A. SIGNIFICANCE Risk Factors for Autism Risk Factors for ADHD/HKD Risk Factors for Delayed Infant Milestone Achievement

Transcription

1 A. SIGNIFICANCE Neurodevelopmental disorders are often characterized by impairments in growth or development of the brain, or central nervous system (CNS), or as disorders of brain function that affect emotion, learning ability, and memory and that unfold with increasing age 55. It has been suggested that organization of the brain is highly plastic in fetal life and that establishment of healthy neural functional systems during the fetal period is critical to normal development 56. In the last few decades there has been tremendous growth in knowledge about neurological development in young children and the vulnerability to adverse exposures during fetal life 49,57. While initial studies of adverse fetal environments relied on phenotypes (i.e. being born early or too small) 58,59, more recent investigations have demonstrated associations between markers of psychological stress, maternal infections, nutritional deficits, and prescription drug use in relation to cognitive outcomes in the offspring During childhood, several neurodevelopmental outcomes emerge that persist through the lifespan; ; here we focus on specific neurodevelopmental diagnoses, autism spectrum disorders (ASD) and attention deficit hyperactivity disorder (ADHD) / hyperkinetic disorder (HKD), but also include IQ and achievement of motor and behavioral developmental milestones in our analyses. Risk Factors for Autism In the U.S. autism is one of the most severe neurodevelopmental disorders in children, defined by impairments in three main domains: social interaction, language and range of interests 64. There have been marked increases in the rate of diagnosed cases of autism over the past two decades and most recent U.S. estimates suggest an increase in prevalence as high as 30% in 8 year olds between 2008 and 2010, approximating an overall prevalence of 1 in 68 children 69. Autism has a heavy male bias, as almost five times more boys are being diagnosed than girls Age specific prevalence across birth-year cohorts from 1990 to 2007 have also increased for autism and ASD in Denmark 73. The condition is presumed to be present at birth or shortly thereafter 74,75, though symptoms are detected at different ages and diagnosis can usually be made by age three 70,71. It has been suggested that this rise in autism prevalence could at least partially be explained by declining age at diagnosis and inclusion of milder cases, while many experts maintain that etiologic factors may also contribute 76. Large population based studies have established a correlation between ASD and family history of autoimmune diseases, associations with MHC complex haplotypes, and abnormal levels of inflammatory cytokines and immunological markers found in sera 77. Relatively few risk factors have been identified for non-familial autism 71,78,79. Perinatal factors associated with autism include fetal distress, birth injury, multiple birth, preterm birth and low birth weight, congenital malformations, 80,81 and pregnancy complications in general 82. Other parental risk factors for autism include older parental age 80,81,83 86, parental obesity 87 and maternal migration 70,80 ; roles for immunological 88,89 and endocrine factors have also been proposed 90. Risk Factors for ADHD/HKD Attention deficit hyperactivity disorder, characterized by inattention, hyperactivity, increased impulsivity and motivational/emotional dysregulation 91, is one of the most common neurobehavioral disorders in children and has an estimated worldwide prevalence of 4-8% 92,93. Hyperkinetic disorder is often characterized as a particularly severe form of ADHD ; however, the key difference is source of diagnostic criteria. ADHD diagnostic criteria are derived from the American Psychiatric Association's Diagnostic and Statistical Manual (DSM-IV, 1994) 97, whereas criteria HKD criteria are derived from the International Classification of Diseases (ICD-10) 98. While there is a substantial difference in the prevalence of ADHD and HKD in clinical samples 95, studies examining the predictive validity of the two main diagnostic schemata find large overlap in terms of important clinical characteristics 92,95,96. The etiology of ADHD/HKD is not well understood but both environmental and genetic factors contribute to disease onset 99,100. As with ASD, it has been argued that the rapid increase in ADHD/HKD observed over the past few decades 101,102 cannot solely be attributable to changes in diagnoses 103 or parental awareness 94, necessitating the search for other causes. The neuropathology of ADHD/HKD may already be present at birth, making exposures during pregnancy and/or infancy of particular interest 104. prenatal exposures associated with ADHD/HKD include maternal diabetes 16, maternal smoking 93,105 ; maternal consumption of alcohol or caffeine 109 ; and a modest contribution is attributed to psychological stress 110,111. Risk Factors for Delayed Infant Milestone Achievement Achievement of developmental milestones as expected is a useful measure of cognitive function in infants and young children 120, and several studies have found significant associations between milestones achievement and outcomes later in life. The Northern Finland Birth Cohort, established in 1966, found that earlier achievement of infant milestones was associated with higher mean scores of teacher ratings, and higher education

2 attainment at age ; other cohorts have found similar findings 122. Delayed achievement of developmental milestones has also been linked to schizophrenia 123 and other mental health illnesses 124. Some maternal factors have been identified with delayed achievement of milestones during infancy; these include maternal diabetes 16, smoking, alcohol consumption 125, depression 126, use of antidepressants 125,127, and short (or absent) duration of breastfeeding 128. Risk Factors for Lower IQ Scores The validity of IQ (intelligence quotient) scores as a neurodevelopmental marker has been debated 129, but is still widely used. Several prenatal factors have been identified to have harmful effects on offspring intelligence. A study on cognitive function found that infants born small for gestational age had lower IQ scores compared to their non-sga counterparts 130. Likewise, low gestational age and birth weight has been associated with reduced school performance 131 and IQ test scores in young adults 132. Maternal measures of lipid and glucose metabolism during gestation have been related to poorer performance on standardized IQ tests in the offspring 22. Family composition such as birth order has also been associated with cognitive development; though more recent studies have found this association to disappear once family size is accounted for 133,134. It has been shown that maternal intelligence is an important predictor of offspring intelligence, more so than race, education, age, poverty status, smoking, the home environment, the child's birth weight or birth order 135 ; this is evident even after controlling for breastfeeding, a factor often positively associated with IQ 135. Evidence that Maternal Diabetes Affects Neurodevelopment There are several plausible biological mechanisms by which maternal hyperglycemia can program fetal neurodevelopment. First, persistent excess exposure to elevated maternal glucose can lead to fetal hyperinsulinemia, which triggers the fetus to increase consumption of oxygen and metabolism, which then induces chronic intrauterine fetal tissue hypoxia Second, the increase in fetal insulin secretion can also hyperactivate the fetal insulin signaling pathway PI3K, which has been postulated to interfere with neurodevelopment 136. Third, excess adiposity associated with type 2 diabetes and gestational diabetes can induce chronic inflammation. Animal models where pups have been injected with inflammatory cytokines IL-1 26, IL-2 27, IL-6 28 or interferon-γ (INF γ) 29 have resulted in abnormalities in white matter of the brain as well as behavioral and cognitive deficits in the offspring 30. Furthermore, when increased levels of maternal IL-6 cross the placenta, deficits in fetal brain development have been seen such as onset of seizures, impairments in spatial learning, increases in HPA activity, disturbances in cholinergic input to the hippocampus, and reduced neurogenesis in adult hippocampus 137. In prenatal maternal diabetes, it has been noted that an increased level of IL-6 and other proinflammatory cytokines are produced 138. A vast amount of literature also exists to suggest that prenatal exposure to infectious agents and the associated inflammation can contribute to neurochemical, hormonal, cytoarchitectural and/or behavioral abnormalities associated with autism, schizophrenia, cerebral palsy, blindness, and/or mental retardation 30,139. Moreover, it has been suggested that acute neuroinflammation during early phases of fetal development may induce neuropathological phenotypes shared by both schizophrenia and autism, and that subsequent latent and persistent inflammation may respectively lead to phenotypic characteristics of schizophrenia and autism 140. Fourth, excess exposure to elevated maternal glucose has been associated with increased free-radical production, impaired antioxidant defense, an acceleration in oxidation reactions and degradations of lipids 31,32. Adverse associations between oxidative stress and neurodevelopment in the offspring have been reported for several ASD investigations Finally, altered maternal autoimmunity found in mothers with type 1 diabetes may influence brain development in the offspring by creating an adverse intrauterine environment 36,37 or by altering offspring autoimmunity in early neurodevelopment through immunoglobulin G 38. Increasing Incidence of Diabetes and Changes in Diagnostic Criteria Over Time The worldwide prevalence of diabetes has been increasing ; in the U.S. about 9% of women of childbearing age have diabetes 146 and during pregnancy an additional 2-10% will develop gestational diabetes 147. It has been estimated that diabetes affects up to 15% of all pregnancies 148 ; and approximately 87.5%, 7.5% and 5% are due to gestational, type 1 and type 2 diabetes, respectively. Recent data from the U.S. show that the incidence of gestational diabetes over the past 20 years has been increasing by % among certain ethnic groups. The frequency of gestational diabetes typically reflects both the frequency and the number of women at risk of type 2 diabetes in the underlying population 149 ; established risk factors include advanced maternal age, obesity and family history of type 2 diabetes 150. Gestational diabetes is defined as diabetes or impaired glucose tolerance recognized for the first time during pregnancy and this definition applies regardless of whether insulin or diet modification is used for treatment. But it does not exclude the possibility that unrecognized glucose intolerance may have antedated or arisen concomitantly with the pregnancy 151. Details of diagnostic criteria for diabetes have changed overtime and reflect the best evidence

