John B Carlin

University of Melbourne, Melbourne, Victoria, Australia

Are you John B Carlin?

Claim your profile

Publications (324)1300.79 Total impact

  • [show abstract] [hide abstract]
    ABSTRACT: The aim of this study was to describe the distribution of magnetic resonance imaging (MRI) patterns in a large population sample of children with cerebral palsy (CP) and to examine associations between MRI patterns, and antenatal and perinatal variables. Data were retrieved from the Victorian CP Register for 884 children (527 males, 357 females) born between 1999 and 2006. Postneonatal MRI was classified for 594 children. For 563 children (329 males, 234 females) for whom classification was to a single MRI pattern, the frequency of each variable was compared between patterns and with the population frequency. White matter injury was the most common MRI pattern (45%), followed by grey matter injury (14%), normal imaging (13%), malformations (10%), focal vascular insults (9%), and miscellaneous patterns (7%). Parity, birth gestation, level of neonatal care, Apgar score, and time to established respiration varied between MRI patterns (p<0.01). Nulliparity was most strongly associated with focal vascular insults, whereas multiparity was associated only with malformations. Grey matter injury was not associated with birth in a tertiary unit, but was strongly associated with severe perinatal compromise. The frequency of neonatal seizures and of nursery admissions was lowest among children with malformations. As known risk factors for CP are differentially associated with specific MRI patterns, future exploration of causal pathways might be facilitated when performed in pathogenically defined groups.
    Developmental Medicine & Child Neurology 02/2014; 56(2):178-84. · 2.68 Impact Factor
  • [show abstract] [hide abstract]
    ABSTRACT: Most adults with common mental disorders report their first symptoms before 24 years of age. Although adolescent anxiety and depression are frequent, little clarity exists about which syndromes persist into adulthood or resolve before then. In this report, we aim to describe the patterns and predictors of persistence into adulthood. We recruited a stratified, random sample of 1943 adolescents from 44 secondary schools across the state of Victoria, Australia. Between August, 1992, and January, 2008, we assessed common mental disorder at five points in adolescence and three in young adulthood, commencing at a mean age of 15·5 years and ending at a mean age of 29·1 years. Adolescent disorders were defined on the Revised Clinical Interview Schedule (CIS-R) at five adolescent measurement points, with a primary cutoff score of 12 or higher representing a level at which a family doctor would be concerned. Secondary analyses addressed more severe disorders at a cutoff of 18 or higher. 236 of 821 (29%; 95% CI 25-32) male participants and 498 of 929 (54%; 51-57) female participants reported high symptoms on the CIS-R (≥12) at least once during adolescence. Almost 60% (434/734) went on to report a further episode as a young adult. However, for adolescents with one episode of less than 6 months duration, just over half had no further common mental health disorder as a young adult. Longer duration of mental health disorders in adolescence was the strongest predictor of clear-cut young adult disorder (odds ratio [OR] for persistent young adult disorder vs none 3·16, 95% CI 1·86-5·37). Girls (2·12, 1·29-3·48) and adolescents with a background of parental separation or divorce (1·62, 1·03-2·53) also had a greater likelihood of having ongoing disorder into young adulthood than did those without such a background. Rates of adolescent onset disorder dropped sharply by the late 20s (0·57, 0·45-0·73), suggesting a further resolution for many patients whose symptoms had persisted into the early 20s. Episodes of adolescent mental disorder often precede mental disorders in young adults. However, many such disorders, especially when brief in duration, are limited to the teenage years, with further symptom remission common in the late 20s. The resolution of many adolescent disorders gives reason for optimism that interventions that shorten the duration of episodes could prevent much morbidity later in life. Australia's National Health and Medical Research Council.
