LAUSR

Abstract The estimation of measurement uncertainty (MU) in the detection of genetically modified organisms is still not a systematic and common practice. Here, the Monte Carlo method (MCM) and the Guide to Uncertainty in Measurement (GUM) approach were simultaneously implemented to evaluate the MU for the relative content of the 35S promoter from cauliflower mosaic virus (CaMV35S) in a mixed sample of genetically modified soybean. The two methods gave a mean estimate of the measure that was very close to the theoretical content of CaMV35S in the mixed sample (3.00%). However, the mean value of CaMV35S estimated by the MCM (2.95%) was smaller than that obtained using the GUM method (3.09%). Moreover, the MCM concluded that the standard uncertainty (2 × 10−4) of CaMV35S was ∼75% smaller than the value (8 × 10−4) estimated by the GUM approach. This suggests that the GUM method overestimated the uncertainty of the CaMV35S content. Additionally, the differences regarding the estimated coverage intervals between the MCM (2.91–3.00%) and GUM (2.93–3.25%) method were determined. The semi-width of the coverage interval (expanded uncertainty for 95% coverage probability) provided by the GUM method (0.16%) assuming a normal distribution was 72% greater than the value (0.045%) estimated by the MCM. Nevertheless, there was no significant difference between the two methods when their calculated uncertainties were rounded to two decimal places. This suggests that the normality assumption in the estimation of MU using the GUM approach is valid and satisfactory.