Please find below a reminder of the forthcoming joint meeting between the Centre for Biostatistics at The University of Manchester, the Royal Statistical Society Primary Health Care Study Group and the Royal Statistical Society Manchester local group.
For refreshment purposes, to register for this free event please contact Wendy Lamb on [log in to unmask] or +44 (0)161 275 5764
Venue: Manchester Dental Education Centre (MANDEC), Higher Cambridge Street, Manchester, M15 6FH
Theme: Developments and challenges in observational data analysis
14.00 - 14.50: Causal inference for dynamic treatment regimens: how analyses of observational data changed international guidelines on when to start antiretroviral therapy
Professor Jonathan Sterne, Head of School of Social and Community Based Medicine, The University of Bristol
At the beginning of 2010 the US Department of Health and Human Services and the WHO issued revised guidelines recommending that HIV-positive people start antiretroviral therapy (ART) at higher CD4 counts. These changes were made after two large HIV cohort collaborations published analyses using novel statistical methods to deal with time-dependent confounding and lead time bias. The ART Cohort Collaboration (Lancet 2009; 373: 1352-1363) concluded that rates of AIDS and death are reduced when ART is initiated at above 350 cells/mm3, but found little evidence of benefit at higher CD4 thresholds. In contrast, NA-ACCORD (N Engl J Med 2009; 360: 1815-26) found strong evidence of reduced mortality when ART is initiated at above 500 cells/mm3. The use of dynamic marginal structural models in that paper was subsequently criticised, and a recent publication using the same approach (HIV-CAUSAL collaboration, Ann Intern Med. 2011; 154: 509-515) did not replicate these findings. I will describe the methods used in these papers, and discuss possible reasons for differences in their results.
14.50 - 15.40: Causal mediation analysis with multiple causally-ordered mediators
Dr Rhian Daniel, London School of Hygiene and Tropical Medicine
In many fields of empirical research, interest lies in decomposing the effect of an exposure on an outcome into its effect via a number of different pathways. For example, the effect of heavy alcohol consumption on systolic blood pressure (SBP) may be separated into an effect via body mass index (BMI), an effect via gamma-glutamyl transpeptidase (GGT), an effect via both BMI and GGT, and an effect via other pathways (not through BMI or GGT)-often called the direct effect. Much progress has been made, mainly due to contributions from the field of causal inference, in understanding the precise nature of estimands that capture these sorts of effects, the assumptions under which they can be estimated from data, and statistical methods for doing so. However, the focus has been mostly on the decomposition of an effect around and through a single mediator, or a set of mediators considered as a block, hence the two components: a direct and indirect effect. In this talk, we define path-specific estimands that permit the decomposition of the total effect of an exposure on an outcome into a sum of numerous path-specific effects through many mediators, where the mediators are permitted to have a causal effect on each other: a setting we call multiple causally-ordered mediators. We show that there are many ways in which this decomposition can be done, and illustrate these using a real data example on alcohol consumption, SBP, BMI and GGT.
15.40 - 16.00: Tea/coffee break
16.00 - 16.50: Examining associations between gestational weight gain, birthweight and gestational age using multivariate multilevel models
Professor Kate Tilling, The University of Bristol
Gestational weight gain (GWG) is associated with a range of perinatal outcomes and increasing evidence supports associations with longer term cardiovascular and metabolic outcomes in mother and child. Relatively little is known about how pre-pregnancy BMI and weight gain during different periods of gestation may interact, although IOM guidelines allow for greater GWG for women of lower pre-pregnancy BMI. In addition, the direction and nature of relationships between GWG, gestational age and birthweight impact on decisions about confounder adjustment. Greater understanding of these complex relationships would help in identifying subgroups of women at higher risk of adverse outcomes. Random effects models were used to examine GWG in a pregnancy cohort in which women had detailed repeat assessment of weight during pregnancy (median number of measures 10 (IQR: 8,11)) measures. A linear spline model identified three distinct periods of GWG: 4-18weeks, 18-28 weeks and 28+ weeks of gestation. Multivariate multilevel models were used to relate GWG to birthweight and gestational age at birth. Critical period, sensitive period and mobility models were used to investigate the epidemiology of GWG with respect to cardiovascular and cognitive outcomes. Coefficients for the relationship between weight at different gestational ages and outcomes were derived from the random effects covariance matrix. Limitations of these approaches, including assumptions of linearity between the random effects in the trivariate model, and sensitivities to high-order interactions in the second-stage epidemiological models, are discussed.
Dr Richard Emsley
Lecturer in Biostatistics/MRC Fellow
Centre for Biostatistics
Institute of Population Health
The University of Manchester
4.304 Jean McFarlane Building
0161 306 8002
Chairman, Royal Statistical Society Manchester Local Group
You may leave the list at any time by sending the command
to [log in to unmask], leaving the subject line blank.