3 available; presently, no universal diagnostic guidelines exist. The U.S., European countries, and Australia have each developed their own guidelines through the years (for Denmark see subaim 1C), all based either on the maternal risk of subsequent diabetes or other statistics. Hence, population specific shifts in gestational diabetes incidence reflect both changes in population risk factors in addition to changes in diagnostic criteria. Summary In summary, given the rising global trends in obesity of all age groups 39, including women of childbearing age, gestational 40 and type 2 39,41 diabetes, the evidence linking these health conditions with ASD, ADHD/HD, IQ levels and delayed development, and the paucity of studies with appropriate data for critically examining the association between maternal diabetes and offspring neurodevelopment, we propose a series of nationwidepopulation cohort studies in Denmark to test the hypothesis that maternal diabetes during fetal development increases the risk of ASD, ADHD/HKD, and specific developmental delays. The proposed population based cohort studies will contribute new knowledge to what is known about the fetal programming of maternal diabetes on the neurodevelopmental of the offspring. B. INNOVATION The Danish National Registers provided a unique opportunity to assemble the largest nationwide population based cohort in existence to date with phenotypically well-characterized cases and with information on personal characteristics 44, including a broad range of lifestyle and workplace exposures. Denmark has a longstanding tradition of maintaining one of the world's largest and oldest health science information data banks, which converted to electronic records in 1973, and has provided tax-supported health care and prescription medication coverage for all residents of the country for more than three decades; resources not available anywhere in the U.S. The resources available for fetal programming research in Denmark are unique in that they span more than three decades of data collection and are available for every inhabitant of Denmark. In this setting, our research group, in collaboration with senior health researchers in Denmark, will conduct this large cohort study of approximately 1.6 million pregnancies for whom we have longitudinal data from national registries including hospitalizations, outpatient visits, and residential histories. Uniquely positioned in this population is a large established nested cohort which provides in depth information on comprehensive maternal histories, environmental, occupational, and lifestyle exposures. Through past, existing and ongoing collaborations with Danish colleagues, Dr. Virk has been granted access to their large nationwide registries and has gained a solid understanding of these data systems. These skills combined with the skills she will acquire from the training outlined in this application will uniquely position her to perform future GxE and GxG analyses that are feasible with the data sources available in the Nordic biobanks. Dr. Dan Geschwind is a recognized leader in neurobehavioral genetics and his close involvement with Dr. Virk s training is a significant strength in this proposal. Dr. Virk s previous training in maternal and child health and epidemiology will give her a unique perspective to conduct future studies in neurobehavioral genetics and greatly advance her potential to make innovative scientific advancements the field in fetal neurobehavioral epidemiology. Overall, the mentors and consultants in this team bring a wealth of experience in fetal programming and population research, neurobehavioral genetics, clinical management of diabetes and statistics that utilize national registers, which in combination with the efficient, and cost effective linkable population registry data permit the assembly of a large, unique study population and identification of an array of outcomes for addressing the etiological impact of maternal diabetes on the neurodevelopment of the offspring. Dr. Ritz has decades of research experience in neurological and perinatal outcomes and has published extensively in highimpact factor journals; she also mentored many doctoral and post-doctoral level trainees and has significant experience in this area. Dr. Olsen has a vast amount of experience with register based research (more than 600 peer-reviewed publications) and is the original primary investigator of the DNBC. Dr. Tania Esakoff, director of the Diabetes in pregnancy program at Cedar Sinai Medical Center will inform the study in clinical aspects of managing diabetes as well interpreting laboratory results for screening. Dr. Diana Schendel s international efforts in registry collaborations has resulted in the formation of multi-national databases (icare and MINERvA) and is a member of the ipsych effort aimed to assess psychiatric outcomes including autism, using interdisciplinary state of the art approaches. She also has long-standing ties with investigators at Aarhus University, the Danish Neonatal Screening Biobank and in Finland, autism researchers using the Finnish Maternal Serum Biobank. She also has extensive experience with the Nordic data sources that will provide insight on cohort and variable construction. Dr. Schendel has published extensively in highly ranked peer reviewed journals and is an international expert on autism spectrum disorders. In summary, this proposal

4 will greatly advance what is known about maternal diabetes and neurodevelopmental outcomes, and also provide training for Dr. Virk to make innovative scientific advancements to further the field of fetal neurodevelopmental epidemiology. C. RESEARCH APPROACH Aim (1): Establish the first, largest and most comprehensive nationwide register-based study to date of specific neurodevelopmental outcomes and maternal diabetes ( ). Utilizing the Danish Civil Registration System and Danish medical registers we will identify all Danish pregnancies during the period , determine which of these offspring were exposed to maternal diabetes [defined by either an ICD-8 ( ) or ICD-10 (starting in 1994) codes for type 1, type 2, or gestational diabetes], link the entire cohort to maternal demographic variables held in the Danish Fertility Register 46, and ascertain offspring with neurodevelopmental disorders (ASD, ADHD, and HKD). Data sources will be linked using each individuals civil registration number (CPR), which is assigned by the national Central Population Registry and provided to each person residing in Denmark. Data linkage and analyses will be performed on a secure server located at the Statistics Denmark (DST), the governing agency of all