Assessing treatment effects in older breast cancer patients: Systematic review of observational research methods

• Observational research will be essential to improve treatment of older breast cancer patients.
• The use of accurate methods is essential to draw valid conclusions from this type of data.
• The majority of observational studies in older breast cancer patients compared treatments directly.
• These studies are therefore likely to be biased due to confounding by indication.

Solid evidence of treatment effects in older women with breast cancer is lacking, as they are generally underrepresented in randomized clinical trials on which guideline recommendations are based. An alternative way to study treatment effects in older patients could be to use data from observational studies. However, using appropriate methods in analyzing observational data is a key condition in order to draw valid conclusions, as directly comparing treatments generally results in biased estimates due to confounding by indication.The aim of this systematic review was to investigate the methods that have been used in observational studies that assessed the effects of breast cancer treatment on survival, breast cancer survival and recurrence in older patients (aged 65 years and older). Studies were identified through systematic review of the literature published between January 1st 2009 and December 13th 2013 in the PubMed database and EMBASe. Finally, 31 studies fulfilled the inclusion criteria. Of these, 22 studies directly compared two treatments. Fifteen out of these 22 studies addressed the problem of confounding by indication, while seven studies did not. Nine studies used some form of instrumental variable analysis.In conclusion, the vast majority of observational studies that investigate treatment effects in older breast cancer patients compared treatments directly. These studies are therefore likely to be biased. Observational research will be essential to improve treatment and outcome of older breast cancer patients, but the use of accurate methods is essential to draw valid conclusions from this type of data.