Reducing the length of mechanical ventilation with significance: a case study of sample size estimation trial design using Monte-Carlo simulation

The power of a study can be established with estimation of total effective sample size (Ntotal). In this study, the impact of the length of mechanical ventilation (LoMV) distribution shape in intensive care on the estimated N_total is investigated. This study provides an overview on the study design involving LoMV, the resulting potential limitations, and the criteria for a 'successful' design. Data from mechanical ventilated patients in a single-center intensive care unit were used in this study. The N_total was estimated using two methods: 1) Model-based Altman's nomogram (a standard); and 2) Monte-Carlo simulation. Using Monte-Carlo simulation, a patient selection criteria is imposed to estimate N_total from 'realistic' patient cohorts. The Altman nomogram shows that the N_total to detect a 25% change in LoMV (ΔLoMV) at power of 0.8 is ≥1000 patients. For the Monte-Carlo simulation, a N_total ≥260 patients is needed to detect similar changes. It is important to consider the LoMV distribution shape and variability, particularly relative to target patient groups who might benefit from the intervention. Assessment of ΔLoMV in response to treatment should be carefully considered to avoid an under-powered studies. The Monte-Carlo simulation combined with objective patient selection provides better design of such studies.