Danish registers. Data will be accessed remotely using a VPN connection through Aarhus University with Dr. Virk s existing accounts, which grants all necessary privileges (see letters from collaborators for permissions). The proposed study would make data linkages from the Central Population Registry (CPR) to other medical registers which include the National Patient Register, Clinical Laboratory Information System, Birth, Psychiatric and Conscript Registers (see figure-1 for population estimates 73 ). The Central Population Registry (available electronically since 1968) acts as source material for all administrative systems in Denmark. Each Denmark inhabitant, including foreign born immigrants and temporary residents, are assigned a unique 10 digit identification (CPR) number. Updates are made continuously (weekly starting in 1968 and daily since 1989) by local registration offices. Among other information, the Population Registry links parents to their children (and siblings) and contains information on date and place of birth, vital status, date of death, emigration, date of marriage and divorce, current job-title, current and previous addresses. 44 Information on for each individual can be linked to all other Danish registers using the unique CPR number. The Danish National Patient Register 45,152 contains information on almost all hospital admissions in Denmark electronically since 1977, as well as records for all outpatient visits and emergency room consultations since Each record includes the person s ID, date of admission or first outpatient visit, date of discharge or last outpatient visit, surgical and diagnostic procedures undertaken, the primary diagnosis, and up to 19 auxiliary diagnoses. Diagnoses are coded according to a modified Danish version of the eighth revision of the International Classification of Diseases (ICD-8) up to 1993, and thereafter to the tenth revision of the Classification (ICD-10). The register is managed by the Statens Serum Institute (SSI), receives new data from all Danish hospitals once a month and is immediately subjected to a primary validity check and corrections are made as appropriate. In-depth validation has been performed on an ad-hoc basis 153,154 ; a validation study performed in 1990 showed that information on whether or not an admission was acute was correct in 98.5% of all compared records. Diagnoses for gestational diabetes are reported by midwives and primary care physicians, into the Danish National Patient Register. No existing longitudinal population trend estimates for gestational diabetes are available, though presently it is estimated to occur in 2-3% of pregnant Danish women. 155 Other adults or children suspected of having diabetes are seen by an endocrinology specialist and once evaluated and a formal diagnosis has been established, reports are entered into the register database without regard to treatments. It been estimated that national and regional registers identify 89-96% of all patients diagnosed with diabetes 156.

5 The Danish Clinical Laboratory Information System (LABKA) 157 contains laboratory results from every blood specimen obtained from any public or private hospital or by any general practitioner and submitted to any clinical chemistry department located in three of the five Danish administrative regions (North Demark, Central Denmark and Copenhagen Region) from 1985 onwards. This data source accounts for approximately 43% of the Danish Population. Clinical data on OGTT and HbA1c measurements are reported into the regional clinical databases by midwifes and physicians. Information is recorded according to the international Nomenclature, Properties and Units coding system 158 which provides terminology for identification of clinical laboratory test values, and ensures international uniformity. Additional information includes patient CPR numbers, date and a code identifying the physician/hospital department requesting the analyses. The Danish Medical Birth Registry collects information on all births in Denmark with the mothers personal identifying number as the key identifier; it was established in 1968 and has been computerized since Data for this register are collected by midwives on maternal pregnancy history, maternal residence, partner cohabitation status, whether the mother smoked in pregnancy (available from 1995 onwards), birth weight, birth length, gestational age, maternal pre-pregnancy BMI (available from 1997 onwards), birth complications, APGAR score and other gestational factors. Validation studies have demonstrated good agreement between the registry and medical records 43. Linkage to this registry will provide important information on potential confounding variables. The Danish Psychiatric Central Research Register (established in 1938, computerized in 1970) contains data on psychiatric admissions to psychiatric hospitals and wards in Denmark, the Faroe Islands and Greenland. Since 1995, this data also includes psychiatric outpatients 159. Children who are suspected of having an ASD or HKD are referred by general practitioners or school psychologists to a child psychiatric ward, where they are evaluated by a multidisciplinary team and assigned a final diagnosis by a child psychiatrist. All cases of ASDs and HKDs are registered in the Psychiatric Register once a formal diagnosis is established and without regard for treatment or educational provisions. A recent validation study shows 94% accuracy in diagnosed autism cases from this register, of the remaining 6%, 5% were diagnosed with an autism spectrum disorder other than autism 160. Linkage to this registry and the patient registry will identify ASD and HKD cases as well as the date of diagnosis. Danish registry-based population prevalence estimates for ASD, HKD and autism in 5-year olds are provided in figure 2, 73 which illustrate the increasing prevalence of children diagnosed ASD, autism and HKD over time. The register contains patient