Design of Experiments Assessment for the Determination of Moisture Content in Five Herbal Raw Materials Contained in Tea Products

Research interest in natural raw materials is rapidly growing due to the high demand for natural products like herbal teas. Their quality control has a direct impact on safety and efficacy. The aim of this study was to evaluate the impact of sample’s mass and temperature on moisture content in Camellia sinensis (Black tea), Cassia fistula (Senna), Chamaemelum nobile (Chamomille), Lippia alba (Juanilama) and Tilia platyphyllos (Linden) with a gravimetric method developed through a full factorial 32 DoE. A response optimizer was executed in order to establish the test conditions that allow obtaining a response according to a target value from a certified method. DoE’s ANOVA shows reproducibility for Camellia sinensis, Cassia fistula, and Lippia alba. Also, the method’s model is able to explain the response variability for all samples based on the R2 (adj). The composite desirability for the proposed conditions of analysis for the five herbal materials is satisfactory according to each target value. However, the lack of reproducibility in Chamaemelum nobile and Tilia platyphyllos and also, the response prediction problems according to the R2 (pred) for Cassia fistula and Chamaemelum nobile, suggest the execution of further studies for them. Therefore, the present method is considered to be adequate for the analysis of moisture content in Camellia sinensis and Lippia alba raw herbs.


INTRODUCTION
Currently, many plants and herbs are used for the manufacture of natural products and for the synthesis of novel medicines, due to their therapeutic properties.
Herbal raw materials are used almost by 80% of the world population, based on empiric knowledge or ancient traditional medicine (Asghar et al., 2016;Orphanides et al., 2013;Rodino & Butu, 2019;Singh et al., 2019). The high demand and the fact that herbs also contain toxic substances imply a critical role in quality control of these raw materials, since natural products must fulfill quality, safety and efficacy requirements (Carvalho et al., 2018;Mukherjee, 2019a).
In 1998, the World Health Organization (WHO) published the guideline 'Quality Control Methods for Medicinal Plant Materials', which addresses the need of quality assurance for natural raw materials with the aim of guaranteeing safety of herbal drug products.

Design of Experiments Assessment for the Determination of Moisture Content in Five Herbal Raw Materials Contained in Tea Products
According to this, not only a quantitative analysis of the active compounds has to be done, but also a qualitative analysis in order to evaluate physicochemical characteristics such as color, volatile compounds, ash values, moisture content and even taste and aroma (Mandal et al., 2017;Mukherjee, 2019b;Sahoo et al., 2010;Thomas & ElSohly, 2016;World Health Organization, 1998).
In this paper, special attention will be given to the In the manufacture of herbal drugs, alimentary and natural products; the starting materials are usually represented by fresh whole plants or their parts, which will face different manipulation processes in order to make them a suitable input material. One of these operations is the drying process. Direct drying methods, such as the gravimetric method, determine the moisture content by weighing a sample before and after drying, where all the weight loss is assumed to be explained by the removal of water (Mukherjee, 2019b;Zambrano et al., 2019 (Cheng et al., 2013;Kaur et al., 2014;Mandal et al., 2017;Teles et al., 2012).
However, when working with herbal raw materials it is necessary to define how dry is dry enough. A perfect way of achieving a target value is through Design of Experiments (DoE). Basically, DoE is a statistical tool used for the organization, conduction and interpretation of the results obtained through the execution of a small, but well designed, number of tests, so that useful information can be collected in the most efficient way (Zambrano et al., 2019).
The application of this multivariate analysis technique requires the level establishment of the analyzed factors.
Therefore, the selection of an experimental design depends on the previous knowledge and nature of the problem. Moreover, regarding their quality evaluation, independent variables are usually factors of the analytical method, while dependent variables are linked to the properties or parameters that reflect the performance of the raw material or product. As a result, a well-executed study can lead to the identification of the optimal conditions for the operation of a certain process or the best method for an analysis (Djuris et al., 2013).
The aim of this study was to evaluate the impact of Additionally, a response optimizer was executed based on a moisture target value provided by a private quality control laboratory with a certified gravimetric method, which ensures the quality of the herbal raw materials for the manufacture of a physicochemical stable product.

Materials
For the execution of this study it was decided to evaluate the moisture content of the five herbal raw materials shown in Table I, that are commonly used in Costa Rica for the manufacture of herbal teas.

Design of experiments
The DoE was done according to the following: 1. Controlled variables: Drying process and equipment, moisture balance, analyst and raw materials' batches.

