiological, and report measures.
2.Examine the interaction of child emotional reactivity with qualities of the childcare environment to and assess impact on social-emotional regulation and behavioural development.
3.Identify the meaning of individual difference in children's reactivity for policy and practice in childcare settings.
The specific aims relevent to the use of the ALSPAC data will be to valiate genotypic markers identified within our target cohort, by examining the same geneotypes and their association with phenotypic variables of temprement, sleep, behaviour and behavioural diagnoses within the ALSPAC data set.
VARIABLES
Genotypes of interest will be ientified via an indepenent cohort study to be conducte as part of the grant application and are not known at this time. Phenotypic variables of interest will be temperament, behaviour, behavioural diagnoses and sleep characterisitcs. Socio-economic status and childcare charatersitics of the children are also requested.
B2334 - Effect of Socioeconomic Depreciation on Child Development A Longitudinal Study - 20/11/2014
Socioeconomic stability plays an important role in the normal development of a child. So, it is not surprising that adverse socioeconomic conditions arising from difficult financial circumstances have been associated with a number of mental health and cognitive problems. The motivation for the current line of thinking is a recent study published in JAMA: Psychiatry wherein we noted that macroeconomic shocks such as recessions can have long-term effects on the mental and emotional health of children between the ages of 0-18 months. Over the last two years, we have been expanding our work on the concept of socioeconomic depreciation (SED) or drop in family income. In this proposal, we will examine the mechanism behind the influence of SED on future behavior problems. Specifically, we are proposing that SED will affect attachment, which in turn will affect social development. Since attachment variables are not available in the ALSPAC data, ALSPAC will not be used to develop this particular hypothesis. However, ALSPAC will be used to explore the role of social development in future behavior problems. (Figure 1 available on request).
Secondly, we propose to explore the effects of prenatal exposure to SED. Borrowing from literature on fetal programming, we propose that prenatal exposure may lead to changes in temperament and increase vulnerability to behavior problems. Temperament has been described as a stable system of traits that moderates susceptibility to behavioral and emotional problems. In line with this argument, this study will examine if temperament moderates the link between SED and behavioral and emotional problems (Figure 2 available on request).
This proposal also extends our work into SED by testing for gene by environment interactions using the large public data sets from the Psychiatric Genomics Consortium (PGC). Because Dr. Faraone is on the PGC's Coordinating Committee, this aspect of the work will be straightforward. For each disorder, the PGC computed a SNP heritability, which indexes the amount of heritability due to common SNP variants. These SNP heritabilities were all much greater than zero and were statistically significant. These analyses also showed substantial and significant correlations between the heritabilities of some of these disorders. For example the correlation between schizophrenia and each of the mood disorders was greater than 0.4. This cross disorder polygenic overlap, along with prior evidence of overlap for single SNP variants and for CNVs, has confirmed prior work from genetic epidemiology indicating shared genetic risk factors among multiple psychiatric disorders. This work is relevant to genetic studies of SED because it allows us to create molecular polygenic scores for use in testing gene by environment interaction. These scores may explain, for example, whey two children raised in identical SED circumstances have different outcomes. When added to knowledge about SED, these genetic data could help allocate scarce resources where they are most urgently needed.
For this project, we propose to use ALSPAC. Details about our proposed work with ALSPAC are discussed below. ALSPAC data will be used to test the following hypotheses. The scientific outline (Section 8) provides a list of specific variables requested, and how those variables will be used in hypothesis testing. In that table, DV refers to Dependent (or Outcome) Variable, IV to Independent (or Exposure) Variable, MV to Moderating Variable, and CV to Confounding Variable.
B2333 - Association between asthma/allergies and depression/anxiety/self harm - 20/11/2014
Several reports indicate that asthma and depression co-occur(1). The association has been reported in countries with different social and cultural histories (2) suggesting a true association between these two conditions. In addition, some studies report a higher rate of suicide or suicide-related behaviours among asthma patients(3). Several pathways that include both directions of association (asthma leading to depression or depression leading to asthma) or common aetiological process that would increase the risk of both conditions have been proposed.
AIM. To determine whether there is an association between asthma/allergy and depression/anxiety/self-harm.
Specific objectives are:
1.Establish whether there is a cross-sectional association between asthma/atopy and depression/anxiety self-harm in children and adolescents.
2.Establish whether children with asthma/atopy have higher risk of developing depression/anxiety self-harm (prospective).
3.Establish whether cross-sectional and prospective associations between asthma and depression vary by gender.