CPR-numbers, dates of admission and discharge, type of admission (emergency room, outpatient or inpatient) and primary and secondary discharge diagnoses is reported to the register 161. The Danish Conscript Registry collects information on military conscripts 162, IQ is estimated using the Danish military draft board s Børge Priens Prøve (BPP) test 47. This test is administered to all Danish men 18- years of age, with the exception of individuals providing medical documentation for conditions that would disqualify them for military service which include epilepsy, spinal osteochondrosis, mental retardation, diabetes and severe asthma (approximately 5-10% of boys from each birth cohort 163 ). The test takes less than one hour to complete, has four sub-tests (letter matrices, verbal analogies, number series and geometric figures) and a total score ranging from 0 to 78. The BPP has remained unchanged since its first use in the 1950s, including its paper and pencil format. The total BPP score has a correlation of 0.82 with the full scale IQ of Wechsler s Adult Intelligence Scale, indicating that the BPP is a suitable measure of general intelligence 164. Sub Aim (1A): Measure the association between maternal diabetes during pregnancy and risk for autism spectrum disorders, attention deficit hyperactivity disorder and IQ. Using the population medical (which include hospital discharge, outpatient, and psychiatric) registers we will ascertain which offspring in our cohort have a diagnosis for: (1) Autism Spectrum Disorders (ASD); (2) Hyperkinetic disorder (HKD)/Attention Deficit Hyperactivity Disorder (ADHD); and link all male children with the military conscript data to obtain IQ test scores. Statistical Methods. We will use standard analytical methods for cohort studies 165,166 to examine the associations between prenatal maternal diabetes identified in the hospital and patient registers and defined as preexisting type 1 diabetes (ICD-8 codes: 249; ICD-10 codes: E10, O24.0), pre-existing type 2 diabetes (ICD-8 codes: Figure 2: Prevalence estimates for ASD, Autism and HKD (95% confidence intervals) per 10,000 children for each birth cohort at 5-yrs Birth Cohort ASD Autism HKD ( ) 4.7 ( ) 6.5 ( ) ( ) 9.5 ( ) 7.0 ( ) ( ) 14.1 ( ) 10.4 ( ) ( ) 19.3 ( ) 15.8 ( ) ( ) 19.3 ( ) 20.2 ( ) ( ) 22.1 ( ) 19.0 ( )

6 250; ICD-10 codes: E11, O24.1) or diabetes with gestational onset (ICD-10 codes: O24.4, O24.9); with autism (ICD-8 codes: 299.0, ; ICD-10 codes: F84.0, F84.1), ASD (ICD-8 codes: 299.0, ; ICD-10 codes:f84.0-f84.9), and HKD (ICD-10 codes: F90.0-F90.9), see figure 2 for prevalence estimates. In our analyses, we will stratify by type of diabetes (pre-existing type1, type 2 diabetes, and gestational diabetes) and furthermore we will stratify by trimester of onset for those women diagnosed with diabetes during gestation. In supplementary analyses, we will evaluate a diagnosis for an abnormal glucose tolerance test (ICD 10 codes: R73.0) with a diagnosis for autism, ASD and HKD. The birth rate has remained fairly constant in Demark resulting in approximately 66,000 births per year. Thus, the age range of children in our cohort will span from birth to 35 years. We will employ Cox proportional hazards regression, to estimate hazard ratios (HR) to obtain point and 95% confidence interval estimates. We will use person-years as the offset variable, which will begin at date of birth and end at the first date of diagnosis (for any of the outcomes considered in this application), death, emigration or end of follow-up whichever comes first. In all models, we will adjust for other potential confounding risk factors, such as maternal and paternal age, maternal education, maternal income, maternal pre-pregnancy BMI (available from 1997 onwards), maternal pre-pregnancy history of hypertension and mental illness diagnosis. We will use either regression-based product terms or stratification to examine effect modification, selecting variables based on preliminary results (but could possibly include age of disease onset and/or gender). As part of a planned extensive bias analysis, we will multiply impute for missing data and re-run all analyses 167. Multiple imputation inference has three distinct phases, first missing data will be filled to generate 5 complete datasets; second the 5 complete datasets are analyzed using standard procedures; and third the results from 5 datasets are combined for inference. We will use Monte Carlo simulation techniques, based bias formulas and background knowledge, to investigate the potential impact of uncontrolled confounding, measurement error and selection bias on our findings 168,169. In addition to the demographic variables available in the national registers the Danish National Birth Cohort (DNBC) will allow us to perform a sensitivity analysis on risk for autism spectrum outcomes identified in the children of the DNBC (N= 720) influenced by key variables such as maternal smoking and alcohol consumption during pregnancy, maternal pre-pregnancy BMI, history of hypertension and other environmental factors including noise exposures. Power Calculations. We expect exposure prevalence between 1.5% and 4.5% for maternal diabetes. Our estimates of ASD and ADHD are from the existing literature 111. We estimated the power to detect various size odds ratios (OR) for 8087 ASD cases and 1,213,000 controls for the main effects of maternal diabetes using the program Quanto (Version 0.5). At a p-value of 0.05 we have ~87% power to detect a main effect for ASD of 1.2 for an exposure frequency of only 3.5% (see table 1); similarly, we have ~85% power to detect a main effect for ADHD of 1.2 for an exposure frequency of only 3.5% (see table 2). Table 1: Power to detect an odds ratio of an environmental main effect given its prevalence for 8087 ASD cases/1,213,000 controls (based on 2 sided, alpha=0.05) Odds Ratio Expo. Prev > >.99 > >.99 > >.99 > >.99 >.99 Table 2: Power to detect an odds ratio