Impact of succussion on pharmaceutical preparations analyzed by means of patterns from evaporated droplets

The aim of the present study was to investigate if patterns obtained from evaporating droplets of pharmaceutical preparations reveal the impact of succussion on such medicinal products. For this purpose, five pharmaceutical preparations (Echinacea 10−2, Baptisia 10−3, Baptisia 10−4, Luffa 10−4, and Spongia 10−6) were prepared according to the European Pharmacopoeia guidelines for the production of homeopathic remedies, in three variants each: with varying numbers of succussion strokes (i) 100, (ii) 10 (succussed samples), and (iii) zero (gently mixed, unsuccussed sample). System stability was studied by means of systematic positive control experiments. Patterns were evaluated by means of computerized image analysis regarding grey level distribution, texture, and fractality. For all investigated pharmaceutical preparations, significant differences were found between the succussed and gently mixed samples; whereas, all three samples (prepared with 100, 10 and zero succussion strokes) could be significantly differentiated for Luffa 10−4 and Spongia 10−6 for one image evaluation parameter each. Control experiments showed a reasonable stability of the experimental set-up.


Results
Qualitative description of the patterns. When analyzed by means of DEM, the five here investigated pharmaceutical preparations created visually recognizable and easily identifiable patterns (Fig. 1). In case of Echinacea 10 −2 , Baptisia 10 −3 , and Luffa 10 −4 the patterns consisted of dendritic, fractal-like structures placed in the droplet center. Echinacea 10 −2 created large, dense networks of very fine ramifications, Baptisia 10 −3 created rather small, roundly shaped structures, and Luffa 10 −4 structures made out of rather few and thick dendrites. Baptisia 10 −4 created unspecific patterns consisting of lines, smears, and, in some cases, single dendrites distributed all over the droplet. Whereas, Spongia 10 −6 created one to five filled, wavy forms per droplet, characterized by a concave and a convex side, placed near to each other and facing each other with the concave sides.
In general, in all pharmaceutical preparations, the impact of succussion on the patterns was visually perceptible in a varying, but rather small degree, and it seemed to decrease the structure's ordering.  (Table 1) application of succussion significantly increased the pattern evaluation parameters grey level distribution (GLD) and entropy (for N S = 10, 100). Also, the fractality parameters local connected fractal dimension (LCFD) and lacunarity increased following the succussion, however, LCFD only for N S = 10 and lacunarity only for N S = 100.
All systematic control experiments performed did not show any significance between the randomization groups for the main effects. Baptisia 10 −3 . As shown in Table 2, Baptisia 10 −3 succussed samples (N S = 100, 10) were characterized by significantly lower GLD, entropy, FP, and LCFD values compared to the unsuccussed samples, whereas lacunarity was significantly higher.
The systematic control experiments yielded a significant main effect for the parameters FP and entropy; the other three image analysis parameters did not show statistically significant differences between the randomization groups for the main effects. Thus, the main experiments' outcome regarding FP and entropy can be distorted due to chamber gradients (see below) and was excluded from further evaluation.
The systematic control experiments yielded a significant main effect for the parameter FP; the other four image analysis parameters did not show statistically significant differences between the randomization groups for the main effects. Thus, the main experiments' outcome regarding FP can be distorted due to chamber gradients (see below) and was excluded from further evaluation. the droplet residues dried on a glass substrate were photographed by means of dark-field microscopy in magnification 100×. In rows patterns obtained from Echinacea 10 −2 , Baptisia 10 −3 , Baptisia 10 −4 , Luffa 10 −4 , and Spongia 10 −6 are represented, whereas in columns varies the preparation manner consisting in the performance of vertical succussion strokes (N S = 100, 10, or 0). Pattern examples derive from main experiments (cf. Fig. 2) and were selected based on an image-analysis parameter value (grey-level distribution (GLD), lacunarity (LAC), and local connected fractal dimension (LCFD)), which is close to the mean value of the corresponding parameter. Different letter codes (a, b, c) are significantly different (p < 0.05).
No systematic control experiment performed showed significant main effects between the randomization groups. Spongia 10 −6 . In case of Spongia 10 −6 ( Table 5) parameter LCFD differentiated all samples and ranked them in the order N S 0 > 100 > 10; whereas lacunarity yielded significantly higher values only for the sample N S = 10. Parameter entropy differentiated the succussed samples (N S = 100, 10) from the unsuccussed ones. The parameters GLD and FP did not differentiate the samples.
