«Attributable fractions used for partitioning risk and evaluating disease prevention Concepts, estimation, interpretation and presentation Geir Egil ...»
Attributable fractions used for
partitioning risk and evaluating
Concepts, estimation, interpretation and presentation
Geir Egil Eide
Dissertation for the degree doctor philosophiae (dr. philos.)
at the University of Bergen, Bergen, Norway
Geir Egil Eide Attributable fractions
Bergen, Norway 2008
Geir Egil Eide Attributable fractions 3
Most of the work with this thesis has been performed in my full-time position as Biostatistician at the Centre for Clinical Research, Haukeland University Hospital, Bergen, from 1999. The centre has provided all administrative and office facilities, computer resources, and expenses for software, literature and publication as well as most of my travel support. Part of the work has been carried out in my Associate Professor II position in the same period at the Section for Epidemiology and Medical Statistics, Department of Public Health and Primary Health Care, University of Bergen. The University of Bergen has also given financial support for the presentation of results at scientific meetings. The first article was written while I was Associate Professor at the Department of Mathematics and Statistics, Norwegian School of Economics and Business Administration, Bergen (1988-1999) and was also supported by the Lung Epidemiological Research Group, Department of Thoracic Medicine, Haukeland University Hospital (1992-1999). I am also indebted to the Department of Statistics, University of California at Berkeley, USA where I spent a sabbatical year 1995/96.
Geir Egil Eide Attributable fractions 4 Geir Egil Eide Attributable fractions 5 Acknowledgements Completing a doctoral project may in many ways be a lonely ride. However, without continuing support and regular ”drinking holes” manned by enthusiastic colleagues and friends the lone rider would probably perish underway. Their cheering and interest at various stages have been instrumental to the fulfilling of this thesis and I am deeply indebted to all.
An important background for the work herein is my collaboration with Professor Amund Gulsvik and his pulmonary epidemiology research group since 1984. At that time I was at the Section for Medical Informatics and Statistics (SMIS) at the University of Bergen (UoB) and my interest in attributable fractions was first evoked in 1988 by a note from his student, now professor, Per S. Bakke (Appendix I) about how to estimate attributable risks of lung symptoms in the Hordaland Study of Obstructive Lung Disease (HSOLD). Although I changed working place in 1988 to the Norwegian School of Economics and Business Administration (NHH), Amund and his disciples have been the most prominent reason for my continuing involvement in clinical-epidemiological research at the Haukeland University Hospital (HUH) and UoB. Amund was also instrumental in establishing the Centre for Clinical Research (CCR) where I became the first employed person in august
1999. His never-fading faith and encouragement has been crucial for my own belief in fulfilling this thesis, and the data from the first cross-sectional sample of the HSOLD has become the common basis of the examples in four of the five articles included.
Also crucial to the completion of this first article was a young and ambitious lad named Olaf Gefeller whose interest in my work at a conference in Brussels (1991) saved it from falling into oblivion. This was the beginning of our long-lasting cowork and friendship including also his cohabitant and statistician Annette Pfahlberg.
Olaf is now a professor in biometry and epidemiology at the University of Nuremberg, Erlangen and the theme of attributable fractions is only one of his many Geir Egil Eide Attributable fractions 6 specialties. Nevertheless, through the years he has continuously involved me in his activities concerning attributable fraction methodology and has been my most important international co-worker. His more or less anonymous marks can be found on most of the articles.
At NHH I am much indebted to the liberal and methodologically inspiring milieu at the small, now defunct, Institute of Mathematics and Statistics. Teaching statistics and mathematics for large and tiny classes was learning-by-doing the hard way. It was in this period the first article was produced, as I wanted to prove to myself that I could be something more than a statistical consultant and co-author.
As a declared doctoral project my work on attributable fractions was not started until 1999 when I started in the position as biostatistician at the CCR and an associate professor II position at the Section for Epidemiology and Medical Statistics (SEMS, previously SMIS) at the Department of Public Health and Primary Health Care, UoB.
I am grateful to the leaderships of both units for providing the legitimacy for giving priority to work with own methodological research besides serving the statistical needs of the medical researchers in the Western Health Trust (Helse Vest) and at the university.
Also, I am extremely thankful for the effort of Professor Ivar Heuch, my sparring partner and supervisor for Articles II-IV, whose meticulous reading and pertinent work with the many unfinished manuscripts since 1999 has improved the quality immensely. Maybe surprisingly, our cooperation has been almost solely by email, ensuring that every problem posed has been commented and resolved, often in latenight emails at weekends and holidays. Ivar is impressing focussed, on mathematical and linguistic details as well as on the whole.
For the cooperation on the final paper I am in debt to Associate Professor Sven Ove Samuelsen who came up with the proposal of a joint project on survival data in 2001 which suited perfectly to my own ideas of further methodological developments. This has been an interesting and stimulating project pointing into new directions for the Geir Egil Eide Attributable fractions 7 applications of attributable fractions. Sharing musical interests and concerts with Sven Ove has given extra flavour to our friendship and cooperation.
I am further most grateful to my colleagues and friends at the CCR for providing such inspirational and joyful work environment. These include from the beginning Gary J.
Allan, Linda Stoltz Olsvik, Ernst R. Omenaas and Tore Wentzel-Larsen, and later also Reidar C. Thorstensen, Ane Johannessen, Tove A. Hanssen and Sølvi Lerfald.
Special thanks go to Ernst for his constructive and including manner of leadership allowing space for special personalities and working styles.
Any researcher will have to stand on the shoulders of others and I take this opportunity to thank also my statistical mentors through the years, i.e. Trygve S.
