Conflict of Interest has become an area of increasing interest in research in recent years, but most of this attention has been focused on financial areas. Our new paper sheds light on some other conflicts, which we contend are potentially hidden, and yet still cause results to be skewed in a number of ways. For instance, in a study about sex education, a researcher need not declare their religious background that might influence their beliefs and findings. In most cases, readers will be unaware of the (potential) biases that are present and this is a worry for our fields that needs attention.
We already know that effect sizes are larger when programme developers conduct studies compared to independent researchers. These disparities are sometimes attributed to biases linked to a potential conflict of interest. These effects can be viewed as either intentional – meaning there is a conscious decision to manipulate or in extreme cases defraud – or unconscious, wherein the researcher does not recognise his or her own ‘self-serving bias’ (Eisner, 2009). Confirmation bias, where researchers and authors have difficulties ignoring their own self-interest in favour of disinterested and impartial evaluation, has also been recognised.
Conflict of interest (CoI) is often identified by the Risk of Bias Tool (RoB) included in Cochrane systematic reviews and is designed to identify and measure key sources of bias (Higgins, 2011). However, this tool may not be an adequate means of recognising all potential conflicts of interest and the biases it can cause, when those conflicts are not strictly financial in nature, mostly because these biases are more subtle, rarely reported and just as likely to lead to flawed results.
Our paper highlights key arguments about non-financial CoI in the field of social and behavioural intervention research and presents three case studies of widely-disseminated child and family interventions which have been the subject of critical appraisals and debates related to potential non-financial CoI as a source of bias (Multisystemic Therapy, Functional Family Therapy and Triple P). The case studies illustrate how evidence appraisal can differ depending on the independence of the appraisers and how this may affect decisions around what is considered ‘good’ evidence. The case studies highlight the complex issues involved in trying to substantiate non-financial CoI, the unsatisfactory and fragmented ways that it is currently addressed, and the need for improved guidance to prevent and redress it.
We suggest developer-led work in social interventions is at greater risk of bias. Current guidelines to address the problem of non-financial conflict of interest bias effects are inadequate. This calls for a need for better reporting and reviewing of studies to capture this more subtle, but still potentially damaging, form of bias.