Newcastle University

Aims

To provide an appreciation of the need for, and an understanding of, the principal statistical methods required in those parts of biostatistics covered by clinical trials and the analysis of binary and survival outcomes.
On completion of this module students should understand the necessity of randomization for the assessment of new treatments; know how to calculate the size of a trial with either Normal or binary outcomes, have a working knowledge of how to allocate patients to treatment; know how to analyse data allowing for baseline information in the analysis of data and the analysis of binary and survival data. They should also have some understanding of the problems posed by subgroups and multiple outcomes and how to deal with simple protocol deviations by `intention to treat'. They should also be aware of the simple crossover trial for continuous data. The students should also be familiar with the main ideas of binary data analysis and survival analysis, including for the latter the causes of censoring. They should be familiar with methods for plotting and comparing survival curves and for assessing the effects of covariates on survival. Students should be able to use R to apply these to real data. The students will develop their skills in data analysis and interpretation.

Original Summary:
The term ‘biostatistics’ has come to mean the application of statistics to areas of medicine. This course will focus on two important areas, namely clinical trials and analysis of commonly arising forms of medical data.
New drugs, new types of operations, new screening programmes are some examples of new medical procedures whose efficacy needs to be assessed before their value to patients and to society can be assured. Over the last fifty years, the randomized controlled trial has come to have a dominant role in this area of medical research. The ideas behind clinical trials are essentially statistical, and statisticians are closely involved with all aspects of their design, analysis and implementation. This course presents the main statistical ideas and the principal methods that are employed in the area. This will involve issues of design, analysis and interpretation. The initial analyses presented will concern Normally distributed data but binary data and survival data often arise in medical applications and methods for handling these will be discussed. For binary data the outcomes are just 0 or 1, such as whether a patient responds to treatment or not. Inferential methods based on the Binomial distribution will be reviewed and the use of logistic regression to adjust for other variables, such as baseline values, will be introduced. Survival data are, e.g, the time a patient survives after treatment. The main complication with survival data is that many observations will be ‘censored’ or only partially observed. For example when a trial of a new treatment for cancer is terminated many of the patients will still be alive. Therefore the survival times of those who died will be known exactly whereas for those still alive their survival time is only known to exceed their present survival. Methods for dealing with this form of data will be considered.

Outline Of Syllabus

The historical context of clinical trials and the need for randomization in the assessment of treatments. The idea of bias. Patient selection, ethical constraints and allocation to treatment. Power calculations for normal and binary outcomes. Methods for controlling bias in assessing outcomes. Analysis of data from clinical trials: baselines as covariates. Subgroups and multiple outcome variables. Protocol deviation and “Analysis by Intention to Treat”. Crossover trials.
Non-parametric survival analysis: calculation of Kaplan-Meier estimates, use of log-rank statistics. Parametric survival analysis: exponential, Weibull and log-logistic distributions; likelihood analysis of effect of covariates; basic ideas of proportional hazards models. Review of comparing proportions using results form the Binomial distribution and the use of logistic regression and interpretation of adjusted odds ratios.

Teaching Rationale And Relationship

Lectures are used for the delivery of theory, illustrated with examples. Tutorials are used to help develop the students’ abilities at applying the theory to solving problems. Practicals are used both for solution of problems and work requiring extensive computation and for simulation to give insight into the ideas/methods studied.

Assessment Rationale And Relationship

A formal unseen written exam is the most appropriate method of assessment for this type of material. The coursework (approximately 4 written assignments of approximately equal weight), project work (a single assignment but submitted with two preliminary submissions before the full project is submitted) and the tests (approximately 2 tests of approximately equal weight) will give the students the opportunity to practise the methods introduced in the course, assess their progress with it and prepare for the exam.