Background: The typical degree of observer bias in randomised clinical trials with non-blind outcome assessors of continuous outcomes is not known. Previous analyses which normally involve comparisons between trials with blind assessors and 'similar’ trials with non-blind assessors, are susceptible to confounding. In an earlier work, based on trials that use both blind and non-blind outcome assessors of the same binary outcome, we found that substantial observer bias could result from misclassification of few patients. Trials with continuous outcomes may have a different susceptibility to observer bias, and an estimate of the typical degree of bias in such trials is important for the assessment of risk of bias.

Objectives: To estimate the degree of observer bias in randomised clinical trials with continuous outcomes, and to identify factors associated with large observer bias.

Methods: A systematic review of randomised clinical trials with both blind and non-blind assessment of the same continuous outcome. We searched PubMed, EMBASE, PsycINFO, CINAHL, The Cochrane Trials Register; HighWire Press and Google Scholar. We plan to extract data from one outcome per trial, and to calculate the difference between the standardized mean difference (SMD) based on blind assessments and the SMD based on non-blind assessments: dSMD. We plan to pool the dSMDs with inverse variance meta-analysis using random effects models. Furthermore, we will study whether the size of dSMD is associated to a)high degree of outcome subjectivity, e.g. global clinical improvement; b)close trial involvement of the non-blind assessors (e.g. they also treat patients); and c)high degree of sensitivity to reporting and behavior of non-blind patients (e.g. symptom interviews).

Results: At the time of abstract submission we had identified 23 eligible trials. We expect full outcome data from 17trials.

Conclusions: Pending.