Table 5 Challenges and recommendations

Study population

Dealing with participation bias

• Keep non-response or loss to follow-up to a minimum

• Responders and non-responders

• Characterize non-responders or those lost to follow-up

• articipants and non-participants

• Control for extent and direction of bias in final data analysis

• Different types of participants

• Participants lost to follow-up for the clinical assessment Finding an adequately sized and well-matched control group

• In case the number of controls is insufficient: incorporate other types of control subjects

• Choose types of controls that are representative of the study population

• Characterize and control for differences between survivors and controls

• Compare self-reported data with an more objective source, such as medical records or registries

Data collection

Validating instruments for data collection

• Conduct reliability studies to account for inter- and intra- observer variation

• If possible, use data collection instruments that allow for one investigator to analyse collected data (observer bias)