Residual Biases

Readers of this CLAHRC WM News Blog know that the Director is more interested in how we get to think we know things, than in what those things are. A pervasive issue for clinical research is that of residual biases. By residual biases I mean bias that exists after statistical adjustment for known confounders in non-experimental studies. But how to get a better idea about the extent of such biases?

The standard method since the Schulz et al. classic [1] is to compare the results of non-experimental and experimental studies of the same intervention under the premise that the RCTs are the gold standard. There are now myriad examples of such ‘meta-regressions’. This week’s Director’s Choice is another approach to this issue from a group of authors from Harvard, Philadelphia and Stanford Universities.[2] These authors carried out a data base study of patients who were eligible for a certain treatment (implantable cardioverter-defibrillator for heart failure) and who did or did not receive the treatment. They carried out statistical adjustment and then compared outcomes that signify general poor health but that are regarded to be independent of the indication for treatment – fractured hip and admission to a nursing home, for instance.

Did the non-intervention group have higher incidence of such health problems? The answer is yes, the difference was massive and the curves deviate from day one. When I teach medical students I use examples of the ‘hidden biases’ that cannot be controlled statistically. Doctors, it seems, have a tacit ability to discern human frailty.I will add this to my example list. I would be delighted to disseminate further such examples in the lofty pages of this News Blog.

— Richard Lilford, Director CLAHRC WM


  1. Schultz KF, Chalmers I, Hayes RJ, Altman DG. Empirical Evidence of Bias: Dimensions of Methodological Quality Associated With Estimates of Treatment Effects in Controlled Trials. JAMA. 1995; 273(5): 408-12.
  2. Setoguchi S, et al. Influence of healthy candidate bias in assessing clinical effectiveness for implantable cardioverter-defibrillators: cohort study of older patients with heart failure. BMJ. 2014; 348: g2866.



Leave a Reply

Fill in your details below or click an icon to log in: Logo

You are commenting using your account. Log Out /  Change )

Google+ photo

You are commenting using your Google+ account. Log Out /  Change )

Twitter picture

You are commenting using your Twitter account. Log Out /  Change )

Facebook photo

You are commenting using your Facebook account. Log Out /  Change )


Connecting to %s