Introduction

The use of herbal remedies as first choice in the empirical treatment of minor health conditions is widespread worldwide today.

The World Health Organisation (WHO) estimated that about 70–80% of the world’s population trusts alternative therapies, in which herbal medicines are predominantly used [1]. These percentages are expected to increase with the globalisation process and given the current socio-economic conditions. Most of these plants are grown in the wild, with only a few being cultivated. This requires rigorous toxicological control of herbal products for therapy, including pollutants absorbed by the plants from the environment in which they grow. Otherwise, instead of the expected benefits, herbal cures could cause serious damage to human health.

Medicinal plants can be easily contaminated with toxic pollutants from soil, water, and air. Soil is usually contaminated by intensive agricultural activities using pesticides to control pests, by various industrial activities, by rainwater, or by atmospheric dust. The accumulation of pollutants such as toxic elements and pesticides in the soil is increasing with the accelerating growth of pollution in the natural environment [2].

The uptake of toxic elements from polluted soils by plant roots depends on both soil characteristics (pH, organic load) and species [3]. These are metals that are more rapidly taken up by the plants, such as Cd, Tl, and Zn, which have a higher transfer coefficient. Other metals such as Cu, Cr, Co, and Pb are strongly bound to soil structures and have low transfer coefficients [4].

Some of these elements, when in concentrations above the permitted limit, can cause cancers and central nervous system damage (Hg, Pb, and As), while others can cause liver or kidney damage (Pb, Cd, and Cu) or affect the development of bones and teeth (Ni, Cd, Cu, and Cr) [5, 6].

Organochlorine pesticides (OPCs) have been widely used over the last century for agricultural and non-agricultural purposes worldwide. Even though they have been banned in most countries, due to their high persistence in the environment, their residues are still found in many matrices, including medicinal herbs [7].

Pesticides contained in medicinal herbs from contaminated soil and water can also lead to hepatotoxicity, nephrotoxicity, developmental disorders in children, and respiratory and cardiovascular system damage [8, 9].

Taghizadeh et al. [10] evaluated the potential poisoning risk to consumers of some food products (dates), in which the authors identified 22 pesticides, 6 toxic elements, and 16 polycyclic aromatic hydrocarbons.

A study conducted by Luo et al. [11] on medicinal plant crops in China showed that more than half of the samples analysed had pesticide content above the European Pharmacopoeia (EP) limit, with some of the products contain 35 types of banned, highly toxic pesticides detected in concentrations well above the maximum permitted level.

As for toxic elements contained in herbal therapeutic products, Yang et al. [12] quantified the metals contained in 279 traditional Chinese herbal remedies; the authors reached the alarming conclusion that some of the analysed products contained amounts above the permitted limits of Pb, As, Cd, and Hg. Other authors, Filipiak-Szok et al. [13], also investigated the content of toxic elements in European herbal medicines and herbal nutritional supplements and showed low levels of contamination. Also, Kulhari et al. [14] investigated the content of nine mineral elements (Mn, Cr, Pb, Fe, Cd, Co Zn, Ni, and Hg) in ten medicinal herbs collected from natural areas in northwest India, concluding that, except for Cr, all other elements were within permissible concentrations.

Currently, there are few studies attempting to correlate toxic elements content with pesticides in natural therapeutic products.

In this paper, was investigated the content of five elements (Hg, Pb, As, Cd, and Cr) and nine organochlorine pesticides (HCB, lindane, heptachor, aldrin, dieldrin, endrin, p,p’DDE, p,p’DDD, and p,p’DDT) in herbs available in Romanian pharmacies. The correlations between the concentrations of toxic elements and pesticides were determined by a statistical analysis. A hierarchical clustering method was used to detect the natural grouping between toxic elements and pesticides in herb samples. Clustering analysis is a statistical technique of partitioning variables or observations based on a given measure of similarity. The method was applied to illustrate the source of higher mineral levels in rice herbs [15], agricultural soil [16, 17], urban soil [18], or groundwater and soil samples affected by surgical instrument industries discharges [19].

