Quantile-BEAST algorithm

Commercial aspirin tablets were analyzed at 18 wavelengths using the modified sample apparatus and two other configurations. Principal component analysis was followed by discriminant analysis with the quantile-BEAST algorithm. Cluster separation was greatest for the spectra collected with the modified sample cell and least for tablets that had been powdered and placed in a traditional sample cup. 

The SIMCA analysis provided highly variable results, while the quantile-BEAST gave better results overall and more consistent prediction. The best results were obtained when the quantile-BEAST algorithm used the full spectra, with no principal axis transformation. 

The quantile-BEAST algorithm is a nonparametric bootstrap method based upon the work of Efron. In the 1988 study, four individual benzoic acid derivatives were analyzed as were mixtures of the four derivatives. The active ingredient concentrations were varied between 0 and 25% of the sample, with aluminum oxide used as a diluent. The samples were ground and passed through a 100-mesh screen prior to analysis. Spectra were collected in triplicate at three wavelengths on an InfraAlyzer 400. 

Continuing the analysis of intact dosage forms with NIR and the quantile-BEAST algorithm, Lodder and Hieftje published an article using the technique for the quantitative and qualitative characterization of capsules with low concentrations of contaminants. Using quantile-quantile (QQ) plots, detection of subpopulations in NIR spectral clusters was possible. These subpopulations were defined as samples whose distance from the center of the training group was less than three SDs. 

In this study, tablets were stored in a hydrator for up to 168 h with tablets withdrawn at regular intervals. After removal from the hydrator, the tablets were weighed and NIR spectra collected prior to the HPLC analysis. Spectra of the intact tablets were collected on an InfraAlyzer 500 in the 1100 to 2500 nm region, using the double-reflecting sample apparatus described by Lodder and Hieftje [82]. The spectra were processed by principal component analysis, and the scores analyzed by the quantile-BEAST algorithm. 

Drennen and Fodder [116] published a paper in 1991 comparing the performance of the improved quantile-BEAST algorithm with that of the Mahalanobis distance in the qualitative analysis of carbamazepine tablets. While the Mahalanobis distance calculation assumes that spectral variations associated with both the calibration and test set are random, in complex pharmaceutical mixtures this may not be the case. The bootstrap algorithm, on the other hand, is a nonparametric test which can be used with nearly any spectral data distribution. 

The qualification of tablet characteristics was evaluated in a study of two algorithms soft independent modeling of class analogies (SIMCA) and quantile-BEAST [95]. The study involved tablet hardness, moisture content, dissolution rate, and degradant concentration. 

In NIR calibration, the introduction of unexpected components (e.g., incorrect chemicals or contaminated raw materials) into a sample may go undetected, causing erroneous results. This type of false sample is potentially more serious than samples in which the correct constituents are present in the wrong concentrations. In a 1988 paper, Lodder and Hieftje [33] discussed in detail the application of a new algorithm, the quantile-BEAST (Bootstrap Error-Adjusted Single-sample Technique), which was more sensitive to false sample detection. The quantile-BEAST was proposed as a new method to assess pharmaceutical powder blends qualitatively. 

The first report of NIR applied to the analysis of intact dosage forms came as a direct result of the deaths caused by cyanide-laced capsules in the early and middle 1980s. Following these poisonings, the Food and Drug Administration analyzed two million capsules by a variety of methods. In 1987, Lodder et al. [81] published a landmark paper in which intact capsules were analyzed by NIR. In this study, the quantile-BEAST cluster-analysis algorithm was used in the analysis of adulterated and unadulterated capsules. 

---