This meeting will have a medical flavour, with two sessions each of two speakers. The programme for the meeting, including abstracts of the papers being presented, is given below.
Session 1: Life course modelling | |
| 2:00 p.m. | Bianca L De Stavola (London School of Hygiene and Tropical Medicine) |
|---|---|
| Statistical issues in life course epidemiology | |
| 2:45 p.m. | Daniel S. Nagin (Carnegie Mellon University and University of Cambridge) |
| Analyzing developmental trajectories: a semiparametric, group-based approach | |
| 3:30 p.m. | Tea |
Session 2: HRT | |
| 3:50 p.m. | David Purdie (Centre for Metabolic Bone Disease, Hull) |
| Statistics and the media - a clinician's viewpoint | |
| 4:35 p.m. | Stephen Evans (London School of Hygiene and Tropical Medicine) |
| Randomised trials and observational tribulations - lessons from HRT | |
The focus of chronic disease epidemiology has recently moved from adult risk factors, such as smoking or obesity, to pre-natal and childhood exposures, birth weight possibly being their most recognisable example. The study of how these temporally ordered factors, which often include both biological and social dimensions, jointly influence the disease of interest is currently referred to as life course epidemiology.
The adoption of a life course approach is not novel, indeed it is well known in the study of behaviour and of psychiatry disorders. However it is met by specific difficulties when applied to chronic disease epidemiology. The talk will describe them, and in particular focus on the availability, consistency and completeness of data that often span over 50 years, and on the intrinsic correlations among the biological and social early life factors of interest (e.g. childhood growth, nutrition and puberty). Some possible strategies of analysis will be illustrated and then discussed in terms of both accessibility to non-technical users and the insights they potentially offer. The strategies include multiple imputations of observations missing at random, latent variables for capturing the underlying correlations, sensitivity analyses to assess the impact of losses to follow-up and structural equation models to jointly study the exposures and disease processes.
Daniel S. Nagin (Carnegie Mellon University and University of Cambridge)
Analyzing developmental trajectories:
a semiparametric, group-based approach
A developmental trajectory describes the course of a behavior over age or time. The aim of this lecture is to provide an overview of a method, based on finite mixture modeling, that colleagues and I have developed for analyzing developmental trajectories. The method provides the capability to (1) identify rather than assume distinctive groups of trajectories, (2) estimate the proportion of the population following each such trajectory group, and (3) relate group membership probability to individual characteristics and circumstances. In addition, two important extensions of the method will be presented - the capability to add time-varying covariates to trajectory models and the capability to estimate joint trajectory models of distinct but related behaviors. The former provides the statistical capacity for testing whether a contemporaneous factor, such as an experimental intervention or a non-experimental event like pregnancy, deflects a pre-existing trajectory. The latter provides the capability to study the unfolding of distinct but related behaviors such as childhood problem behavior and adolescent drug abuse.
David Purdie (Centre for Metabolic Bone Disease, Hull)
Statistics and the media - a clinician's viewpoint
The reporting, in July 2002, of data from the Women's Health Initiative (WHI) Trial of oestrogen with and without progestogen, has illustrated the problems faced by physicians in the wake of serious inaccuracies in coverage of data releases by the print and broadcast media. The guidelines for scientific and medical reportage, issued by the Royal Society in collaboration with the media, seem to be being regularly ignored. Perhaps due to the innate tensions between journalists writing copy and Sub-editors writing headlines, the essential balance between valid conclusion and appropriate headline are often lost in the need to attract attention, dramatise findings and secure sales. For example, the regular use of percentage values to inflate small absolute differences - without the absolute numbers themselves being reported - resulted, in the WHI coverage, in large discontinuations of valid treatment by alarmed patients.
This paper will take the WHI trial reports as an example and will set out the press reports, the actual published data, and the implications for practice. Observations from delegates on the issue of how complex trial results should be set before the general public will be welcome.
Stephen Evans (London School of Hygiene and Tropical Medicine)
Randomised trials and observational tribulations - lessons from HRT
Hormone Replacement Therapy (HRT) is licensed for the relief of menopausal symptoms (and in some cases prevention of osteoporosis), but neither in the UK nor the USA has it been licensed for prevention of cardiovascular disease. Randomised trials (RCTs) that are used for licensing of medicines are generally relatively small. New benefits and new harms may be discovered when a product is in widespread use, but post-marketing studies of medicines are rarely done systematically and, as they are usually observational, may be subject to bias and confounding.
Adverse effects of HRT in terms of breast and endometrial cancer and venous thromboembolism (VTE) were identified relatively early in observational studies. Major observational studies of these adverse effects were produced in the late 1990s, but some authorities (medical and governmental) were unconvinced that the associations were causal.
Many observational studies have found reduced risks of coronary heart disease but not everyone believed these either. There was a slight tendency for believers in cardio-vascular benefits to be disbelievers in the adverse effects.
Recent RCTs of combined HRT against placebo have confirmed the increased risks of VTE and breast cancer but have found no evidence of any cardiovascular benefit, and increased risk of stroke. In contrast, reductions in fractures and colo-rectal cancer found in observational studies have been replicated in trials. The balancing of risks and benefits is complex and strongly related to duration of treatment. An overview of the history and current knowledge regarding HRT will be given. Some suggestions as to when observational studies can and cannot be trusted will be given.
Page maintained by R. A. Bailey
Page updated 19/2/03