Therefore, we consider that the evaluation of pollutants in herbs as raw materials applied in the pharmaceutical industry is necessary from a regulatory point of view. Considering the fact that these pollutants do not offer any therapeutic benefit, their content in the final products is important from a toxicological point of view.

Experimental

Sampling

The analysed samples consisted of 14 medicinal plants preserved by drying, known and commonly used in Europe, having monographs in the European Pharmacopoeia [20]. They were purchased only from Romanian pharmacies and come from certified Romanian producers. Table 1 presents information on the studied plants. In order to reduce particle size and satisfy the conditions for homogeneity, the samples were ground in an electric grinder, using a grinding time of 2–3 min. Once plants were powdered, they were kept in capped polypropylene flasks until analysis.

Table 1 General information on analysed medicinal herbs
Full size table

Toxic Element Analysis

The energy dispersive X-ray fluorescence (EDXRF) is a technique used to identify mineral elements, being considered a simple and non-destructive method for multi-element analysis [21]. The concentrations of mercury (Hg), lead (Pb), arsenic (As), cadmium (Cd), and chromium (Cr) were determined according the method proposed by Jyothsna et al. [22] using EDXRF technique (XEPOS C spectrometer, Na-U, Spectro Germany). The instrument is equipped with an X-ray tube using Rh anode and a silicon drift detector (SDD) operated at 50 kV and 1000 uA. The quantitative analysis is carried out using the X-LAB Pro software. The accuracy and precision of the EDXRF analysis were checked by testing a certified reference material (NIST SRM 1571-Orchard leaves). The results for the recoveries ranged from 90 to 104%, which indicated good correlation between the certified and measured values.

Organochlorine Pesticide Analysis

Two grams of homogenised sample was placed in a 50-mL recipient. Thirty millilitre hexan were added, and the mixtures were submitted to microwave extraction for 50 min at 120 °C. Subsequently, the extracts were cleaned up and fractionated by passing through a florisil column, previously conditioned with solvent. Extracts were then concentrated using the Kuderna-Danish concentrator to 1 mL before injection in gas chromatograph.

The analytical determination of organochlorine pesticide content was conducted by the gas-chromatographic method with a Perkin Elmer gas chromatograph CLARUS 500, equipped with electron capture detector. This model was equipped with a Perkin Elmer column, Elite 5MS having the following dimensions 30 m × 0.25 mm × 0.50 μm, and temperature range −60 to 330 °C. The carrier gas is helium; sample injection volume is 1.0 μl. Detector and injector temperature was initially set at 180 °C for 0 min, ramped 1–7 °C/min to 230 °C for 10 min, and ramped 2–15°/minute to 250°C for 10 min.

The identification of each compound was done by comparing the retention time of the peaks in the chromatograms with the retention time of the standards. Quantification was carried out by interpolation from the calibration curves, obtained by analysis of five different standard concentrations. The correlation coefficients derived from the linear regressions were higher than 0.9800, showing the relationships between peak area and analyte concentration. The detection limits were ranged between 0.002 and 0.003 μg·g−1, and recovery percentages corresponding to each compound were found in the range 89–103, indicating the accuracy of the method.

Statistical Analysis

The statistical software used in the analysis is R version 3.4.3 [23]. The Pearson correlation matrix between metals and pesticides was computed using the function cor, while the function hclust was applied for the hierarchical clustering based on the correlation matrix. Both functions are found in the R package stats. The chosen similarity measure for clustering is:

$$d\left(i,j\right)=1-\left|{\rho}_{i,j}\right|$$
(1)

where |ρi, j|is the absolute value of the Pearson correlation coefficient between two elements i and j and. Further properties of this similarity measure can be found in study of Gu and Wang [24]. To decide the optimal number of clusters, the function kgs in R package maptree was used to compute the Kelley-Gardner-Sutcliffe penalty function. For the cluster with the larger number of nonzero values, the statistical significance difference between the components of the clusters was assessed based on Kruskal-Wallis Rank Sum test at 5% level, implemented in kruskal.test in the package stats. Before applying the test, the normality of the samples was statistically checked using the Shapiro-Wilk Normality test shapiro.test at level α = 0.05. Furthermore, the Wilcoxon Rank sum test with Benjamini and Hochberg correction [25] pairwise.wilcox.test was employed to decide which element in the cluster is statistically different from the others.