of an environmental main effect given its prevalence for 7671 HKD cases/1,213,000 controls (based on 2 sided, alpha=0.05) Odds Ratio Expo. Prev > >.99 > >.99 > >.99 > >.99 >.99 Strengths and limitations. Our study has a number of strengths. First, we will use only cases that have been confirmed by a medical diagnosis and recorded in the heath registers; and existing literature supports high validity for the outcomes we are assessing (see text above for respective percentages on accuracy). Furthermore, since diagnoses can be found in more than one register, we can identify cases with multiple diagnoses (diagnosis in more than one register) and apply this higher threshold of exposure and outcome status in our sensitivity analyses. Second, we will have access to the entire population, avoiding selection and attrition bias. Third, our design is efficient and our large sample size gives us sufficient statistical power to detect small effect sizes for exposures that are rare. Fourth, in our analyses of maternal diabetes and ADHD we can include maternal history of medical illness diagnosed by a medical professional, which has been a limitation of previous studies. Fifth, due to our record linkage approach, we will not be hampered by differential participation bias or by differential or non-differential recall bias. And finally, a unique advantage of this study is the ability to asses an array of neurodevelopmental outcomes in the same population cohort, minimizing many sources of bias (selection, diagnostic, unmeasured background factors) and aiding the interpretation of these findings. Some important limitations of this study should be acknowledged. First, there will be some missing data (see statistical methods section for our statistical approach to handling missing data) and we will need to impute these values using statistical software packages (though from our previous experience we anticipate the proportion of missing values will be negligible). Imputed values are

7 subject to error and assumptions are required for their imputation, which may not be correct. Second, measurement of maternal pre-pregnancy BMI is only available in the birth register from 1997 onwards and we will only partially adjust for this variable, but will include this in the sensitivity analysis outlined in sub aim 1B. Similarly, we will perform additional sensitivity analyses that will allow us to examine the effects of control of diabetes in Aim IC. Third, it is possible that some cases of gestational diabetes will not be detected and some women will be misclassified as unexposed. Prior to 1991 systematic screening for gestational diabetes was only performed at larger university hospitals though it became more broadly adopted after 2003; and in some cases, gestational diabetes may have been classified as unspecified diabetes in pregnancy. Sixth, Denmark is a homogenous, predominantly Caucasian population and will not allow for betweenethnic and racial analyses that may increase generalizability to the U.S. population; and we will not be able to examine any differences in biologic/genetic vulnerabilities for the neurodevelopmental outcomes that interact with maternal diabetes. Finally, while some validation studies exist on diagnostic accuracy, we have limited information on the validity of other variables used in analysis. Sub Aim (1B): Perform sensitivity analyses of risk variation (unmeasured confounder bias) using the Danish National Birth Cohort (DNBC). The DNBC is nested in the nationwide registry population of Denmark, and has additional extensive risk factor information that includes nutritional and environmental pregnancy exposures that have been collected during pregnancy for ~100,000 women and children 48. One limitation of the analyses outlined in the nationwide population based studies in Aim 1A, is that we will not be able to control for some important confounders such as maternal smoking, alcohol consumption and maternal pre-pregnancy BMI for the entire study period, factors associated with cognitive impairments during pregnancy. However, the DNBC will be used to perform a sensitivity (unmeasured potential confounder) analyses that will measure the association between maternal diabetes, smoking, and alcohol consumption, maternal IQ and the outcomes under investigation. Final output from cox regression models will be adjusted for any associations that have been identified in the sensitivity analyses, and will correct for the bias of unmeasured confounding. Sub Aim (1C): Measure the association between clinical control of diabetes during pregnancy and risk for ASD, ADHD/HKD and IQ. We will obtain information regarding the management of diabetes, specifically OGTT (oral glucose tolerance test) and HbA1c measurements over the course of pregnancy from the Clinical Laboratory Information System (LABKA) (1985 onwards) and the Hospital Register (1997 onwards). In Denmark, women at high risk of gestational diabetes (defined by a diagnosis of gestational diabetes in a previous pregnancy, previous delivery of an infant weighing 4500g, maternal and/or family history of diabetes or obesity defined as BMI 27) undergo an OGTT first around weeks of gestation, and also repeatedly throughout the pregnancy. The main diagnostic criteria for gestational diabetes are based on the Diabetes in Pregnancy Study Group recommendation from , i.e. a fasting capillary whole blood glucose 6.1 mmol/l (corresponding to fasting plasma glucose 7.0 mmol/l) and/or 2h capillary whole blood glucose 9 mmol/l (corresponding to capillary plasma glucose 10 mmol/l and venous plasma >= 9 mmol/l) after a 75g OGTT. Prior to 1991, systematic screening for gestational diabetes was only performed at larger university hospitals though it became more broadly adopted after In some cases gestational diabetes may have been classified as unspecified diabetes in pregnancy and may have also been diagnosed using HbA1c measurements. We will obtain