Abstract
A recently reported “prospective, randomized study into amniotomy and oxytocin as induction methods in a total unselected population” was examined for selection bias and bias after entry into the study. The null hypothesis that clinical attitudes to amniotomy as a means for inducing labor had no influence on the decision to enter women into the trial and allocate them to either amniotomy or oxytocin was rejected at p < 0.00025. Clinical attitudes were further found to statistically significantly influence the prescribed assessments 4 h after entry into the trial and the selection of the second intervention that was required in the absence of acceptable progress (p < 0.0005). Bias at the time of this prescribed assessment was large enough to result in an inverse relationship between “acceptable progress within 4 hours” and “delivery within 24 hours” after induction.
A subanalysis of the nulliparae entered into the trial further substantiated both bias at entry and bias in following the prescribed protocol. As hypothesized, these biases reached a greater statistical significance in nulliparous than in parous women. The likelihood that all of these observations would be encountered in a truly randomized study of this size can be estimated to be less than one in a billion (or p < 0.000,000,000,000,1). The study, therefore, provides a classical example of the dangers of non-blind allocation to different treatment groups in clinical trials. It is further concluded that no randomized controlled studies between amniotomy and oxytocin in a “total unselected population” are available.
Key Words: