Similarity analyses of chromatographic herbal fingerprints: A review

Highlights

• Similarity analyses of herbal fingerprints are reviewed.

• Different (dis)similarity approaches are discussed.

• (Dis)similarity-metrics and exploratory-analysis approaches are illustrated.

• Correlation and distance-based measures are overviewed.

• Similarity analyses illustrated by several case studies.

Abstract

Herbal medicines are becoming again more popular in the developed countries because being “natural” and people thus often assume that they are inherently safe. Herbs have also been used worldwide for many centuries in the traditional medicines. The concern of their safety and efficacy has grown since increasing western interest. Herbal materials and their extracts are very complex, often including hundreds of compounds. A thorough understanding of their chemical composition is essential for conducting a safety risk assessment. However, herbal material can show considerable variability. The chemical constituents and their amounts in a herb can be different, due to growing conditions, such as climate and soil, the drying process, the harvest season, etc. Among the analytical methods, chromatographic fingerprinting has been recommended as a potential and reliable methodology for the identification and quality control of herbal medicines. Identification is needed to avoid fraud and adulteration. Currently, analyzing chromatographic herbal fingerprint data sets has become one of the most applied tools in quality assessment of herbal materials. Mostly, the entire chromatographic profiles are used to identify or to evaluate the quality of the herbs investigated. Occasionally only a limited number of compounds are considered. One approach to the safety risk assessment is to determine whether the herbal material is substantially equivalent to that which is either readily consumed in the diet, has a history of application or has earlier been commercialized i.e. to what is considered as reference material. In order to help determining substantial equivalence using fingerprint approaches, a quantitative measurement of similarity is required. In this paper, different (dis)similarity approaches, such as (dis)similarity metrics or exploratory analysis approaches applied on herbal medicinal fingerprints, are discussed and illustrated with several case studies.

Introduction

Traditional herbal medicines (THM) have been used for centuries by billions worldwide, but recently a revival is seen in Western nations. According to the World Health Organization (WHO) definition, herbal medicines include herbs, herbal materials, herbal preparations and finished herbal products that contain as active ingredients parts of plants, or other plant materials, or combinations [1].

A survey conducted in 2005 revealed that 71% of the Canadians were using natural health products, a term which includes not only Herbal Medicinal Products (HMPs) but also vitamins and minerals; of the persons surveyed 11% used herbal remedies and algal/fungal products [2]. In the United States in 2002, approximately 19% of the adult population was using HMPs [3]. Another study has shown that approximately 36% of pregnant women in Norway, a potentially vulnerable sub-population, use herbs [4]. Although a wide variety of herbal medicines have been used across the centuries, many mass produced plant-based products commercially available in the United States, Europe, Canada, and Australia bear little resemblance to the traditional, small scale and typically water based preparations [5]. Furthermore, unfortunately, in most countries the traditional medicines have not been officially acknowledged, maybe because of a lack of attention in this area or because of insufficient research. The quality, safety, and efficacy data supporting many herbal materials are far from adequate to meet the criteria set by western regulatory authorities. These criteria were also the consequence of some accidents that happened with herbal formulations. For instance, the replacement of Stephania tetrandra by Aristolochia fangchi in a dietary formulation with renal failure as a consequence, due to the presence of aristolochic acid in the latter plant.

The major mystery of a herbal medicine is its chemical composition, which varies depending on several factors, e.g. botanical species, used chemotypes, the anatomical part of the plant used (seed, flower, root, leaf, and so on), but also on storage conditions, amount of sun, humidity, type of ground, time of harvest, geographic cultivating area [6]. Thus, a herbal medicine may vary from batch to batch in the concentrations of their chemical constituents. The consequence hereof is that those batches may significantly differ in pharmacological activity: involving both pharmacodynamic and pharmacokinetic issues. In order to assure the safety, the efficacy and the authentication of properties and to create a link between the components and the traditional use of a plant, a botanical technique is insufficient, which makes a multi-technique approach necessary [7], [8].