RESULTS AND DISCUSSION
Pharmaceutical analysis is an important approach to develop novel analytical and control methodologies for the quality assessment of herbal products. In pharmaceutical and food industries such as the tea one, Moreover, optimal experimental designs are particularly important in these sciences, since they allow to identify combinations of factors for a proper estimate of the parameters of a model or system. In other words, they can lead to an optimum profile response by selecting the best set of processing conditions. The present DoE is conceived as a 3 2 full factorial design. The presence of a third level for two continuous factors helps to determine a quadratic relationship between the response and both input variables (Harbourne et al., 2009;Mead et al., 2012;Wagner Jr. et al., 2014).
Despite the number of runs to measure, error will always be present in any study. As the results obtained through the analysis of natural raw materials are influenced in some degree by noise or uncontrolled variables, there must be a strategy for the minimization of the effects caused by them. That's why randomization was employed in this study, because it allows a better statistical distribution of the error attributable to those factors among the results (Castillo-Henríquez et al., 2019a;Mead et al., 2012;Reichert et al., 2019).

Preliminary aspects
The main objective of this DoE was to establish the best conditions for the determination of the moisture content in herbal raw materials, based on a target value.
However, in order to discuss the developed DoE, it must be addressed based on the previous operations that led to the definition of the input variables and the response of interest, even though those processes are not part of the design. The investigation's flow diagram is presented in  As can be seen in the flow diagram presented in Figure 1, initially we have the drying and blending process.
Drying is the most critical operation since it has been found that many of the enzymes responsible of the  Mizukami et al., 2006;Toontom et al., 2012).
On the other hand, blending has a great impact on the homogeneity of moisture distribution among the whole crude drug that is being processed. That is explained basically because only the region that is in contact with the heated air flow is suffering the removal of moisture or water, while the other sections are stuck to each other and are transferring their moisture. In addition to that, individual parts of herbs used for the manufacture of tea products vary in shape, size and consistency, so it is reasonable to find differences in terms of moisture content. As a consequence, we addressed that situation by analyzing three different amounts of mass consisting of 1, 2, and 3 g, with the aim of determining whether this factor is significant, and to reduce waste as well (Chan et al., 2012;Fomeni, 2018;Schinabeck et al., 2019).  Figure 2 to 6, respectively.    The first herbal raw material analyzed was Camellia sinensis. In this case as can be seen in Table III, blocks are not significant, so the method for this herbal represents a good reproducibility. In contrast, both factors and the interaction between them have values lower than 0.05, thus they are significant. However, for situations like this when an interaction is presented, no matter how many terms are statistically different; the interaction is the one that will govern the effect on the response. Whereas, it is said that two factors interact significantly on the response variable when the effect of one depends on the level at which the other one is (Lin et al., 2015;Megías-Pérez et al., 2019).

Analysis of variance
According to the ANOVA on Table III However, it is important to take into consideration that these limits affect the composite desirability, which is a parameter that evaluates the way in which the proposed model configuration optimizes a response. The composite desirability has a range of 0-1, where 1 represents the ideal situation, while 0 indicates that the output variable is outside of the acceptable limits (Gutiérrez-Pulido & De la Vara Salazar, 2008;Reichert et al., 2019).
An important factor to take into consideration in the optimization design is the weight; this determines how the composite desirability is distributed in the interval between the lower or upper limit and the target value. It is possible to choose between 0.1, 1, or 10. A value lower than 1 means that less emphasis is placed on the target value, 1 is the neutral configuration which gives the same importance to the target and to the limits and finally, a value higher than 1 emphasizes more on the objective which makes difficult to achieve the optimization. Since the present study involves natural raw materials, there's a certain variability that can't be controlled, so we worked with a weight of 1 (Mead et al., 2012;Wagner Jr. et al., 2014). The summary of DoE's results is presented in Table IV. As can be seen in Table IV (Wagner Jr. et al., 2014).

Model summary
The correlation coefficient (R 2 ) and the adjusted correlation coefficient (R 2 adj) allow to measure the overall quality of the regression model. These coefficients make a comparison of the variability explained by the model against the total variation. In general, for prediction purposes a R 2 adj of at least 70% is recommended. Such statistic is preferred over the R 2 because the latter is falsely increased with each term incorporated into the model (Arvidsson & Gremyr, 2008). Detailed information about the model summary is presented in Table V.

CONCLUSION
The worldwide expansion in production and use of natural products like herbal teas has made their quality, efficacy and safety a major concern for the health authorities. As a result, pharmaceutical analysis introduces itself as a solution for the development of novel quality assessment and control methods. The gravimetric method developed through a DoE for the evaluation of the moisture content, showed reproducibility for Camellia sinensis, Cassia fistula, and Lippia alba. An adequate approximation to the target value based on the composite desirability was done for the five herbal materials and the confidence interval for their response was established in order to guarantee a physicochemical parameter for stability. However, the lack of reproducibility in Chamaemelum nobile and Tilia platyphyllos and also, the response prediction problems according to the R 2 (pred) for Cassia fistula and Chamaemelum nobile, suggest the execution of further studies for them. Therefore, the present method is considered to be adequate for the analysis of moisture content in Camellia sinensis and Lippia alba raw herbs, for which a robust experimental design is recommended as a final step before its approval.

ACKNOWLEDGMENT
No potential conflict of interest was reported by the authors.