clinical OGTT and HbA1c values for all women where data is available and measure the association between diabetes during pregnancy (a continuous measure of OGTT and HbA1c) and ASD, ADHD/HKD and IQ; employing logistic regression to estimate odds ratios (ASD, autism, HKD); and Pearson correlation coefficients for IQ (continuous outcome). Aim (2): Examine the association between maternal diabetes during pregnancy and neuromotor and behavioral traits in the Danish National Birth Cohort. Utilize the well-established Danish National Birth Cohort of approximately 100,000 pregnant women and their offspring, (follow-up data currently at 7 years), to test the hypothesis that women with maternal diabetes [defined by an ICD code in the patient and hospital register for either pre-existing type 1 diabetes (ICD-8 codes: 249; ICD-10 codes: E10, O24.0), pre-existing type 2 diabetes (ICD-8 codes: 250; ICD-10 codes: E11, O24.1) or diabetes with gestational onset (ICD-10 codes: O24.4, O24.9)] is related to (1) developmental milestones (motor skills development); and (2) parental assessment of strengths and difficulties measured by SDQ (Strengths and Difficulties Questionnaire). In our analyses, we distinguish between type of diabetes (pre-existing type1, type 2 diabetes, and gestational diabetes) and furthermore we will stratify by trimester of diagnosis for those women diagnosed with diabetes during gestation. The DNBC, is a longitudinal population-based cohort of pregnant women and their offspring established between 1996 and Information on study design and data collection methods for the DNBC have been described elsewhere 48. Information was collected twice during

8 pregnancy using computer-assisted telephone interviews, at gestational age 12 and 30 weeks, and again when the children reached 6 and 18 months of age. The DNBC has existing data on approximately 100,000 pregnancies and is completing data collection for the 11 th year of follow-up. Using the 18 month and 7-year follow-up interviews, we will assess the following outcomes: Developmental Milestones. Developmental milestones in infancy were assessed in the 18-month interview using telephone interviews. There were twelve questions pertaining to the developmental milestones, consisting of three questions on gross motor skills (sitting without support by nine months of age, walking without support by 16 months of age, climbing stairs without support); two questions on fine motor skills (ability to take off socks and shoes, drinking from an ordinary cup without assistance); one question regarding child s attention (ability to be occupied alone, ability to make a mark on a table/paper, turning the picture right when looking at a book); and three questions on language (using word like sounds to tell what he/she wants, ability to use 10 or more names/words, two word sentences). Questions were developed by child neuropsychologists, with extensive consideration of age-appropriate developmental scales (interview questionnaires can be accessed at however, their psychometric properties have not been validated. Parental assessment of Strengths and Weaknesses. At the 7-year follow-up, all DNBC participants were asked 25 questions to assess their child s ADHD-like behaviors using the standardized Strength and Difficulties Questionnaire (SDQ). The SDQ is a screening tool that assesses five areas of social-behavioral development that consist of emotional symptoms, conduct problems, hyperactivity, peer relationship and prosocial behavior in children and adolescents ages 4 to ; it has been shown to detect children with psychiatric disorders with reasonable efficiency 113, and also a reliable screening instrument for emotional and behavioral problems in school-age children 114,115. The interview was conducted after the child s 7 th birthday with either the mother or the child s primary caregiver and the questions ascertained behavior from the previous 6 months. Included in this assessment were six parent questions, about their own personal behavior problems during childhood. Properties of the SDQ have been tested in many populations and languages 171 including Nordic countries 115,172,173. Furthermore, a recent study of the Danish SDQ among 71,840 parent and teacher raters of 5, 7, 10 and 12 year old children, which utilized 4 large Danish cohorts one of which was the DNBC, confirmed the internal reliability of the original five factor structure and its usefulness as a screening tool to assess emotional and behavioral problems in children 174. Previously, the SDQ has been employed to evaluate maternal prenatal smoking, drinking and acetaminophen use on children s ADHD-like behaviors Statistical Methods (Developmental Milestones). We will use standard analytical methods for cohort studies 165,166 to examine the associations between prenatal maternal diabetes and developmental milestones. We will employ logistic regression models for all dichotomized variables and ordinal logistic regression for outcomes on the ordinal scale, and cox survival models to compare the age at which the child could sit or walk without support. We will use person-years as the offset variable, which will begin at date of birth and end at the age the milestone was reached. Estimates will be adjusted for maternal age at birth, gender of child, child s birth year, gestational age, birth weight, parity, socio-economic status of mother, maternal smoking and alcohol drinking during pregnancy, maternal pre-pregnancy BMI and history of hypertension, parent s behavioral scores in childhood, and mother's ever had mental health problems. We will also assess child s age at mother s interview, breast-feeding after the child turned 6 months of age, out-of-home day care (i.e., in day nursery or regular child care), as potential confounders. Statistical Methods (SDQ).We will follow the recommendations for scoring the SDQ (Youth in mind and create a total difficulties score (range 0-40) by summing over four subscales (emotional symptoms, conduct problems, hyperactivity, and peer problems) that range from 0-10 each, with higher scores indicating an increasing number of behavioral problems with the exception of the pro-social