Results and Discussion

Toxic Elements in Medicinal Herbs

A number of mineral elements important for nutrition can accumulate in medicinal herbs [26], as well as toxic elements such as Cd, Hg, and Pb, which are not directly used by plant and are harmful to human health [27]. Table 2 shows the results of determinations made on 14 of the most popular herbal medicinal products sold in Romanian pharmacies. As show in Table 2, the herbs contained toxic elements in the following order Cr > Pb > Cd > Hg > As.

Table 2 Toxic elements content in herbs (mg·kg−1)
Full size table

Arsenic is considered a non-essential element. Its contamination is caused by geological sources, pesticide application, and some industrial processes. In this study, As was present only in Calendula officinalis L; in the rest of the herbs, arsenic was not detected, indicating that the activities carried out in the areas where plants were harvested do not lead to arsenic accumulation in the soil. Similar results were reported by Sindhu and Beena [29] in medicinal herbs (Aloe vera L) available in the local market.

Cadmium is an element with unknown essential functions in both plants and humans [30]. Major sources of Cd in soil and plants are phosphate fertilisers, sewage sludge application, and fossil fuel combustion. It has been reported that the toxicity of Cd on the plants can be observed in the form of stunt growth [31]. In this study, the level of cadmium in 64% of herbs exceeded the WHO maximum permissible limit of 0.3 mg·kg−1 [28], while in the rest of plants, the content of this element were below the detectable minimum values (<0.001). Similar results of high levels of cadmium have been reported in Iranian medicinal herbs and some medicinal plants commercialised in Turkey [32, 33].

Lead is known as the most common and stable toxic element in nature, being very dangerous for plants, animals, and organisms [34]. Continuous fertiliser application, fuel burning, and sewage sludge are major sources leading to increased Pb pollution. The WHO maximum permissible limit of lead in medicinal herbs is 10 mg·kg−1 [28]. In our study, Pb was detected in all herb samples and the values were around 1 mg·kg−1. It is possible that the low value of the soil-plant transfer coefficient of this metal leads to the accumulation of lower amounts of Pb in the aerial parts of plants. It was reported that some of factors such as root surface area and rood exudation can affect the availability and uptake of Pb [35]. Contrary to the data from this study, high level of Pb has been reported in medicinal plants in Jordan and traditional herbs consumed in the United Arab Emirates [32, 36].

Chromium is another toxic element, being released by steel industries, sewage sludge applications, and tanneries [37, 38]. The permissible level for this metal in raw herbals is 2.0 mg·kg−1 [28]. In the present study, 57% of samples contained Cr concentrations above the permissible limits defined by WHO. The highest contents have been obtained for Centaurea cyanus L. and Taraxacum officinale Webb, while the lowest ones for Rosa canina L.

Due to its multiple uses (gold mining, batteries, industrial products, and pesticides), mercury is accumulated in different sites and represents a global pollutant. It has been described that the major symptoms of Hg toxicity in plants are decrease of growth process, reduced root development, and inhibition of photosynthesis activity [39]. In the investigated samples, 28% of the plants contained Hg above the limit proposed by WHO, of 0.2 mg·kg−1. Previous researchers, Sindhu and Beena, [29] and Begum et al. [40], reported the presence of mercury below limit levels in medicinal plants.

Minerals can get into medicinal plants and their products from contaminated agricultural sources (soil, water, agrochemicals), during processing or deliberate addition for their supposed medicinal value [41].

In the present work, the data show that toxic elements were present in varying concentrations in the herbs, in some cases exceeding the maximum permitted levels for medicinal plants [28]. Similarly, ten medicinal plants analysed by Ajasa et al. [42] collected from different areas of Ogbomoso (Nigeria) showed variable element levels. In addition, Gajalakshmi et al. [43] studied the presence of toxic elements in some medicinal herb species and revealed similar results. They reported that Cr content was at higher level in all plants.