4.Determine the pathways underpinning an association between asthma (and/or atopy) and depression (anxiety/self-harm).
B2332 - Autism sleep mental health parent distress and chronic pain in childhood - 20/11/2014
Aims:
To describe the prevalence of pain within those with diagnosed autistic spectrum disorder.
To describe the prevalence of pain within those with autistic like symptoms including social and communication difficulties, prematurity and delays reaching developmental milestones.
To explore if depression and anxiety are higher amongst those with autism and pain.
To explore if people with autism and pain suffer from sleep deprivation.
To explore how prevalent sleep deprivation is in people who present with persistant pain compared to those who present with autism and persistent pain.
To explore if sleep is dependant on anxiety scores in those with pain more than those without pain.
To explorwe if children with a diagnosed autistic spectrum disorder have other chronic sleep, pain or anxiety conditions.
B2331 - Translational Psychiatry - genetic variation and dysfunction of human ARC gene - 20/11/2014
The immediate early gene, ARC, is a key regulator of protein synthesis-dependent synaptic plasticity. Synaptic plasticity functions in memory as well as in adaptive changes related to fear, anxiety, and reward. Dysfunction of synaptic plasticity is increasingly implicated in neuropsychiatric conditions including depression, schizophrenia and autism spectrum disorders. Our recent work links several SNPs in Arc to cognitive function. We tested whether ARC variants are associated with six measures of cognitive functioning in two healthy samples-the Norwegian Cognitive NeuroGenetics (NCNG) sample and the Swedish Betula sample. Of a total 71 ARC variants, 21 were nominally associated with human cognitive functions involving semantic knowledge, visuospatial abilities, delayed episodic memory, and general cognitive abilities (IQ).
Preliminary meta-analysis of large-scale genome-wide association studies' (GWAS) results across two neuropsychiatric conditions showing the most genetic overlap - SCZ and BP (http://www.med.unc.edu/pgc) - reveals a synergetic effect of two SNPs located in the regulatory downstream 3' region of the ARC gene (rs7465272 and rs7835613, p = 2.31e-06 and 9.01e-06 respectively). Adding MDD to the joint meta-analysis further reinforces the involvement of one of those SNPs in the susceptibility to neuropsychiatric conditions (rs7835613, p = 7.90e-06). Taking these findings together with known neuronal functions of the ARC gene, we hypothesize that its genetic variants may accentuate such common genetic architecture of cognitive abilities in neuropsychiatric conditions.
Our planned analyses:
1. Genetic association of ARC gene complex with cognitive function (cross-sectional and longitudinal examination)
1.1 ARC gene and cognitive functioning
AIM AND HYPOTHESIS
Here we plan to perform regression analyses to investigate whether common ARC variants are associated with cognition or other aspects of brain plasticity. Since previous studies are suggestive to this association, we hypothesize that various aspects of cognition , plasticity, and/or neuropsychiatric disorders will be associated with ARC variants.
B2330 - An investigation of maternal risk factors for the development of paediatric obsessive-compulsive disorder - 13/11/2014
BACKGROUND:
Obsessive-compulsive disorder (OCD) commonly has its onset in childhood, affecting 1-4% of young people under the age of 18 (Douglass et al, 1995; Flament et al., 1988; Heyman et al, 2001) . The disorder is highly disabling across domains (Piacentini et al, 2003) and is predictive of a range of negative outcomes (Wewetzer et al., 2001; Micali et al, 2010). Furthermore, if left untreated the disorder typically follows a chronic course (Skoog & Skoog, 1999) and is major source of economic burden, ranking among the top 10 most disabling illnesses recognized by the World Health Organization (Murray & Lopez, 1996). There is a pressing need to further understand the mechanisms underlying the development and maintenance of pediatric OCD, in order to facilitate effective prevention and intervention.
OCD results from a complex interplay between genetic and environmental factors. Twin studies have shown that genetic factors explain 45-65% of the variance in OCD symptoms in children, with the remainder being accounted for by shared and non-shared environment (Pauls, 2008). The specific environmental risk factors remain unknown, but it has been proposed that aspects of family environment contribute to OCD (Van Noppen et al, 2009). Consistent with this, previous studies have demonstrated that parents of young people with OCD show elevated rates of critical and over-intrusive parenting styles (Barrett et al, 2002; Hibbs et al, 1991) and psychiatric disorders (Lenane et al, 1990). However, a major limitation of the research in this field to date is the reliance on cross-sectional designs, meaning that the direction of effects between childhood OCD and parenting style/parental psychopathology remain unclear. To date, no prospective studies have been conducted to test whether such family factors predict the onset of OCD in children.
