Issue link: https://www.e-digitaleditions.com/i/1331851
Tablets & Capsules January/February 2021 41 indicates the presence of a trend in the beginning of 2020. This trend cannot be related to the addition of a second roller dryer, as the graph shows a high degree of similar- ity between data points from both roller dryers. The trend in data is explained by a transfer of the qual- ity control laboratory to a new loca- tion in quarter 1 of 2020. This is visualized in the score plot in Fig- ure 2, which shows clustering of data before (blue) and after (red) the laboratory move. It's important to understand that all batches pro- duced were well within specification and that the clustering of colors represents a small trend in the data. Qualification of the existing qual- ity control equipment after relo- cation proceeded prospectively according to IPEC guidelines. All parameters were well within speci- fications before and after the move, and univariate comparative test- ing of different analyses on differ- ent products revealed no differences. The observed trend in the data is an example of the strength of MVA, as it is able to identify minor trends by using a large data set. Early obser- vation of minor trends in data sets is important for excipient suppliers. It allows for quick adjustment of pro- cesses within the validated range, enabling year-to-year and batch-to- batch consistency. Apart from the ability to identify patterns, PCA also provides informa- tion on key drivers for differences in the data set. The loading plot of this analysis indicates key drivers, such as particle size (%w/w <150 µm, %w/w <45 µm), bulk and tapped density, moisture content (Karl-Fisher and LOD) and color (400 nm). Figure 3 shows a PCA score and loading plot for batches of anhy- drous lactose analyzed after the labo- ratory move. A high overlap between the data points for both roller dryers shows that there is a high degree of consistency between these batches. The difference between roller dryer 1 and roller dryer 2 in Figure 3 is significantly smaller than reported differences between batches of one two components is 0.32. This indi- cates little trend in the data, as only 32 percent of the data is explained by the two first components. A high overlap between data points from roller dryer 1 from 2017, 2018, and 2019 shows that there is a high degree of consis- tency between these batches. The separate clustering of 2020 data production lines. More than 150 batches of anhydrous lactose from roller dryer 1 (RD1) were taken over the four years from 2017 to 2020. All data from roller dryer 2 (RD2) originates from 2020. The ellipse indicates the 95 percent hoteling's interval. All analyzed batches are well within specifica- tion. The combined R 2 (x) of the Figure 3 Principal component analysis score plot and loading plot for all batches of SuperTab 21AN produced after the laboratory move (Overlap between the two data sets indicates a high degree of similarity between data points from both roller dryers.) a. Score plot t[2] 4 3 2 1 0 -1 -2 -3 -4 t[1] -5 -4 -3 -2 -1 0 1 2 3 4 5 Hotelling's T2 95% confidence limit RD1—2020 RD2—2020 R2x[1] = 0.21 R2x[2] = 0.164 b. Loading plot p[2] 0.6 0.5 0.4 0.3 0.2 0.1 0 -0.1 -0.2 -0.3 -0.4 p[1] -0.4 -0.3 -0.2 -0.1 0 0.1 0.2 0.3 0.4 0.5 0.6 R2x[1] = 0.21 R2x[2] = 0.164 Color 270 nm Color 210 nm Color 400 nm Beta content (%) Bulk density Acidity (ml) Specific rotation (°) PSD %<45 µm PSD %<150 µm Tapped density LOD (%w/w) PSD %<250 µm PSD %<355 µm KF (%w/w)