    The Lancet 01/2014; · 39.06 Impact Factor
  • Source
    [show abstract] [hide abstract]
    ABSTRACT: Multiple imputation (MI) is becoming increasingly popular as a strategy for handling missing data, but there is a scarcity of tools for checking the adequacy of imputation models. The Kolmogorov-Smirnov (KS) test has been identified as a potential diagnostic method for assessing whether the distribution of imputed data deviates substantially from that of the observed data. The aim of this study was to evaluate the performance of the KS test as an imputation diagnostic. Using simulation, we examined whether the KS test could reliably identify departures from assumptions made in the imputation model. To do this we examined how the p-values from the KS test behaved when skewed and heavy-tailed data were imputed using a normal imputation model. We varied the amount of missing data, the missing data models and the amount of skewness, and evaluated the performance of KS test in diagnosing issues with the imputation models under these different scenarios. The KS test was able to flag differences between the observations and imputed values; however, these differences did not always correspond to problems with MI inference for the regression parameter of interest. When there was a strong missing at random dependency, the KS p-values were very small, regardless of whether or not the MI estimates were biased; so that the KS test was not able to discriminate between imputed variables that required further investigation, and those that did not. The p-values were also sensitive to sample size and the proportion of missing data, adding to the challenge of interpreting the results from the KS test. Given our study results, it is difficult to establish guidelines or recommendations for using the KS test as a diagnostic tool for MI. The investigation of other imputation diagnostics and their incorporation into statistical software are important areas for future research.
    BMC Medical Research Methodology 11/2013; 13(1):144. · 2.21 Impact Factor
  • Source
    [show abstract] [hide abstract]
    ABSTRACT: Background Language impairment (LI) in the preschool years is known to vary over time. Stability in the diagnosis of LI may be influenced by children's individual variability, the measurement error of commonly used assessment instruments and the cut-points used to define impairment. AimsTo investigate the agreement between two different age-based versions of a language assessment instrument and the stability of the classification of LI using the two measures over a 12-month period. Methods & ProceduresA total of 945 participants completed the Clinical Evaluation of Language Fundamentals (CELF—Preschool 2 or 4th Edn) at 4 and 5 years of age. Agreement and stability were analysed using Bland–Altman plots, correlation and odds ratios. Sensitivity and specificity were calculated for two thresholds of the CELF-P2 using the diagnostic category on the child's subsequent CELF-4. Outcomes & ResultsFor all CELF scores, mean differences for the cohort between 4 and 5 years were within 1.5 scale score units. In contrast, at the individual level variability was found across the range of scores and was of a greater magnitude than previously reported. Stability in LI classification was low, with 36% of 5-year-olds with LI (defined as a standard score below –1.25) classified as typical at 4 years, even though odds ratios calculated from classifications at the two time points suggested that 4-year-olds with LI had 23 times greater odds than their typical peers to receive a diagnosis of LI at 5 years. The CELF-P2 did not demonstrate adequate levels of diagnostic accuracy for LI at 5 years: sensitivity of 64% and specificity of 92.9%. Conclusions Substantial variability across the entire range of possible CELF scores was observed in this community cohort between the ages of 4 and 5 years. The stability of LI classification was lower than that reported in previous research conducted primarily on smaller clinical cohorts. The current study's results suggest that the variability observed in developmental language pathways is the result of a combination of limitations in measurement instruments, individual children's abilities and the arbitrary nature of the boundaries defining LI.
    International Journal of Language & Communication Disorders 10/2013; · 1.44 Impact Factor
  • [show abstract] [hide abstract]
    ABSTRACT: Background. Estimates of the risk of intussusception (IS) associated with currently licensed rotavirus vaccines (RV1, Rotarix®, GSK, and RV5, RotaTeq®, Merck) diverge. Contemporaneous introduction of both vaccines in Australia enabled a population-based assessment of risk. Methods. Confirmed cases of IS in infants 1-<12 months were identified from national hospitalisation databases, supplemented by active hospital-based surveillance from July 2007-June 2010. Vaccination histories were verified by the Australian Childhood Immunisation Register, which was also used to identify age-matched controls. Self-controlled case-series and case-control methods were used to assess risk of IS associated with both vaccines in pre-specified periods post-vaccination. The estimated burden of vaccine-attributable IS was compared with estimated reductions in gastroenteritis hospitalisations. Results. Based on 306 confirmed cases of IS, the relative incidence (RI) of intussusception in the 1-7 day period following the first vaccine dose, was 6.8 (95% confidence interval [CI] 2.4 to 19.0, p<0.001) for RV1, and 9.9 (95% CI 3.7 to 26.4, p<0.001) for RV5. There was a smaller increased risk 1-7 days following the second dose of each vaccine. The case-control analysis gave similar results. We estimate an excess of 18 IS cases and >6,500 fewer gastroenteritis hospitalisations in young children annually in Australia following vaccine introduction. Conclusions. We found a similarly increased risk of IS following both vaccines but the balance of benefits and risks at population level was highly favourable, a finding likely to extend to other settings despite varying incidence of IS and potentially higher morbidity and mortality from both gastroenteritis and intussusception.
