Students were randomised using simple randomisation.
Patients were randomised using minimisation, balancing on the size of the pleural effusion at baseline, whether the infection was acquired in hospital or not, and the presence of purulent pleural fluid.
Patients were randomised using a Bayesian algorithm.
Following completion of the baseline battery women will be randomised using a computer-generated algorithm to either the intervention or control condition.
After completion of the baseline assessments, participants will be randomised using a computerised random number generator (block size=3) in a 1 1 1 ratio to one of the three intervention groups: (1) exercise 1 day/week, (2) exercise 3 days/week or (3) education.
Within one week of the baseline telephone interview, participants will be randomised using computer generated permuted blocks of varying sizes (unknown to investigators) stratified for study centre (Wales and London) and for 'medication evaluation' only.
After written consent is obtained, baseline assessment is completed and then participants are randomised using sealed sequential, opaque envelopes.
After baseline measurements, women were randomised by using a computer programme, stratified for municipality to a diet-induced weight loss group (diet group; n = 97), a weight loss induced mainly by exercise group (mainly exercise group; n = 98) or a stable weight control group (control group; n = 48).
After consent and completion of baseline data collection, the participants were individually randomised using a random block size, stratified by centre and a web-based remote randomisation system into either a 1 week course or a 5 week course.
Eligible patients were randomised at baseline using a minimisation algorithm that was stratified by site and baseline DAS28 based on erythrocyte sedimentation rate (ESR) (≤ or >5.5).
