Please activate JavaScript!
Please install Adobe Flash Player, click here for download


116 ISCB 2014 Vienna, Austria • Abstracts - Poster PresentationsWednesday, 27th August 2014 • 11:00-11:30 Monday25thAugustTuesday26thAugustThursday28thAugustAuthorIndexPostersWednesday27thAugustSunday24thAugust Methods: We analysed 404 Follicular lymphomas cancer patients (FL) di- agnosed between January 1, 1980 and December 31, 2005. Each FL was matched on age and gender with 4 individuals randomly selected from the general population. Causes of death were divided into 3 groups: car- diovascular disease (CVD), cancer lymphoma related and other. For the FL patients we have investigated overall survival (OS), progression free sur- vival (PFS) and time to next treatment (TNT). When comparing with the general population, we have computed overall survival and cumulative incidences stratified by the three causes of deaths using competing risk methodology. Results: This population-based study revealed larger overall mortality risk but not due to CVD in FL patients than in the general population. The cu- mulative incidence of TNT was still elevated for stage II patients compared to Stage I, but both estimates were much lower than when modelled with Kaplan Meier methodology. Conclusions: Many cancer patients live longer and respond well to treat- ment so they might experience a relapse/need for new treatment long time after the initial diagnosis when also other competing events might interact with the main event of interest. Therefore, the choice of the cor- rect methodology is crucial.   P3.1.149 Cox model with multiple events: an application to mammography screening intervals in the Portuguese primary health care system A Sousa1 , DP Tavares2 , H Mouriño1 , P Nicola3 1 Faculdade de Ciências - Universidade de Lisboa, Lisboa, Portugal, 2 Universidade de Lisboa, Lisboa, Portugal, 3 Faculdade de Medicina - Universidade de Lisboa, Lisboa, Portugal   According to the World Health Organization, breast cancer is the top can- cer in women both in the developed and the developing world. In 2012, breast accounted for 522 000 deaths worldwide. Although these numbers, breast cancer mortality has been falling in many European countries due to the combined effects of breast screening and better treatments. So far the only breast cancer screening method that has been proved ef- fective is mammography screening. In 2003, the European Council has suggested to all the Member States to undertake screening for women aged 50-69 years and the Portuguese Directorate-General of Health has adopted this recommendation and women must be screened every two years. Nowadays the screening covers 60% of the territory. This study aims at identifying variables associated with an increase of the time interval between screenings events. We focus on all women enrolled at Family Health Units (FHU) from Lisbon. It covers the period from 2000 to 2013.The variables used are age at study entry, body mass index, age at menarche, alcohol consumption, smoking, menopausal status, contracep- tive use and the number of primary care visits. The Cox model with multiple events has been estimated. This model al- lows for multiple mammographies per woman. We found out that high body mass index, hormonal contraceptive use, menopausal status and number of primary care visits are related to the time between screening examinations.   P3.1.150 Statistical modelling of biomarkers incorporating non-proportional effects for survival data J Stephen1 , G Murray1 , J Bartlett2,3 , D Cameron3 1 University of Edinburgh, Edinburgh, United Kingdom, 2 Ontario Institute for Cancer Research, Toronto, Canada, 3 Edinburgh Cancer Research Centre, Edinburgh, United Kingdom   Personalised medicine is replacing the one-drug-fits-all approach with many prognostic models incorporating biomarkers available for risk strati- fying patients. Evidence has been emerging that the effects of biomarkers change over time and therefore violate the assumption of proportional hazards when performing Cox regression. Analysis using the Cox model when the assumptions are invalid can result in misleading conclusions. We report the results of a review of existing approaches for the analysis of non-proportional effects with respect to survival data which identi- fied a number of well-developed approaches but a lack of application of these approaches in practice. The review indicated there is a need for more widespread use of flexible modelling to move away from standard analysis using a Cox model when the assumption of proportional hazards is violated. We further illustrate the use of two key approaches; the multivariable frac- tional polynomial time (MFPT) approach by Sauerbrei et al. and flexible parametric models proposed by Royston & Parmer, to develop a model for predicting survival of patients with early breast cancer. We illustrate their respective advantages and disadvantages in the development and evalu- ation of a prediction model.   P3.1.151 Impact of length of follow-up on the evaluation of prognostic scores with an example using two breast cancer studies J Stephen1 , G Murray1 , J Bartlett2,3 , D Cameron3 1 University of Edinburgh, Edinburgh, United Kingdom, 2 Ontario Institute for Cancer Research, Toronto, Canada, 3 Edinburgh Cancer Research Centre, Edinburgh, United Kingdom   Background: We investigate the impact of follow-up duration on two residual risk models, IHC4 and Mammostrat, for predicting risk in early breast cancers using two studies with different lengths of follow up; the Edinburgh Breast Conservation Series (BCS) and the Tamoxifen versus Exemestane Adjuvant Multinational (TEAM) trial. Methods: The multivariable fractional polynomial time (MFPT) algorithm was used to determine which variables had possible non-proportional ef- fects and the best fitting fractional polynomial to model these effects. The performance of the scores was assessed at various lengths of follow-up using measures of discrimination and calibration. Results: We observed a strong time-dependence of both the IHC4 and Mammostrat scores. Both scores were significant independent predic- tors of outcome restricted to the first five years of follow-up, after which the scores were not associated with distant recurrence free survival. The models performed statistically better with shorter follow-up compared to full follow-up with differences in D statistic between 0.4 and 0.5 and R2 between 7 and 13%. Conclusion: Our analyses confirm that it is important to consider the length of follow-up and violations of the Cox proportional hazards as- sumption when evaluating prognostic models. Longer follow-up resulted in strong degradation of the performance of the scores.  

Pages Overview