Type II Error in the Spine Surgical Literature

In Brief Study Design. A literature review. Objectives. To determine the frequency of potential type II errors published in the spine surgical literature. Summary of Background Data. The randomized controlled trial is the strongest clinical evidence available in investigational medicine. Unfortunately, it is common for randomized controlled trials published in peer-reviewed journals not to report a primary question or a sample size calculation. When the null hypothesis is accepted and the power of a study is unreported, the validity of a study’s findings may be significantly limited. To our knowledge, the spine literature has not been appraised to determine the frequency of type II errors. Methods. A literature search was conducted of MED-LINE, PubMed, and Cochrane databases, using the key words of “spine” and “surgery” between 1967 and 2002. Trials were included if they were of a 2-group randomized controlled trial design, which reported a nonsignificant difference in the primary outcome. The frequency of reporting the primary outcome and sample size calculation was determined. The sample size was assessed to determine whether the trial had sufficient patients to detect a 10%, 25%, and 35% relative difference in the primary outcome for a power of 80%. Results. A total of 37 studies satisfied the inclusion criteria. Six studies reported a sample size calculation (17%). Of the remaining 31 studies, 5 explicitly stated a primary outcome (14%). The mean type II error (beta error) was 82%. Conclusion. The spine surgical literature is plagued with a high potential for type II error. A trial’s methodology should be scrutinized to prevent misinterpretation of the results. A review of the spine surgical literature revealed 37 randomized controlled trials that met the inclusion criteria. Seventeen percent of these articles calculated a sample size a priori. The remaining 31 trials had an average power of 18%. Only 17% of these studies were appropriately powered to detect a clinically significant difference between treatment groups.