    Clinical Infectious Diseases 08/2013; · 9.37 Impact Factor
  • [show abstract] [hide abstract]
    ABSTRACT: The aim of this study was to review the distribution of neuroimaging findings from a contemporary population cohort of individuals with cerebral palsy (CP) and to facilitate standardization of imaging classification. Publications from 1995 to 2012 reporting imaging findings in population cohorts were selected through a literature search, and review of the titles, abstracts, and content of studies. Relevant data were extracted, including unpublished data from Victoria, Australia. The proportions for each imaging pattern were tabulated, and heterogeneity was assessed for all individuals with CP, and for subgroups based on gestational age, CP subtype, and Gross Motor Function Classification System level. Studies from three geographic regions met the inclusion criteria for individuals with CP, and two additional studies reported on specific CP subtypes. Brain abnormalities were observed in 86% of scans, but were observed least often in children with ataxia (24-57%). White matter injury was the most common imaging pattern (19-45%), although the proportions showed high heterogeneity. Additional patterns were grey matter injury (21%), focal vascular insults (10%), malformations (11%), and miscellaneous findings (4-22%). This review suggests areas where further dialogue will facilitate progress towards standardization of neuroimaging classification. Standardization will enable future collaborations aimed at exploring the relationships among magnetic resonance imaging patterns, risk factors, and clinical outcomes, and, ultimately, lead to better understanding of causal pathways and opportunities for prevention.
    Developmental Medicine & Child Neurology 08/2013; · 2.68 Impact Factor
  • Katherine J Lee, John B Carlin
    [show abstract] [hide abstract]
    ABSTRACT: The self-controlled case series method is a statistical approach to investigating associations between acute outcomes and transient exposures. The method uses cases only and compares time at risk after the transient exposure with time at risk outside the exposure period within an individual, using conditional Poisson regression. The risk of outcome and exposure often varies over time, for example, with age, and it is important to allow for such time dependence within the analysis. The standard approach for modelling time-varying covariates is to split observation periods into blocks according to categories of the covariate and then to model the relationship using indicators for each category. However, this can be inefficient and can lead to problems with collinearity if the exposure occurs at approximately the same time in all individuals. As an alternative, we propose using fractional polynomials to model the relationship between the time-varying covariate and incidence of the outcome. We present the results from an analysis exploring the association between rotavirus vaccination and intussusception risk as well as a simulation study. We conclude that fractional polynomials provide a useful approach to adjusting for time-varying covariates but that it is important to explore the sensitivity of the results to the number of categories and the method of adjustment. Copyright © 2013 John Wiley & Sons, Ltd.
    Statistics in Medicine 07/2013; · 2.04 Impact Factor
  • [show abstract] [hide abstract]
    ABSTRACT: The concept of missing at random is central in the literature on statistical analysis with missing data. In general, inference using incomplete data should be based not only on observed data values but should also take account of the pattern of missing values. However, it is often said that if data are missing at random, valid inference using likelihood approaches (including Bayesian) can be obtained ignoring the missingness mechanism. Unfortunately, the term "missing at random" has been used inconsistently and not always clearly; there has also been a lack of clarity around the meaning of "valid inference using likelihood". These issues have created potential for confusion about the exact conditions under which the missingness mechanism can be ignored, and perhaps fed confusion around the meaning of "analysis ignoring the missingness mechanism". Here we provide standardised precise definitions of "missing at random" and "missing completely at random", in order to promote unification of the theory. Using these definitions we clarify the conditions that suffice for "valid inference" to be obtained under a variety of inferential paradigms.
    Statistical Science 06/2013; 28(2). · 2.24 Impact Factor
  • [show abstract] [hide abstract]
    ABSTRACT: To evaluate the effectiveness of school-based physical activity interventions on fitness, adiposity and cardiometabolic outcomes among schoolchildren. Medline, Embase, EBSCOhost CINAHL and ERIC databases were searched up to October 2012. Inclusion criteria: intervention delivered at school with controls having no intervention or usual physical education classes; participants aged 5-18 years; outcomes spanning some or all of the above. We assessed levels of evidence for identified trials based on methodological quality and sample size. Dose of the interventions (a total summary measure of intensity, frequency and duration) were considered. Eighteen randomized controlled trials (RCTs, total participants = 6,207) were included, of which six were large, higher quality trials with high dose of the intervention. The intervention was consistent in increasing fitness with large, higher quality studies and high dose of intervention providing strong evidence. Dose of school-based physical activity is an important determinant of trial efficiency. Some large, higher quality RCTs provided strong evidence for interventions to decrease skin-fold thickness, increase fitness and high-density lipoprotein cholesterol. Evidence for body mass index, body fat and waist circumference, blood pressure and triglycerides, low-density lipoprotein cholesterol and total cholesterol remain inconclusive and require additional higher quality studies with high dose of interventions to provide conclusive evidence.
    Obesity Reviews 06/2013; · 6.87 Impact Factor
  • [show abstract] [hide abstract]
    ABSTRACT: The concept of missing at random is central in the literature on statistical analysis with missing data. In general, inference using incomplete data should be based not only on observed data values but should also take account of the pattern of missing values. However, it is often said that if data are missing at random, valid inference using likelihood approaches (including Bayesian) can be obtained ignoring the missingness mechanism. Unfortunately, the term "missing at random" has been used inconsistently and not always clearly; there has also been a lack of clarity around the meaning of "valid inference using likelihood". These issues have created potential for confusion about the exact conditions under which the missingness mechanism can be ignored, and perhaps fed confusion around the meaning of "analysis ignoring the missingness mechanism". Here we provide standardised precise definitions of "missing at random" and "missing completely at random", in order to promote unification of the theory. Using these definitions we clarify the conditions that suffice for "valid inference" to be obtained under a variety of inferential paradigms.
  • Source
    [show abstract] [hide abstract]
    ABSTRACT: Background: Preterm infants ≤32 weeks' gestation are increasingly being managed on continuous positive airway pressure (CPAP), without prior intubation and surfactant therapy. Some infants treated in this way ultimately fail on CPAP and require intubation and ventilation. Objectives: To define the incidence, predictors and consequences of CPAP failure in preterm infants managed with CPAP from the outset. Methods: Preterm infants 25-32 weeks' gestation were included in the study if inborn and managed with CPAP as the initial respiratory support, with division into two gestation ranges and grouping according to whether they were successfully managed on CPAP (CPAP-S) or failed on CPAP and required intubation <72 h (CPAP-F). Predictors of CPAP failure were sought, and outcomes compared between the groups. Results: 297 infants received CPAP, of which 65 (22%) failed, with CPAP failure being more likely at lower gestational age. Most infants failing CPAP had moderate or severe respiratory distress syndrome radiologically. In multivariate analysis, CPAP failure was found to be predicted by the highest FiO2 in the first hours of life. CPAP-F infants had a prolonged need for respiratory support and oxygen therapy, and a higher risk of death or bronchopulmonary dysplasia at 25-28 weeks' gestation (CPAP-F 53% vs. CPAP-S 14%, relative risk 3.8, 95% CI 1.6, 9.3) and a substantially higher risk of pneumothorax at 29-32 weeks. Conclusion: CPAP failure in preterm infants usually occurs because of unremitting respiratory distress syndrome, is predicted by an FiO2 ≥0.3 in the first hours of life, and is associated with adverse outcomes.