No systematic control experiment performed showed significant main effects between the randomization groups.
Influence of succussion on DEM patterns. In order to summarize the experimental results, in Table 6 we considered as relevant only cases where the corresponding image analysis parameter was experimentally stable, which means that (i) the systematic positive control experiments were not significant, and (ii) in the F-test of analysis of variance of the main experiments the F value for the factor N S was higher than the F value for the interaction N S and day. This means that 16 out of 25 parameter/preparation combinations were retained.   www.nature.com/scientificreports www.nature.com/scientificreports/ Overall, we observed significant differences for at least one sample (N S = 100 or 10) compared to N S = 0 in all analyzed comparisons (100%, 16/16). In most cases (68.75% of comparisons, 11/16), the difference was between the succussed (N S = 100, 10) and unsuccussed (N S = 0) samples, without differentiating between the succussed samples. In 12.50% (2/16) of cases all samples (N S = 100, 10, and 0) could be significantly differentiated; in 12.50% (2/16) of cases the N S = 10 sample differed from the two others (N S = 0, 100); and in one case (6.25%, 1/16) the N S = 100 sample differed from the two others (N S = 0, 10).
Generalizing, it can be said that the GLD did not show a general direction of the influence of the succussion on the patterns; whereas in patterns from the succussed samples the pattern evaluation parameter entropy increased, and LCFD decreased. Lacunarity was the unique parameter showing significant differences for all pharmaceutical preparations and in general showed increased values in the succussed samples. FP differentiated the samples only in case of one remedy (Luffa 10 −4 ). climatized chamber gradients. Results of the F-test of the two-way analysis of variance with independent factors row and column from the systematic positive control experiments performed with Echinacea 10 −2 , Baptisia 10 −3 , Baptisia 10 −4 , Luffa 10 −4 , and Spongia 10 −6 are shown in Table 7. As it can be noticed, factor row showed significance for most image evaluation parameters of the patterns obtained from the five pharmaceutical preparations (14 results out of 25; 14/25), whereas factor column was significant only in one case (Luffa 10 −4 , parameter lacunarity). The interaction between factors row and column resulted also significant in 8/25 cases, however mostly with lower F values than those observed for factor row.

Discussion
The results of the present study show that in all five analyzed pharmaceutical preparations the succussion strokes applied during production significantly influenced the DEM patterns. It can be summarized that succussion induced the formation of structures characterized by a greater disorder (parameter entropy) and smaller complexity (parameter local connected fractal dimension), at the same time increasing the gaps between the structure elements (parameter lacunarity). In case of two preparations (Luffa 10 −4 and Spongia 10 −6 ), significant differences could be found between all samples (N S = 0, 10, and 100). The here chosen parameters have already been applied in structure analysis of patterns formed in course of phase transition of liquid pharmaceutical preparations 11 ; moreover, raw material surfaces present in pharmaceutical triturations were also analyzed by means of fractal dimension 18 .
DEM patterns in the here analyzed dilution range 10 −2 -10 −6 are in a first place a function of solute dry residue. Differences found between the patterns of succussed vs. not succussed samples might be linked with succussion-induced aggregation of large-size molecules 2,5 , or, in case of Spongia 10 −6 (consisting only of mineral substances, since the sponge is roasted) through the introduction of air bubbles and/or particle formation 2 .
Whereas the patterns of Echinacea 10 −2 , Baptisia 10 −3 , Luffa 10 −4 , and Spongia 10 −6 were concentrated in the central part of the droplet residue and fitted entirely on the photographed in 100× image, in case of Baptisia 10 −4 the structures were rather unspecific and distributed almost evenly through the entire droplet residue (Fig. 1). In order to keep the magnification equal in the whole experimentation series, the part to be photographed was chosen by the experimenter (based on a visual check of the pattern, the part with most evident structures was  Table 6. Graphical representation of relevant differences found in the image evaluation parameters in pharmaceutical preparations prepared with varying numbers of succussion strokes N S = 100, 10, or 0. Different letters (a, b, c) indicate significant differences at p < 0.05. LEGEND: GLD -grey level distribution; FPforeground pixels; LCFD -local connected fractal dimension; LAC -lacunarity.    (Table 7). In most cases (13/14) this systematic error could be successfully eliminated (Tables 1-5) by the application of a quasi-randomization design, consisting in the randomization of the samples only within the columns, keeping simultaneously an even distribution of the samples within the rows. In future experiments, however, a better isolation of the inner-chamber should be aimed at to improve the homogeneity of evaporating conditions. The influence of the factor day was significant in most of the here presented experiments (24/25 differentiation and 23/25 control experiments) (Tables 1-5). A significant influence of the experimentation day has been reported in many previous studies concerning methods based on phase-transition-induced pattern formation 11,13,14,[19][20][21][22] . This fact might be due to some day-to-day variations in the experiment performance or experimental conditions; or to other yet unknown and uncontrolled influences.