Some researchers have been reported that these content differences can be attributed to the differences between the absorption and translocation capacities of minerals in plants [36, 44]. The uptake, accumulation, and amount of metals in plants can be influenced by a number of factors, such as atmospheric (traffic density, metal smelting operations), bioavailability of metals in the soil (through the addition of pesticides, sewage sludge or manure), the nature of the soil in which grasses grow (pH, organic matter), plant species (plant maturity, time of harvest), and plant processing conditions [14, 45, 46].

In their research, Srivastava et al. [47] compared the mineral content of different species of herbs collected from their natural habitats and from market samples, in India. Results showed that market samples were more contaminated with toxic elements than natural samples, which may indicate contamination during processing or storage.

At the same time, there are studies which indicate that metal content varies in different parts if the same plant [48]. Thus, Xu et al. [49] showed that accumulation of metals is higher in the leaves than in the root, while Bawa et al. [50] reported higher concentrations of metals in the root than in the shoots.

However, the risk of toxic elements entering a food chain depends on the metal mobility and availability in the soil [51]. In soil, metal cations are bound to negatively charged particles such as clay and organic matter; when metals detach from these particles and enter the soil solution, they become bioavailable with the potential to accumulate in plants and other soil-dwelling/growing organisms [52].

In order to favour the decontamination process of polluted soils, there are now concerns about phytoremediation by growing species capable of extracting and concentrating certain toxic elements [2]. The example of plants of the family Asteraceae, alternatively Compositae, capable of accumulating and concentrating Cd, Zn, and Cu from soil is known [53].

These results confirm these hyperaccumulative properties of the plants of this family, as both Centaurea cyanus and Taraxacum officinale were found to contain the highest amounts of toxic elements of all the plants analysed, being the most popular and frequently used medicinal plants. There are several plants of the Asteraceae family frequently used in therapy (e.g., Tagetes patula, Achillea millefolium, Cichorium intybus) that could have a decontamination effect through bioconcentration of toxic elements. Therefore, it becomes necessary to introduce control protocols on the toxic element content of herbs and standardised plant extracts for therapeutic use [54]. In addition, the investigation of the specific bioaccumulation properties of each of the herbs used in therapy would be of great value for the improvement of the control protocols developed by the EP [20].

In the case of Taraxacum officinale Webb species, studies show that this plant not only accumulates and concentrates toxic elements in soils but its growth and mineral accumulation are stimulated by the addition of fungicides such as thiuram [55]. In nature, fungi and microorganisms are largely responsible for the immobilisation of toxic elements in soil. The application of fungicides thus facilitates the release of these metals, which become bioavailable to plants growing on these treated soils. Recent studies on Asian phytotherapeutic products have shown that toxic elements (As, Cd, Pb) are present in the composition of these products, but below the toxicity limit [56].

Organochlorine Pesticides

OCPs have been widely used in the last century for agricultural and non-agricultural purposes worldwide. Even though they have been banned in most countries due to their persistence, OPCs are still found in areas where they were previously used and often contaminate plants growing nearby. Some of these substances are also still used for public health purposes, for example to control disease vectors such as mosquitoes and are often applied near agricultural fields. Pesticide residues can then pass through the air to crops, contaminating medicinal plants growing in nearly fields.

OPCs contain in their structure carbon, chlorine and hydrogen, and the carbon-halogen bond being highly resistant to degradation. Therefore, organochlorine compounds are degraded very slowly and may remain in the environment and inside organisms after exposure for a longer time [57].

Table 3 presents the content of organochlorine pesticides in the herbs selected for the study as commonly used by the population. The results show the presence of OPCs, such p,p’DDD, p,p’DDT, aldrin, and dieldrin in concentrations above the maximum permitted limit established by the European Pharmapoeia (EP), in Hypericum perforatum, Crataegus monogyna, and Epilobium parviflorum herbs [58]. These plants are known to grow in natural areas, and contamination may be due to plant-specific bioaccumulation processes. The case of Epilobium parviflorum species, which became more abundant and vigorous after herbicide (glyphosate) application, is presented in the literature by Matulevičiūtė, [59].