Getting useful chemical information from samples with many components has long been a challenging task to chemists. Herbal medicines are complex mixtures, containing often hundreds of chemically different constituents but only a few, if not one, have been acknowledged by scientific experiments to be responsible for the beneficial and/or hazardous effects [9]. There may be hundreds of unknown components in a herb and its herbal medicinal extract, of which many occur in low amounts. Moreover, large variability may exist within the same herbal material, as discussed above. Additionally, the number of standardized herbal medicines available is very limited. For instance, in the 1990 Chinese Pharmacopoeia, only 100 documented herbs have markers described [10] and these markers represent a limited percentage of the material as a whole.

Recently, more attention has been focused on safety and efficacy, and more specifically on the quality control of herbal medicines. For instance, in 1991, the World Health Organization accepted the chromatographic fingerprint technology as an identification and qualification technique for medicinal herbs [1]. In 2000, the State Food and Drug Administration of China (SFDA) allowed developing fingerprints of Traditional Chinese Medicines as the standard of quality control [11], [12], [13]. Besides the SFDA, the Food and Drug Administration (FDA) of the United States of America [14] and the European Medicines Agency (EMA) [15] have also accepted fingerprints as an alternative approach to evaluate the quality of herbs and their preparations. A fingerprint is a characteristic profile or pattern which chemically represents the sample composition and in which, usually, as much information as possible is reflected. Generally, fingerprints can be obtained using several techniques, both chromatographic and spectroscopic. Several analytical strategies to obtain and handle herbal fingerprints have already been listed and reviewed [16], [17], [18], [19], [20].

Spectroscopic fingerprints can be obtained using Infrared (IR), Raman or Nuclear Magnetic Resonance (NMR) spectroscopy. Mass spectrometric (MS) fingerprints also can be developed. Chromatographic fingerprints can be developed with Thin-Layer Chromatography (TLC), High-Performance Thin-Layer Chromatography (HPTLC), High-Performance Liquid Chromatography (HPLC), Ultra High-Pressure Liquid Chromatography (UHPLC), or Gas Chromatography (GC). Other, less common, techniques to obtain fingerprints are (Pressurized) Capillary Electrochromatography (pCEC), or Capillary Electrophoresis (CE). Most frequently, UV, DAD or MS detection is used.

Among the various techniques, chromatography is the one most commonly recommended for generating herbal fingerprints [12]. High-Performance Liquid Chromatography is one of the most often used techniques in traditional herbal medicines research (Fig. 1). The UHPLC approach has some advantages, such as a large decrease in analysis time and solvent consumption, and the possibility of obtaining high efficiency and resolving co-eluting compounds, while its drawbacks are an increased back-pressure and the availability of only few stable stationary phases. However, this technique nowadays is emerging in herbal fingerprint research.

Researchers have applied chromatographic fingerprint technology as a tool to evaluate the quality of herbal samples or their derived products [21], [22]. Identification of the samples must be done with great care to preserve the consumer's safety. It is done by eliminating the adulterations (mixed with different plants) or complete misidentifications (incorrect plant) as well as the samples with either low quality (containing lower concentrations of active compounds), or higher concentrations of contaminants (e.g. pesticides) [23], [24].

The purpose of this review is to identify herbs using fingerprints. Two groups of data treatment can be used for that purpose: similarity analyses (which are reviewed here) and classification methods (each identification is also a classification problem), which we do not review here. Though both spectroscopic and chromatographic fingerprints can be treated with the same data-handling techniques, our focus in this review is on the chromatographic. For the past decade or so, a great interest has been dedicated to the similarity analysis of fingerprints, which is important in the context of identification and quality control. These approaches can be used to evaluate and confirm that a sample is originating from the expected herb and to exclude that it is from another. Many methods/parameters have been recommended/applied to evaluate the similarity or difference between (two) fingerprints.

Both unsupervised data analysis techniques (e.g. Principal Component Analysis or Hierarchical Clustering Analysis) and similarity parameters were applied to fingerprint data [25], [26]. A similarity parameter is a statistic calculated to evaluate the (dis)similarity between two fingerprints. The sample is compared with one or several reference or standard profiles. A (dis)similarity can be quantified using either different distance (e.g. Euclidean distance) or similarity measures (e.g. correlation coefficient). Similarity parameters are usually considered more attractive because they are easier to understand and apply. Below, a number of such similarity parameters are discussed.

The main goal of this review is not only to discuss the (dis)similarities and exploratory analysis approaches used in similarity analysis, but also to provide an overview of recent applications in herbal medicine fingerprint case studies for data sets either small or large in number of samples (references).