EuroCoord Methodology Meeting

March 2015

The statisticians involved in EuroCoord met in London on 5 December 2014 for their annual meeting to discuss statistical methodology.

During the morning session Virginie Rondeau (University of Bordeaux) gave a presentation on FRAILTYPACK, an R package for the analysis of frailty models using penalized likelihood estimation. Virginie was the original developer of FRAILTYPACK and has supervised improvements and extensions to the program. Frailty models were developed for the analysis of correlated survival data – such data can arise because of clustering in the data (e.g. families, hospitals, trials) or because recurrent events are observed in the same individual. The presentation included a blend of theory and examples, such as: re-hospitalizations of patients with colorectal cancers; effect of air pollution on cardiopulmonary mortality by geographical area; placebo-controlled randomized trial of gamma-interferon in Chronic Granulomatous Disease; and a meta-analysis of randomized trials in patients with head and neck cancers.

                FRAILTYPACK is freely available at http://cran.r-project.org/web/packages/frailtypack/

The afternoon session started with an exploration of issues in the analysis and interpretation of competing risk models, led by Ronald Geskus. Ronald illustrated his presentation with examples from his own career,as an analyst and statistical reviewer. Much of the discussion hinged over the question of whether attendees were interested in the cause-specific hazard or sub-distribution hazard (i.e. ignoring the occurrence of the competing event). Ronald pointed out that this primarily depended on whether the analysis was being conducted for the purposes of prediction or to give insight into disease aetiology (in practice, both may be of interest).

The day concluded with several shorter presentations of ongoing work: “Late entry into the cohorts when analyzing longitudinal CD4 data from vertically HIV infected children” by Giota Touloumi; “Does HIV viremia copy years have a causal effect on morbidity/mortality?” by Al Cozzi Lepri; “Viral load and biomarkers: are blips correlated?” by Alex Szubert; “Imputing viral subtype in time to virological failure analyses” by Ellen White; “CD4 and viral load measurements of seroprevalent HIV-positive individuals to improve predictions regarding the time of their infection” by Nikos Pantazis. All the talks were well received and generated useful discussion. A similar event will be held in 2015.