Nilsen (my master’s supervisor), Rolv A. Skjærven (a pioneer in medical statistics in Bergen and my boss at SMIS from 1978-88; now professor at SEMS), and Jostein Lillestøl (inspirator and professor at NHH). Others to thank are Valborg Baste (who back in 1986 provided the original logistic regression analyses of the HSOLD data which pervade the thesis), Stein Emil Vollset (who suggested the journal for Article I, was heavily involved in the establishment of CCR and chair of SEMS until 2006), and finally Egil Haugland (former director at Haukeland University Hospital), and R & D Director Grethe S. Tell at Haukeland University Hospital (former head of Department of Public Health and Primary Health Care), both for supporting my conditions of employment in 1999 and later. Thanks go also to my hosts at the Department of Statistics, University of California at Berkeley where I spent a sabbatical year 1995/6, and to the Centre for Advanced Studies, Oslo for providing financial support and excellent working conditions at the Research Group on Statistical Analysis of Complex Event History Analysis (under the leadership of professors Ørnulf Borgan and Odd O. Aalen) for finishing the work with Article V during the fall semester of 2005 and spring 2007. Also, I thank all my statistical and other colleagues through the years for making the field of statistics and academic research such an interesting and fun place to work.
Geir Egil Eide Attributable fractions 8 Last, but not the least, I thank my family for giving my life other dimensions, sharing their love, sticking up with my heavy work load and absenteeism both physically and mentally, giving me a home to come back to, and sharing adventures and experiences of less scientific value but nevertheless invaluable for coping with real life itself.
AFE Attributable fraction in exposed AHF Attributable hazard fraction AFB Attributable fraction before (time t) AFW Attributable fraction within study
HSOLD Hordaland Study of Obstructive Lung Disease HUH Haukeland University Hospital MLE Maximum likelihood estimator NHH Norges handelshøyskole (Norwegian School of Economics and Business Administration) PMVD Proportional marginal variance decomposition
Summary Background In medical research some fundamental tasks are to study potential harmful exposures that may give increased risk of getting some disease, potential beneficial treatments that may increase chance of recovering from a disease, or interventions that may reduce the extent or effect of a harmful exposure. In epidemiologic research these questions are studied by collecting individual data for representative samples of the population. For a specified disease (e.g. breast cancer) there will usually be many risk factors, some may be modifiable (e.g. life style factors like smoking habits, physical activity, dietary factors) and other factors not so easy to modify (like reproductive factors, aging, genetic factors). Provided that enough data for the individuals in the sample is collected on the occurrence of disease and the relevant risk factors, statistical models are identified to estimate the effects of the various risk factors on the prevalence or incidence of the disease in the population. Estimating the factual situation in the population and quantifying the uncertainty in the estimates are thus important aims of such statistical analyses. Having done so, a natural next question of importance is what kind of exposures can be avoided, or how many diseased cases can be prevented, if such exposure could be completely or partially eliminated. A statistical concept that can be used to quantify this is the attributable fraction. For a single disease caused by a single exposure the attributable fraction due to this factor is the proportion of diseased subjects that could have been prevented if the specified exposure had not been present. Or, in other words, one questions what would the proportionate reduction in diseased subjects in the population be if the exposure distribution had been different from what it actual is? For illustration, an Italian study estimated that 15.0 % of the breast cancer cases might have been avoided if the betacarotene intake had been increased to at least 3366 μg/day for everyone while not changing the distribution of a number of other risk factors (low vitamin E intake, residence, alcohol habits, physical activity, age, educational level, calorie intake and menopausal status). Increasing also vitamin E intake (to at least 8.5 mg /day) for all Geir Egil Eide Attributable fractions 12 subjects gave a combined attributable fraction per cent of 21.5%. Sometimes eliminating a common exposure with a moderate increased risk of disease may have the same effect in the population at large as eliminating a rare exposure with a highly increased risk of disease. Thus, an attributable fraction depends both on the risk of disease if exposed and the extent of the exposure in the population studied.
In general, the attributable fraction quantifies the proportion of cases prevented if the factual exposure distribution were replaced with a hypothetical, so called counterfactual, exposure distribution. The attributable fraction can also be crudely defined as excess proportion of diseased in the population relative to the total proportion. The attributable fraction has also several other applications, e.g. to quantify the proportion of diseased that can be ascribed to one or more exposures (epidemiology), to predict the effect of planned preventive interventions (health policy) and to apportion the responsibility for the disease to various agents responsible for the exposure (liability law). It has been used in regional and national research, as well as in global studies like the Global Burden of Disease and Comparative Risk Assessment projects of the World Health Organization.
With multiple risk factors attributable fractions can be defined in many ways depending on how the counterfactual situation is hypothesized. This thesis describes how attributable fractions can be defined, interpreted and estimated for various scenarios, e.g. one factor is eliminated while the rest is kept fixed; several factors are eliminated; and, multiple factors are removed sequentially may be in different orderings. It also describes convenient graphical methods to illustrate the potential impact on disease load in a population from interventions on one or more risk factors.
The statistical and graphical methodology is potentially useful as tools in health policy discussions illustrating possible effects of different preventive strategies under evaluation and may ease the communication between researchers, decision takers and the public. Which strategy will have the largest effect in a public health perspective?
Which factors should be given priority in a public health intervention or in Geir Egil Eide Attributable fractions 13 legislation? How much can be achieved by changing personal habits versus general prevention of environmental exposure locally, nationally or globally? Methodology for computerized, and possibly interactive, manipulations of different scenarios is developed to depict the estimates of possible consequences.