    Neonatology 04/2013; 104(1):8-14. · 2.57 Impact Factor
  • Source
    [show abstract] [hide abstract]
    ABSTRACT: PURPOSE: Early puberty has been linked to higher rates of mental health problems in adolescence. However, previous studies commencing after the initiation of puberty have been unable to explore whether early puberty is preceded by higher rates of these problems. In a large national study, we aimed to determine whether difficulties in behavior and psychosocial adjustment are evident before as well as during the early pubertal transition. METHODS: The Longitudinal Study of Australian Children recruited a nationally representative cohort of 4,983 children at age 4-5 years in 2004. This analysis includes 3,491 of these children (70.1%) followed up at ages 6-7, 8-9, and 10-11 years, with a completed parent report of stage of pubertal maturation at age 8-9 years. Parents reported behavior difficulties (Strengths and Difficulties Questionnaire) and psychosocial adjustment (Pediatric Quality of Life Inventory) at all four waves from ages 4-5 to 10-11 years. RESULTS: Both boys and girls who entered puberty early (i.e., by age 8-9 years) also experienced poorer psychosocial adjustment at this age. These psychosocial differences were already evident at ages 4-5 and 6-7 years, and persisted to at least age 10-11 years. Similar patterns were evident for behavior difficulties, but only for boys; early puberty was not related to behavior difficulties in girls. CONCLUSIONS: Children with early puberty have different patterns of behavior and social adjustment from the preschool years through early adolescence. At least in part, the association between early-onset puberty and poor mental health appears to result from processes under way well before the onset of puberty.
    Journal of Adolescent Health 03/2013; · 2.97 Impact Factor
  • Source
    [show abstract] [hide abstract]
    ABSTRACT: AIM: To explore the link between pre-school children's general home computer use and their letter knowledge. METHODS: As part of the Early Language in Victoria Study, a community cohort of 1539 four-year-old children was tested on letter knowledge as well as on non-verbal intelligence, oral language, articulation and phonological awareness. Performance on these measures was examined in relation to parent-questionnaire responses exploring home literacy environment and the amount of time children spent using the computer. RESULTS: A positive correlation between computer use and letter knowledge was found, and this association was still evident after controlling for other cognitive and environmental factors known to predict the development of letter knowledge in young children. CONCLUSIONS: Greater computer use in pre-school children appears to have a positive association with emerging literacy development. Future research needs to examine the nature of that association.
    Journal of Paediatrics and Child Health 02/2013; · 1.25 Impact Factor
  • Source
    [show abstract] [hide abstract]
    ABSTRACT: BACKGROUND: Newborn screening allows novel treatments for cystic fibrosis (CF) to be trialled in early childhood before irreversible lung injury occurs. As respiratory exacerbations are a potential trial outcome variable, we determined their rate, duration and clinical features in preschool children with CF; and whether they were associated with growth, lung structure and function at age 5 years. METHODS: Respiratory exacerbations were recorded prospectively in Australasian CF Bronchoalveolar Lavage trial subjects from enrolment after newborn screening to age 5 years, when all participants underwent clinical assessment, chest CT scans and spirometry. RESULTS: 168 children (88 boys) experienced 2080 exacerbations, at an average rate of 3.66 exacerbations per person-year; 80.1% were community managed and 19.9% required hospital admission. There was an average increase in exacerbation rate of 9% (95% CI 4% to 14%; p<0.001) per year of age. Exacerbation rate differed by site (p<0.001) and was 26% lower (95% CI 12% to 38%) in children receiving 12 months of prophylactic antibiotics. The rate of exacerbations in the first 2 years was associated with reduced forced expiratory volume in 1 s z scores. Ever having a hospital-managed exacerbation was associated with bronchiectasis (OR 2.67, 95% CI 1.13 to 6.31) in chest CT scans, and lower weight z scores at 5 years of age (coefficient -0.39, 95% CI -0.74 to -0.05). CONCLUSIONS: Respiratory exacerbations in young children are markers for progressive CF lung disease and are potential trial outcome measures for novel treatments in this age group.