9 behavior subscale (range 0-10) for which higher scores indicate positive social behaviors. We will also generate a parental behavioral problems score (range 0-12) of ADHD-like symptoms based on the 2-parent childhood behavior questions. We will use generalized linear models with a log-link function and a Poisson distribution to estimate risk ratios and 95% confidence intervals for prenatal maternal diabetes and ADHD-like behaviors measured from the SDQ 112. Following the advice of Goodman, we will dichotomize SDQ scores using a cut-off point resulting in high specificity for ADHD-like behaviors ( total difficulties scores 17). All models will be adjusted for variables listed above. Strengths and Limitations. This study has a number of strengths. First, the DNBC has collected detailed data early pregnancy data on maternal health status, health behaviors, nutritional and supplemental intake as well as occupational exposures all of which can be examined as potential confounders. Second, the DNBC is a large cohort size, which is necessary since the exposure of interest is not common among most pregnant women. Third, members of the DNBC have aged years allowing researchers to evaluate childhood onset outcomes at the present time. Fourth, measures in the SDQ have been validated in the Danish population and the internal reliability of the original five factor structure and its usefulness as a screening tool to assess emotional and behavioral problems in children has been confirmed. And finally, this highly valuable research can be performed cost effectively since data can be acquired with minimal cost. Some important limitations of the DNBC should be acknowledged. First, the DNBC is a homogenous, predominantly Caucasian population and will not allow for between-ethic analyses. Second, women in the DNBC represent a relatively affluent group of women with full access to health care, conditions that differ from the U.S. Third, it is likely that some cases of diabetes will be undetected during pregnancy and some women misclassified as unexposed. Finally, there will be missing data and we will need to impute these values using statistical software packages, imputed values are subject to error and assumptions are required for their imputation, which may not be correct. Meetings between Investigators (see Plan for Career Development and Training) Final Summary. This proposal builds on almost a decade of previously conducted research on adverse prenatal exposures during pregnancy using the Danish Civil Registration System and Danish medical registers, and outcomes in the offspring. Through existing collaborations with Danish colleagues, we have access to their unique resources of large-scale, nationwide registries which collect data on maternal health conditions, demographic variables and outcomes in the offspring, and have gained expertize in the individual data sources as well as their contents. Denmark has a long-standing tradition of maintaining one of the world's largest and oldest health science information data bank and has provided tax-supported health care for all residents of the country for many decades; conditions not available anywhere in the U.S. Our focus on maternal metabolism and fetal programming of neurocognitive outcomes, will advance what is known about fetal programming effects of maternal diabetes, inform women of child bearing age and physicians to take preventative measures, in addition to informing health officials charged with public policy decisions on maternal nutrition during pregnancy. Furthermore, the training outlined in this application combined with Dr. Virk s prior training (in maternal and child health and epidemiology) will uniquely position Dr. Virk to make innovative scientific advancements in the field of fetal neurodevelopmental epidemiology. Project Timeline. Year 1: K99 Year 2: K99 Year 3: R00 Year 4: R00 Year 5: R00 (1) Data linkage and data analysis outlined in aim 1 and subaim 1A; (2) Coursework: Applied Genetic Modeling (Winter M237), Advanced Human Gen. & Genomic Technologies (Spring HG2244); (3) Training in CART biweekly seminars, journal club, and annual retreat; (4) Mentor meetings (Virk, Olsen, Ritz): Spring meeting in Los Angeles, Fall meeting in Denmark; (5) Conference meetings: EuroEpi/EBBS (Euro Brain & Behaviour Society), SPER/SER conference. (1) Data linkage and data analysis outlined in subaim 1B; (2) Coursework: Research Topics in Neurobehav. Genetics (Winter NS215), Phenotypic Measurement of Complex Traits (Spring NS240); (3) Training in CART biweekly seminars, journal club, and annual retreat; (4) Mentor meetings (Virk, Olsen, Ritz): Spring meeting in Los Angeles, Fall meeting in Denmark; (5) Conference meetings: EuroEpi/EBBS, SPER/SER conferences; (6) Apply to new institutions for R00 period. (1) Complete manuscripts from aim 1, subaim 1A and 1B; (2) Begin data analysis outlined in aim 1C; (3) Mentor meetings (Virk, Olsen, Ritz): Spring meeting in Los Angeles, Fall meeting in Denmark; (4) Conference meetings: EuroEpi/EBBS (Euro Brain & Behaviour Society), SPER/SER conference. (1) Complete data analysis and manuscripts that correspond to aim 1C; (2) Begin data analysis outlined in aim 2; (3) Preparation of first R01 submission; (3) Mentor meetings (Virk, Olsen, Ritz): Spring meeting in Los Angeles, Fall meeting in Denmark; (4) Conference meetings: EuroEpi/EBBS (Euro Brain & Behaviour Society), SPER/SER conference. (1) Complete data analysis and manuscripts that correspond to aim 2; (2) Preparation of second R01 submission; (3) Mentor meetings (Virk, Olsen, Ritz): Spring meeting in Los Angeles, Fall meeting in Denmark; (4) Conference meetings: EuroEpi/EBBS (Euro Brain & Behaviour Society), SPER/SER conference.