Table 3 Organochlorine pesticides (OPC) in herbs (mg·kg−1)
Full size table

Profile of individual pesticide was γ- HCH (lindane) > p,p’DDT > p,p’DDD > HCB > endrin > aldrin > dieldrin> heptachlor> p,p’DDE. Environmental factors can influence their levels and distribution. Thus, it has been reported that the application of the pesticides in the wet season leads to an easier uptake by the plant [60]. Furthermore, different physicochemical characteristics of soils (organic matter, microorganism activity), as well as properties of pesticide compounds (hydrophobicity), can influence the degradation, adsorption, and bioaccumulation of pesticides in plants [61]. It was reported that DDT is more hydrophobic than HCH; hydrophobic compounds are strongly bound to root and soil organic colloidal surfaces [62].

Some compounds such as heptachlor, dieldrin, and pp’-DDE were detected in high concentrations in some plants (Epilobium parviflorum Schreb, Hypericum perforatum L., Xanthium spinosum L.), while in others they were in very low levels or not detected. This indicates that some herbs came from areas with lower levels of contamination, or their processing may have contributed to lower pesticide concentrations. Plant processing such as drying, cleaning, grinding, and storage has been reported to reduce pesticide residues in final products [63].

Of the HCHs group, the γ- HCH isomer (Lindane) was detected in most herbs, being above the limit established by EP, in 78% of the plants (Table 3). The results showed variations in lindane concentration from 0.06 ± 0.00 mg·kg−1 to 5.30 ± 1.24 mg·kg−1, with a mean value of 1.34 ± 0.09 mg·kg−1, depending on the degree of exposure of herbs to this compound. The primary source of lindane in the environment is associated with contaminated water runoff containing the compound; it is associated with the management of common insecticides on agricultural land (vector control) and the use of animal pest control products. The results obtained by us are lower than those reported by Murtaj et al. [64] in medicinal plants collected in Albania, where α- and β-HCH isomers predominate. The presence of γ-HCH isomer in most of the herbs analysed can also be confirmed by evaluation the ratio α/β-HCH. Thus, the ratio of α/β-HCH > 3 is an indication of an input of technical HCH and long-range atmospheric transport and deposition, while a ratio of close to <1 is characteristic of lindane sources [65]. In the present study, the α/β-HCH mean ratio was closed, indicated the presence of lindane as source of HCHs.

Due to its persistent nature and hydrophobicity, aldrin is readily metabolised to dieldrin. Aldrin has low phytotoxicity; plants are only affected at high level of concentrations. Dieldrin was present in 48% of the samples, and higher amounts were identified in Artemisia absinthium L., Hypericum perforatum L., and Epilobium parviflorum Schreb.

Hexachlorobenzene (HCB) is considered a very persistent compound due to its chemical stability and resistance to biodegradation. In the soil, HCB binds strongly to the organic part, accumulating in plants and crops. Although HCB was detected in most samples, its concentration was above the permissible limit of 0.1 mg·kg−1 [58], in 14% of the plants. Its commercial production has been banned in the most of the world; however, recent sources of HCB may include waste and sewage sludge incineration processes, production of chlorinated organic solvents, and fossil fuel combustion [66]. It has been mentioned that currently, HCB contamination mainly comes from chemical production processes (where it is used as a chemical intermediate or polymer additive) and incineration processes [67].

Heptachlor is a pesticide that has been widely used mainly against termites and soil insects. It is metabolised by soil bacteria and fungi into different products via independent metabolic pathways [68]. A widespread reaction in the environment is the conversion of heptachlor to the more persistent heptachlor epoxide [69]. In the analysed plants, heptachlor was detected only in Epilobium parviflorum Schreb., at levels above the permissible limit [58].

p,p’DDT was detected in 36.66% of the analysed herbs, in amounts higher than its metabolites p,p’DDD and p,p’DDE. Using the ratio between DDT and its metabolites, it is possible to identify the possible source of DDT; thus, a DDT/DDD + DDE ratio higher than 1 indicates recent use of DDT [70].

