Challenges In Identification And Handling Of Out-of-trend Results - AlphaMD

MUMBAI, India - May 26, 2017 - PRLog -- Variability is common in all areas of pharmaceutical manufacturing and testing. Tablet weight and hardness, blend and content uniformity, target fill volume for liquid and semisolid dosage forms and dissolution variability are common examples of drug product parameters that are subject to variability or spread about a target (or mean) value.

       Eliminating variability is impossible but controlling and understanding the sources of variability forms the basis of root cause analysis. Variability may be introduced by the measurement method or the manufacturing process. Test values that fall outside the specification limits that are usually set by manufacturers or regulatory authorities for products are labeled as Out-of-Specification (OOS) points. Batches that exhibit OOS results are often discarded by the manufacturer along with a thorough investigation of the cause(s) of the OOS result. Handling of OOS results is described comprehensively in a guidance document released by the FDA.

In certain cases the test result may not lie outside the specification limit but may exhibit variability about a mean value that is calculated from a historical batch data. Out-of-trend (OOT) is the term used to describe data that does not follow the expected trend in comparison with other batches. Although there is no clear legal or regulatory basis to define data that is within specifications but out of trend, FDA guidance on OOS states that similar methods may be used to examine OOT results. However, identification of OOT results is a complicated issue as it relies on obtaining data from historical batches that in itself exhibit variability. Common causes for OOT results are changes in operator, equipment, calibration or cleaning methods or slight changes in analytical method.

       Several methods exist for identification of OOT test results such as the by time point method, regression control charts and slope control chart methods. The simplest method to identify OOT results used routinely in manufacturing and analytical settings is to set limits at ±2 standard deviation (SD) and ±3 SD from the mean value (obtained from historical data sets). Test results falling outside these limits are labeled outliers or OOT points and the aberrant batches are routinely subjected to investigation and root cause determination for the observed deviation. However the above-mentioned methods are suitable for normally distributed data sets and are likely to result in erroneous results when data deviates significantly from normality. Another problem is the handling of marginal results: results falling outside set control limits that deviate by a fraction of a percent are to be investigated as per the currently used methods. However marginal differences may not lead to conclusive results as regards the cause of the variation and often result in unnecessary, lengthy and expensive investigations. The question remains: what OOT results are due to assignable causes versus those that do not require investigation?

One of the causes for marginal results being identified as OOT is the erroneous assumption of normality of the data distribution. When investigated, data sets that deviated from normality had fewer results being identified as OOT when a non-parametric method was used for identification of outliers as compared to when the standard ±2 SD and ±3 SD method was applied. In contrast, when the data set followed a normal distribution, the test results identified as OOT were the same when the ±2 SD/± 3 SD method or a non-parametric method provided the data set was large (> 50 results). This enabled batches that deviated significantly from the control limits to be considered OOT to which possible cause(s) of deviation could be assigned and safeguarded marginal results from detailed investigations.

       Handling of OOT results does not stop at application of statistical tests to identify OOT batches. A thorough investigation of the causes leading to OOT results followed by remediation of the causes can lead to better process control and less wastage of resources. Risk assessment of the manufacturing process using Failure Modes and Effects Analysis (FMEA) can help assign possible reasons for observed OOT results. This should be further supported by Corrective and Preventive Action (CAPA) as it is frequently seen that batches that do not follow the expected trend are likely to fall out of specification.

Along with tools such as control charts that can help to identify problematic production batches that deviate from historical data, it is important to understand the cause(s) of failure and suggest corrective measures that can help reduce the occurrence of OOT results. Therefore a thorough handling of OOT results includes both logical statistical analysis as well as a comprehensive risk assessment.

Visit http://www.alphamd.com for details