The procedure has eight steps, shown in sequence in Figure 1. Table 1 lists what each step is for and what it should produce. An itrifal of the simplest kind, built on the dried fruits of Terminalia chebula, Terminalia bellirica and Emblica officinalis, serves as the running illustration. Gallic acid, ellagic acid and the hydrolysable tannins of these fruits are well characterised and are widely used as reference substances for them.
Steps of the proposed procedure, with the purpose and the output of each
| Step | Purpose | Output |
|---|---|---|
| 1. Authenticate ingredients | Confirm botanical identity and purity of every crude drug | Vouchers and single-drug chromatograms |
| 2. Reference batches | Capture genuine variation; trace each ingredient | Complete and minus-one laboratory batches |
| 3. Extraction and clean-up | Remove the sugar base and recover analytes reproducibly | Fixed sample preparation protocol |
| 4. Method development | Resolve as many characteristic constituents as possible | HPTLC and HPLC conditions |
| 5. Validation | Show that the methods are specific, precise and robust | Validation report |
| 6. Reference fingerprint | Define the common pattern of genuine batches | Characteristic peaks and mean profile |
| 7. Acceptance criteria | Turn the pattern into a pass or fail rule | Written specification |
| 8. Routine use | Test production batches and monitor drift | Batch records and trend charts |
Sequence of the proposed fingerprinting procedure for a classical compound formulation
A fingerprint of the product is only as sound as the identity of what went into it. Each crude drug is authenticated by a qualified botanist or pharmacognosist against pharmacopoeial descriptions, using macroscopic and microscopic characters, and a voucher sample is retained. A thin-layer chromatogram of each single drug is run at this stage, following standard practice for plant drug analysis (Wagner and Bladt, 1996). Contaminants are tested separately: heavy metals, pesticide residues, aflatoxins and microbial load fall under their own limits and are not addressed by a fingerprint (WHO, 2007).
The reference against which later batches will be judged cannot be a single preparation. Several batches are made in the laboratory strictly by the pharmacopoeial method, each from authenticated drugs of a different geographical or commercial origin, so that the natural variation of genuine material is represented. For each, a set of "minus-one" batches is also prepared, in which one ingredient at a time is left out. Comparing these with the complete product shows which peaks or bands each ingredient contributes and so allows every component of the formula to be traced in the finished mixture.
The sample preparation has to separate the analytes from the sugar base without losing them. A common approach is to disperse a weighed portion in water and partition it against solvents of increasing polarity, or to extract directly with aqueous methanol and pass the extract through a solid-phase cartridge that retains the phenolic and less polar constituents while the sugars wash through. Whatever route is chosen, it is fixed in writing, including the sample mass, solvent volumes, times and temperatures, because small changes here alter the fingerprint more than any instrumental setting. Volatile constituents, where they matter, need a separate preparation and gas chromatography.
Two techniques are developed side by side. High-performance thin-layer chromatography is inexpensive, runs many samples on one plate under identical conditions and gives an image that can be read at several wavelengths and after derivatisation; it suits the routine laboratory of a Unani manufacturer (Reich and Schibli, 2007). Reversed-phase high-performance liquid chromatography with diode-array detection gives better resolution and quantitative data. Because no single run shows every class of compound, the use of more than one fingerprint for the same product has been recommended (Fan et al., 2006). For the myrobalan itrifal, an acidified mobile phase is needed to keep the phenolic acids in a single ionisation state and so obtain sharp peaks. Conditions are optimised for the number and separation of characteristic peaks and not for the speed of one marker (Tistaert et al., 2011).
The marker assays within the method are validated in the ordinary way for specificity, linearity, range, accuracy, precision, limits of detection and quantitation, and robustness (ICH, 2005). The fingerprint itself is validated on the parameters that apply to a pattern: repeatability of relative retention times or RF values and of relative peak areas, intermediate precision across days and analysts, stability of the sample solution, and robustness to small changes in mobile phase, temperature and plate or column lot. The minus-one batches from Step 2 provide the test of specificity, since the method should detect the absence of each ingredient.
The chromatograms of the reference batches are aligned, since retention times drift slightly between runs. Correlation optimised warping is the most widely used alignment method (Nielsen et al., 1998; Tomasi et al., 2004). Peaks present in every genuine batch are designated characteristic peaks, and those identified against reference substances are labelled. A mean or median chromatogram then serves as the reference fingerprint. Principal component analysis of the aligned data is a helpful check at this point: genuine batches should cluster, and minus-one batches should fall outside the cluster (Wold et al., 1987).
A fingerprint becomes a quality standard only when it yields a decision. Three kinds of criterion are combined. The first is qualitative: every characteristic peak or band must be present at the expected position. The second is a similarity measure, usually the correlation coefficient or the cosine of the angle between the sample chromatogram and the reference, with a threshold derived from the spread among the reference batches themselves (Gong et al., 2003). The third is quantitative: content limits for one or two markers. The threshold must come from the data on genuine batches and should not be borrowed from another product.
In routine use each production batch is run alongside a retained reference batch on the same plate or in the same sequence. Results are charted over time so that a slow drift, such as a change of raw material supplier, is noticed before a batch fails. The reference fingerprint is reviewed when the source of an ingredient changes.