In the present study, the ratio DDT/DDD + DDE in 10 herbs were 0 as there was no detection of DDT in those plants. The rations in the other samples (Hypericum perforatum L., Crataegus monogyna Jacq., Xanthium spinosum L.) exceeded 1 value; this indicates that there was recent input of DDT into the agricultural areas. Also, the ratio of parent DDT to its metabolite DDE, higher than 1 in Hypericum perforatum L. and Xanthium spinosum L., plants, indicates the presence of DDT in the last years [71]. At the same time, the ratio of DDD to DDE, higher than 1 in Hypericum perforatum L. and Artemisia absinthium L species, shows anaerobic degradation of DDT [72].

The results of the present study showed higher detection of OPC residues in medicinal plants than the data reported in Ethiopia and Iran [73, 74] and lower than those in China [75].

Correlation Coefficient Matrices and Cluster Analysis

Cluster analysis was used to group the metals and pesticides into groups with homogeneous structures. The technique reveals hidden clusters among the objects, based on a similarity measure. The advantage of clustering analysis is that the number of groups does not need to be known a priori, but it is instead obtained as a result of some exploratory techniques [76].

In this analysis, the aim was to empirically discover which groups of metals and pesticides are likely to be found together in herbs. The clustering is performed using the distance (1) in section Statistical Analysis. In this way, elements that are highly correlated, negative or positive, will be closed in the distance d(i, j) and thus will belong to the same cluster. Further properties of this similarity measure can be found in study of Gu and Wang [24]. Since Pearson correlation is invariant under linear transformation, the data were scaled prior to computing the distance. Table 4 shows the correlation matrix for the metals and pesticide levels in herb samples. The statistically significant values, corresponding to p-values below 0.05, are indicated in bold. Based on the distance matrix obtained, a hierarchical clustering method with complete linkage was used to group the elements. The method assigns first each element in its own cluster and then, iteratively, it merges similar clusters until there is only a single cluster. The function hclust outputs a binary tree with elements as leaves (Fig. 1).

Table 4 Correlation coefficient matrix for metals and pesticides found in herb samples (n =14)
Full size table
Fig. 1
figure 1

Kelley-Gardner-Sutcliffe penalty for different number of clusters. The optimal number of clusters is marked in red

Full size image

Since the number of clusters is not known a priori, the Kelley-Gardner-Sutcliffe penalty function helps to decide the optimal number of clusters in the hierarchical cluster tree [77]. The procedure kgs generated for a range of clusters k from 2 to 14, the corresponding penalty function. The optimal number of clusters is the one that attains the minimum penalty. Figure 1 shows the result of the Kelley-Gardner-Sutcliffe penalty function procedure and the optimal k = 4. Using 4 clusters, the clustering analysis is shown in Fig. 2, where the red rectangles indicate the structure of each group.

Fig. 2
figure 2

Optimal number of clusters and the structure of each cluster

Full size image

An important step of our research was to study the correlations between selected pesticides and toxic elements in the analysed herbs. Numberless studies have reported that some metals exhibit catalytic properties and can affect the behaviour of co-existing pesticides in soil [16]. To the best of our knowledge, there is no such information at the herb level. In this study, the obtained data show a strong positive correlation between Endrin and As with r-value of 0.96, indicating an increase in Endrin concentration in the presence of higher As concentration for Calendula officinalis L. Similarly, a significant positive correlation was obtained between HCB and Pb with r-value of 0.83. In about 72% of the herbs, it was observed that Pb concentrations increased with increasing HCB levels, thus indicating a similar origin of the pollution source. Thus, it can be mentioned that As, Pb, and even Cr stabilise some pesticides, such as endrin and HCB; in this way, they become more persistent, therefore leading to an increase/accumulation of pesticide content. For the other elements and pesticides, the correlation coefficients were not high, indicating that the correlations are not significant; these metals may facilitate pesticide degradation [15].

The strongest connection was observed between Heptaclor and p,p’DDT, corresponding to the shortest linkage. The significant Pearson correlation coefficient between them is around 0.98 (Table 4), indicating that small levels of Heptaclor in herbs are linearly correlated to small levels of p,p’DDT. This strong correlation between Heptachor and p,p’DDT indicates recent contamination. In Epilobium parviflorum Schreb, both compounds showed concentrations above the maximum permitted limit [57]. Similarly, a strong positive correlation was observed between Aldrin and p,p’DDD with r-value of 0.63.