    Thorax 01/2013; · 8.38 Impact Factor
  • Source
    [show abstract] [hide abstract]
    ABSTRACT: Missing data often cause problems in longitudinal cohort studies with repeated follow-up waves. Research in this area has focussed on analyses with missing data in repeated measures of the outcome, from which participants with missing exposure data are typically excluded. We performed a simulation study to compare complete-case analysis with Multiple imputation (MI) for dealing with missing data in an analysis of the association of waist circumference, measured at two waves, and the risk of colorectal cancer (a completely observed outcome). We generated 1,000 datasets of 41,476 individuals with values of waist circumference at waves 1 and 2 and times to the events of colorectal cancer and death to resemble the distributions of the data from the Melbourne Collaborative Cohort Study. Three proportions of missing data (15, 30 and 50%) were imposed on waist circumference at wave 2 using three missing data mechanisms: Missing Completely at Random (MCAR), and a realistic and a more extreme covariate-dependent Missing at Random (MAR) scenarios. We assessed the impact of missing data on two epidemiological analyses: 1) the association between change in waist circumference between waves 1 and 2 and the risk of colorectal cancer, adjusted for waist circumference at wave 1; and 2) the association between waist circumference at wave 2 and the risk of colorectal cancer, not adjusted for waist circumference at wave 1. We observed very little bias for complete-case analysis or MI under all missing data scenarios, and the resulting coverage of interval estimates was near the nominal 95% level. MI showed gains in precision when waist circumference was included as a strong auxiliary variable in the imputation model. This simulation study, based on data from a longitudinal cohort study, demonstrates that there is little gain in performing MI compared to a complete-case analysis in the presence of up to 50% missing data for the exposure of interest when the data are MCAR, or missing dependent on covariates. MI will result in some gain in precision if a strong auxiliary variable that is not in the analysis model is included in the imputation model.
    Emerging Themes in Epidemiology 01/2013; 10(1):6. · 2.59 Impact Factor
  • [show abstract] [hide abstract]
    ABSTRACT: To examine the prevalence of binge drinking in adolescence and its persistence into adulthood in an Australian cohort. 15-year prospective cohort study. Victoria, Australia. 1943 adolescents were recruited from secondary schools at age 14-15 years. Levels of past-week 'binge' drinking (5+ standard drinks on a day, each 10 g alcohol) and 'heavy binge' drinking (20+ standard drinks on a day for males, 11+ for females) were assessed during six adolescent waves, and across three adult waves up to age 29 years. Half of the males (52%) and a third of the females (34%) reported past-week adolescent binge drinking. 90% of male and 70% of female adolescent-onset binge drinkers continued to binge in young adulthood; 70% of males and 48% of females who were not adolescent-onset binge drinkers reported young adult binge drinking. Past-week heavy bingeing was less common in adolescence than adulthood. Overall, 35% of the sample (95% CI 33% to 38%) reported past-week binge drinking in adolescence and young adulthood and one-third (33%; 30% to 35%) first reported binge drinking in young adulthood; only 7% of the sample (6-8%) had binge drinking in adolescence but not young adulthood. 'Heavy binge' drinking occurred in adolescence and young adulthood for 9% (8% to 10%); 8% (7% to 10%) reported it in adolescence but no longer in young adulthood; and 24% (22% to 26%) began 'heavy binge' drinking in young adulthood. Among adolescent binge drinkers (n=821), young adult binge and heavy binge drinking were predicted by being male, adolescent antisocial behaviour and adverse consequences of drinking in adolescence. Binge alcohol use is common and persistent among young Australians. Efforts to prevent the onset of binge drinking during adolescence may substantially reduce harmful patterns of alcohol use in young adulthood.
    BMJ Open 01/2013; 3(8):e003015. · 1.58 Impact Factor
  • [show abstract] [hide abstract]
    ABSTRACT: Lung lavage using two aliquots of 15 mL/kg of dilute surfactant was performed in 30 ventilated infants with severe meconium aspiration syndrome (MAS). Mean recovery of instilled lavage fluid was 46%, with greater fluid return associated with lower mean airway pressure at 24 h and a shorter duration of respiratory support. CONCLUSION: Recovery of instilled lavage fluid is paramount in effective lung lavage in MAS and must be afforded priority in the lavage technique.
