Abstract |
The conformity assessment of a product or item that require the evaluation of multiple parameters is a challenge, particularly when these parameters are correlated. Correlation may arise due to the material characteristics of the product (intrinsic correlation) or due to how the components are measured (metrological correlation). This work aimed to propose a methodology to decompose the total correlation (experimental correlation) into intrinsic and metrological correlations since both correlation components impact sequentially and differently on the risk of false conformity decisions. The developed methodology was applied to assess the conformity of a medicine with two active pharmaceutical ingredients, using independent and correlated analytical procedures. The correlations were calculated for each type of analysis, using a developed and available MS-Excel spreadsheet. The global total risks were estimated by Monte Carlo Method, considering the intrinsic and metrological correlations and also considering only the total correlation. Metrological correlation can increase or decrease the impact of the original intrinsic correlation on values used to assess conformity. The approach proposed in this paper allows a more realistic quantification of the risk of false decision than assuming the same correlation value attributed to material characteristics and the measurement process. The estimated intrinsic correlation values are statistically equivalent regardless of whether the medicine s experimental characterisation was supported on independent or correlated measurements. The metrological correlation reduced the global total producer s risk significantly from (5.2-5.6\%) to (4.6-5.0\%), with 95\% confidence level. If it is assumed the same intrinsic and metrological correlation, the risk of false decision is underestimated (about 11\% in the case studied in this work). Differences between accurate and simplified correlation management depend on the specific scenario. Thus, intrinsic and metrological correlations affect the global total risk of false decisions differently and should be taken into account separately in the conformity assessment of a product or item that require the evaluation of multiple parameters. |