No significant correlation was observed between the other selected pesticides, showing that the pesticide sources were the same. This may be due to the fact that the herbs show different vegetation periods, hence different contamination. Some pesticides are applied at different stages of crop production, so they reach the soil and plants at different times.

The first cluster consisting of Pb, HCB, Cr, and Hg has a linkage distance around 0.5. Within this cluster, the toxic elements and the pesticide are positively correlated, suggesting that the existence of Pb is correlated to the presence of HCB, Cr, and Hg in herbs. These correlations indicate that the compounds have similar emission sources; even if HCB is not used directly, it may be a by-product of industrial waste incineration processes or sewage sludge use. The clusters formed by Cd, As, and Endrin, on the one hand, and dieldrin, heptachlor, and p,p’DDT, on the other hand, are closer to each other than any other cluster. The relationship between these compounds, especially in the case of Epilobium parviflorum Schreb and Calendula officinalis L, shows a contamination of the area where the plants grow. The geometry of the dendrogram at the top level indicates that the cluster composed by aldrin, p,p’DDD, lindane, and p,p’DDE is substantially different from the other 3 clusters.

The first cluster (Pb, HCB, Cr, and Hg) contained the smallest number of zero values, and it was further analysed to detect any statistically significant difference between the medians of each element. First, the Shapiro-Wilk normality was applied to each component of the cluster to test the null hypothesis that the data are normally distributed. Since all p-values obtained were below the significance level α = 0.05, the null hypothesis can be rejected at 5% level. Since the normality assumption did not hold, the Kruskal-Wallis test was applied to test the null hypothesis H0 : medPb = medHCB = medCr = medHg, where med is the median of the data. The alternative hypothesis is that at least one median is different. The test kruskal.test produced a p-value of around 0.001181 below the significance level 0.05, indicating that there is at least one component in the group with the median statistically different from the others. In the boxplots of the 4 elements, shown in Fig. 3, it can be seen that the medians and the variances vary significantly among the components, with the largest difference being observed for Cr. To test this observation, a pairwise comparison was performed using the Pairwise Wilcoxon Rank Sum test with the Benjamini and Hochberg, [25], correction for multiple testing. The test confirmed that the values of Cr are statistically different from the rest of the group. However, the p-values obtained were approximately computed, due to the ties existing in the data.

Fig. 3
figure 3

Box plot for the elements in the strongest correlated cluster

Full size image

Conclusions

Among the toxic elements that have been determined, Cr, Cd, and Hg exceeded the maximum concentrations allowed for raw plant material in some of the analysed products, especially in Centaurea cyanus L., Verbascum thapsus L., Artemisia absinthium L., Taraxacum officinale Webb., and Betula pendula Roth. Also, some pesticides (endrin, aldrin, p,p’DDT, p,p’DDD) showed concentrations higher than European Pharmacopoeia limits in Hypericum perforatum, Crataegus monogyna, and Epilobium parviflorum. These results raise alarm bells for both producers and consumers. As a follow-up to these determinations, it will be necessary to quantify toxic elements and pesticides in infusions and decoctions prepared from these plant products, as recommended to be consumed for therapeutic purposes.

The results indicate the existence of significant correlation between HCB concentrations and Pb levels, as well as between endrin and As concentrations in Calendula officinalis, a plant obtained from crops. Furthermore, bioaccumulation studies may clarify the cause of the observed correlations.

Cluster analysis indicated that the elements detected are clusters in 4 groups, among which the strongest correlated group consists of Pb, HCB, Cr, and Hg. Within this cluster, the pairwise Wilcoxon rank sum test indicated that the values of Cr are statistically the most different from the other components.

Although the extent to which the transfer of the determined toxic elements and pesticides from dried plants into preparations for therapy has not yet been investigated, the results presented make it necessary to introduce the quantification of toxic elements and pesticides into the quality control protocols of products for phytotherapy.