    Acta Paediatrica 11/2012; · 1.97 Impact Factor
  • [show abstract] [hide abstract]
    ABSTRACT: Context:The benefits of high serum levels of 25-hydroxyvitamin D [25(OH)D] are unclear. Trials are needed to establish an appropriate evidence base.Objective:We plan to conduct a large-scale trial of vitamin D supplementation for the reduction of cancer incidence and overall mortality and report here the methods and results of a pilot trial established to inform its design.Design:Pilot D-Health was a randomized trial carried out in a general community setting with 12 months intervention and follow-up.Participants:Participants were 60- to 84-yr-old residents of one of the four eastern Australian states who did not have any vitamin D-related disorders and who were not taking more than 400 IU supplementary vitamin D per day. A total of 644 participants were randomized, and 615 completed the study (two persons withdrew because of nonserious adverse events).Interventions:The interventions were monthly doses of placebo or 30,000 or 60,000 IU vitamin D(3).Main Outcomes:The main outcomes were the recruitment rate and changes in serum 25(OH)D.Results:Ten percent of those approached were recruited. At baseline, the mean 25(OH)D was 42 nmol/liter in all three study arms. The mean change in 25(OH)D in the placebo group was 0.12 nmol/liter, compared with changes of 22 and 36 nmol/liter in the 30,000- and 60,000-IU groups, respectively.Conclusions:The D-Health pilot has shown that a large trial is feasible in Australia and that a dose of 2000 IU/d will be needed to ensure that a large proportion of the population reaches the target serum 25(OH)D level.
    The Journal of clinical endocrinology and metabolism 10/2012; · 6.50 Impact Factor
  • [show abstract] [hide abstract]
    ABSTRACT: BACKGROUND: Multiple imputation is becoming increasingly popular for handling missing data, with Markov chain Monte Carlo assuming multivariate normality (MVN) a commonly used approach. Imputing categorical variables (which are clearly non-normal) using MVN imputation is challenging, and several approaches have been suggested. However, it remains unclear which approach should be preferred. METHODS: We explore methods for imputing ordinal variables using MVN imputation, including imputing as a continuous variable and as a set of indicators, and various methods for assigning imputed values to the possible categories (rounding), for estimating a non-linear association between an ordinal exposure and binary outcome. We introduce a new approach where we impute as continuous and assign imputed values into categories based on the mean indicators imputed in a separate round of imputation. We compare these approaches in a simple setting where we make 50% of data in an ordinal exposure missing completely at random, within an otherwise complete real dataset. RESULTS: Methods that impute the ordinal exposure as continuous distorted the non-linear exposure-outcome association by biasing the relationship towards linearity irrespective of the rounding method. In contrast, imputing using indicators preserved the non-linear association but not the marginal distribution of the ordinal variable. CONCLUSIONS: Imputing ordinal variables as continuous can bias the estimation of the exposure-outcome association in the presence of non-linear relationships. Further work is needed to develop optimal methods for handling ordinal (and nominal) variables when using MVN imputation. Copyright © 2012 John Wiley & Sons, Ltd.