AIMS:
This primary aim of this study is to test whether maternal parenting style and maternal psychopathology in early childhood predict the subsequent development of OCD symptoms in youth. A secondary aim is to test whether the persistence of OCD symptoms in young people is mediated by maternal parenting style and maternal pscyhopathology.
B2329 - Genic and intergenic heritability analysisof general cognitive ability - 13/11/2014
Aims
Genome Wide Complex Trait Analysis (GCTA) has been applied to Genome Wide Association Study (GWAS) data sets of cognitive ability to show that around half of the phenotypic variation is tagged by common variants (Davies et al., 2011). However, GCTA cannot provide information regarding which variants are more important. This study aims to find regions in the genome that make a significant contribution to the GCTA heritability estimate. Three groups of single nucleotide polymorphisms (SNPs) will be formed; the first group will consist of all SNPs available for use, the second set will consist of SNPs found in genes, whereas the third will be formed from SNPs that are found between genes. GCTA will be perfomed on each group allowing for both the genic and intergenic SNP-sets to be tested against 0 in order to determine if they each make a significant contribution toward intelligence differences. In addition the two groups will be compared against each other to show which regions matter most to human intelligence differences. Using GCTA in this way will help to show that variants within genes account for a significant proportion of the heritability of human intelligence. This will justify the use of gene and gene-set based analyses well as indicate which areas of the genome are of particular importance to cognitive ability.
Hypothesis
It is expected that for general cognitive ability both the geneic and the intergenic SNPs will make significant contributions to the total heritiability estimate. It is also expected that SNPs found within genes will also make a greater contribution toward the heritability estimates than SNPs found between genes as has been demonstrated for height (Yang et al., 2011) and indicated for schizophrenia (Schizophrenia Working Group of the Psychiatric Genomics Consortium, 2014).
Exposure variables
GCTA has been performed previously on general cognitive ability in The Avon Longitudinal Study of Parents and Children (ALSPAC) group (Benyamin et al., 2013) as it has a suitably large sample size (5517 at age 9) and an execellent test to derive an Intelligence Quotent (IQ) estimate, namely the Wecshler Intelligence Scale for Children-III (WISC-III). For the proposed analyses to take place, the genotype data along with Verbal (Information, similarities, arithmetic, vocabulary, and comprehenssion) and Performance (picture completion, coding, picture arrangement. Block design, and object assembly) subscales of the WISC-III will be used to form an IQ score. GCTA analysis will be performed on this IQ score using all SNPs available for analysis. Next, the total set of SNPs will be divided into genic and intergenic SNP-sets before GCTA is performed on both SNP-sets.
Outcome variables
The proportion of phenotypic varainace attributable to the genetic variation in each of the three groups will form the outcome variable.
Confounding variables
One potential source of confounding would be that the two SNP-sets could tag the same variants due to the existence of linkage disequilibrium.
B2327 - Vitamin D in Pregnancy and Perinatal Depression - 12/11/2014
Background:
Results from a recent meta-analysis suggest that low Vitamin D levels are associated with depression1. As pregnancy is a time for increased nutritional demands, precipitation of depression would be more likely during pregnancy if this hypothetical association were causal. Low vitamin D levels are common during pregnancy, particularly in high-risk groups.2,3,4 Some studies have reported an association between low vitamin D levels and depressive symptoms during pregnancy5,6 and in the postnatal period.7,8,9 However, other studies have not supported this hypothesis.10 Previous studies have been limited by sample size7,9, identification of depression10, or follow up time to development of symptoms.5,6,8
Aims:
*To determine the relationship between vitamin D (serum 25-hydroxyvitamin D3 and D2) and perinatal depression.
*To personally develop research skills in preparation for applying for research fellowships
B2318 - Prenatal and early childhood environmental risk factors for psychiatric outcomes using a novel exposure biomarker - 06/11/2014
There is substantial evidence that early environmental insults increase risk for psychiatric disorders,
especially those with strong neurodevelopmental origins. Thus, there is a need to reliably identify these
environmental factors, as well as whether higher exposure level during critical periods for
neurodevelopment are related to greater risk for psychiatric disorders. Moreover, while environmental
toxins have been reliably associated with early neurodevelopment, similar investigations into longer term
and psychiatric outcomes are scarce.