    Statistics in Medicine 07/2012; · 2.04 Impact Factor
  • Source
    [show abstract] [hide abstract]
    ABSTRACT: Retaining participants in cohort studies with multiple follow-up waves is difficult. Commonly, researchers are faced with the problem of missing data, which may introduce biased results as well as a loss of statistical power and precision. The STROBE guidelines von Elm et al. (Lancet, 370:1453-1457, 2007); Vandenbroucke et al. (PLoS Med, 4:e297, 2007) and the guidelines proposed by Sterne et al. (BMJ, 338:b2393, 2009) recommend that cohort studies report on the amount of missing data, the reasons for non-participation and non-response, and the method used to handle missing data in the analyses. We have conducted a review of publications from cohort studies in order to document the reporting of missing data for exposure measures and to describe the statistical methods used to account for the missing data. A systematic search of English language papers published from January 2000 to December 2009 was carried out in PubMed. Prospective cohort studies with a sample size greater than 1,000 that analysed data using repeated measures of exposure were included. Among the 82 papers meeting the inclusion criteria, only 35 (43%) reported the amount of missing data according to the suggested guidelines. Sixty-eight papers (83%) described how they dealt with missing data in the analysis. Most of the papers excluded participants with missing data and performed a complete-case analysis (n=54, 66%). Other papers used more sophisticated methods including multiple imputation (n=5) or fully Bayesian modeling (n=1). Methods known to produce biased results were also used, for example, Last Observation Carried Forward (n=7), the missing indicator method (n=1), and mean value substitution (n=3). For the remaining 14 papers, the method used to handle missing data in the analysis was not stated. This review highlights the inconsistent reporting of missing data in cohort studies and the continuing use of inappropriate methods to handle missing data in the analysis. Epidemiological journals should invoke the STROBE guidelines as a framework for authors so that the amount of missing data and how this was accounted for in the analysis is transparent in the reporting of cohort studies.
    BMC Medical Research Methodology 07/2012; 12:96. · 2.21 Impact Factor

Publication Stats

9k Citations
1,300.79 Total Impact Points


  • 1990–2014
    • University of Melbourne
      • • Department of Paediatrics
      • • Centre for International Mental Health
      Melbourne, Victoria, Australia
  • 2013
    • Drug & Alcohol Services South Australia
      Tarndarnya, South Australia, Australia
  • 2000–2013
    • Murdoch Childrens Research Institute
      • • Research Group for Cancer & Epigenetics
      • • Clinical Epidemiology & Biostatistics (CEBU)
      Melbourne, Victoria, Australia
    • Birmingham Children's Hospital NHS Foundation Trust
      Birmingham, England, United Kingdom
  • 1993–2012
    • The Royal Children's Hospital
      • • Centre for Community Child Health
      • • Department of Respiratory Medicine
      • • Centre for Adolescent Health
      • • Department of Cardiology
      • • Clinical Epidemiology and Biostatistics Unit (CEBU)
      • • Department of Neurology
      Melbourne, Victoria, Australia
  • 2011
    • University of Helsinki
      Helsinki, Southern Finland Province, Finland
  • 2007–2011
    • University of New South Wales
      Kensington, New South Wales, Australia
    • University of Queensland 
      • Institute for Molecular Bioscience
      Brisbane, Queensland, Australia
    • National Maternity Hospital
      Dublin, Leinster, Ireland
    • University of Otago
      • Injury Prevention Research Unit
      Dunedin, Otago, New Zealand
  • 2006–2011
    • King's College London
      • • Department of Health Service and Population Research
      • • Institute of Psychiatry
      London, ENG, United Kingdom
    • UK Department of Health
      Londinium, England, United Kingdom
    • University of Adelaide
      • Discipline of Paediatrics
      Adelaide, South Australia, Australia
    • Cancer Council Victoria
      • Centre for Behavioural Research in Cancer
      Melbourne, Victoria, Australia
  • 1994–2011
    • Victoria University Melbourne
      Melbourne, Victoria, Australia
  • 2010
    • The Royal Hobart Hospital
      Hobart Town, Tasmania, Australia
    • University of Cambridge
      • MRC Biostatistics Unit
      Cambridge, ENG, United Kingdom
  • 2009
    • Ministry of Health, Fiji
      Suva City, Central, Fiji
  • 2008
    • Royal Children's Hospital Brisbane
      Brisbane, Queensland, Australia
  • 2000–2008
    • Royal Hospital for Women
      Sydney, New South Wales, Australia
  • 2003
    • Washington University in St. Louis
      • Department of Psychiatry
      San Luis, Missouri, United States
    • Monash University (Australia)
      • Department of Epidemiology and Preventive Medicine
      Melbourne, Victoria, Australia
  • 2002
    • Royal Melbourne Hospital
      Melbourne, Victoria, Australia