Current approaches use indirect measures of fetal exposure, such as chemical concentrations in maternal
blood and urine, which may not accurately reflect fetal uptake for many toxins due to variable partitioning
across the placenta (Smith et al., 2007). Thus, a direct marker of early exposure is needed to aid our
understanding of the etiology of psychiatric disorders, as well as to aid prevention.
The current study proposes to use a unique and well-characterized birth cohort: the Avon Longitudinal
Study of Parents and Children (ALSPAC) and applying an extensively developed and piloted method to
determine multi-toxin early exposure (Arora et al., 2006, 2011; 2012; 2013; Hare et al., 2011; Austin et
al., 2013). Human deciduous teeth undergo systematic mineralization commencing prenatally and are
composed of a collagen-based, lipid-containing, calcium-rich apatite matrix, which accumulates bone
seeking elements and organic compounds. The methodology combines detailed histological analysis and
associated analytical tools to establish a comprehensive record of weekly to monthly exposure history of
multiple chemicals and metals and history of infections from the second trimester through early
childhood.
Aims
The overall goal of this project is to determine the association of early life environmental metal and
chemical toxin and infection exposures with multiple psychiatric and neurological outcomes: psychotic
like experiences, depression, mild cognitive impairment and autstic-like traits. A detailed history of
exposure during prenatal and early childhood development will be established to distinguish between two
important aspects of the exposure profile:
1) Exposure timing: We will construct weekly to monthly exposure history over the prenatal and early
childhood periods in order to identify the developmental period(s) most strongly associated with risk for
poor psychiatric and neurological outcomes.
2) Exposure intensity: Cumulative exposure may be more important than exposure during any single
developmental window and our biomarker allows measurement of cumulative exposure from the prenatal
period to the time of tooth shedding, which occurs between ages 6 to 12 years.
B2316 - Examining the long-term outcome of early onset depression in 10-13 year olds - 31/10/2014
Aims
1. To investigate the long-term outcomes of early onset depression in children affected at the ages of 10 and 13 years.
2. To examine the potential moderating effect of factors that might predict long-term outcome, including comorbid psychiatric disorder and family history of depression.
3. To examine the course and correlates of depressive symptoms and disorder across adolescence in boys.
Background
Depression in pre-adolescence is a serious, neglected disorder. It is common, has poor outcomes, and has been the subject of very little research. This project will investigate the course of early onset depression in a large population cohort study
B2315 - Bayesian individual knot point models for modelling growth - 31/10/2014
This project is IEU research, linked with Laura Howe's MRC fellowship and the IEU stats theme. Analysis will be carried out by Sam Brilleman with supervision from Laura Howe and Kate Tilling.
Aim: to evaluate the use of Bayesian individual knot point models for modelling growth across childhood and adolescence and estimating age of puberty onset, and to compare these models with alternative approaches such as SITAR
Exposure: age
Outcome: height and weight
Confounding variables: gender
Project justification: Self-reported measures of pubertal status can be inaccurate and there is often lots of missing data due to participant embarrassment. Height-based indices of pubertal development offer an attractive alternative that may be both more accurate and result in less missing data. In this project, we will use all available height and weight data to estimate trajectories of growth across childhood and adolescence, using models that allow person-specific ages at which the rate of growth changes. The idea is that the timing of these knowpoints may be informative about pubertal growth spurts. The models will be compared with other approaches to identifying pubertal growth spurts from growth trajectories, e.g. superimposition by translation and rotation (SITAR), a method developed by Tim Cole.
B2314 - Handling multivariate missingness in health and behaviour data via multiple imputation under MNAR Issues in Practice - 31/10/2014
BACKGROUND:
Data collected by ALSPAC to explore potential predictors of adolscent alcohol, tobacco and cannabis consumption exhibit substantial levels of missingness in both outcomes and covariates and are plausibly missing not at random (MNAR). This is also the case for data collected by ALSPAC to explore potential correlates of adolescent mental health. Standard maximum likelihood and Bayesian full probability modelling procedures can quickly become computationally infeasible in the presence of multivariate MNAR missingness. We propose to tackle this difficulty by extending a popular procedure for performing multiple imputation under MAR, Fully Conditional Specification, to handle MNAR missingness mechanisms. Our method will be broadly applicable to future analyses of the ALSPAC data and to both cross-sectional and longitudinal data.
AIMS:
1. To develop a new statistical method for handling datasets with multivariate MNAR missingness.
2. To use this new method to further explore the missing data mechanism for
a. key variables on alcohol, tobacco and cannabis use in teenagers.
b. key variables on mental health outcomes in teenagers.