To conclude, we observed that the application of the droplet evaporation method on pharmaceutical preparations led to the creation of patterns revealing differences for the parameters grey level distribution, texture, and fractality, dependent on the application of succussion and the number of succussion strokes performed during the pharmaceutical processing.
In the present investigation we performed succussion by shaking the cylinder with the solution by hand freely in the air, with the cylinder being filled to about 2/3 of its capacity. This kind of succussion is a usual procedure applied by many producers of pharmaceutical preparations, however it is not completely standardized and might vary in velocity and dynamic when performed by different persons. Further DEM experiments should be conducted comparing the impact of different methods of succussion, considering besides the quantity of performed movements also their intensity and type of movement.
The here presented experimental protocol might constitute a fairly economic and quick tool to investigate the impact of agitation on solutions, which has great importance for fabrication and distribution of pharmaceutical preparations in general and which is addressed in many recent investigations. In particular, it might serve to compare the role of several factors known for being critical for the solution properties, like for instance the kind of induced flow (e.g. chaotic vs. ordered, vortex-like) 2,3,23 , different surfaces and coatings of the recipient's walls 24,25 , and different volumes of the headspace 2,26 . DEM might be applied alternatively or complementary to established analytical methods used for the characterization of succussed solutions, such as, inter alia, micro-flow imaging, dynamic light scattering, light obscuration method (serving for analyzing the formation of particles), size exclusion chromatography, and tryptic digestion/HPLC (for studying the aggregation of proteins), hydroxyphenyl fluorescein assay (analyzing the formation of free radicals), and fluorescence spectroscopy, Fourier transform infrared spectroscopy and differential scanning calorimetry (characterizing further the solution composition and thermodynamic characteristics) [1][2][3][4][5][6] . Comparison studies of DEM with these methods should be conducted to characterize the DEM patterns better in terms of specific solution properties. Furthermore, investigations on a possible link between the patterns and biological efficacy are needed.     , prepared from the 10 −1 dilutions by applying different numbers of succussion strokes (N S = 100, 10, or 0). These three variations of a given homeopathic preparation were analyzed in one experimental run, consisting of twelve slides with droplets deposited on them (Fig. 3). Four slides were used for each pharmaceutical preparation. The slides were distributed in a climatized chamber following a quasi-randomization design. Each main experiment had a corresponding systematic positive control experiment where the analyzed sample was prepared three times with N S = 10 and analyzed following the same quasi-randomization design as in the main experiment. All experiments were independently repeated three times. preparation of pharmaceutical preparations for analysis. 0.8 g of a pharmaceutical preparation in dilution 10 −1 was weighed and placed in a sterile glass cylinder (SBR-ET, Mix Cyl. 10 ml, B; Brand GmbH + CO KG, Wertheim, Germany) with stopper (untargeted volume 13 ml); subsequently 7.2 ml purified water according to Pharm. Eur. 9.4 12 ("purified water in bulk", X-SEPTRON LINE 10 VAL, BWT AQUA AG, Aesch, Switzerland) was added in order to reach a dilution of 1:9. The cylinder was closed tightly; 10 or 100 succussion strokes were applied by hand. The movement to achieve succussion was performed in the air without hitting against a firm base. For the unsuccussed samples, the content of the cylinder was mixed with a glass stirrer by performing circular movements in order to not create any foam. After the settling of any foam in preparations N S = 10 and 100, the cylinders were re-opened and 0.8 ml of the solution were taken for the preparation of the next dilution, as described previously. In this way three variants (N S = 100, 10, 0) of each preparation (Echinacea 10 −2 , Baptisia 10 −3 , Baptisia 10 −4 , Luffa 10 −4 , and Spongia 10 −6 ) were produced. All samples were prepared fresh for each experiment. The samples were not blinded.

Manufacturing of pharmaceutical preparations in dilution 10
Droplet evaporation method. Microscope slides (76 × 26 mm, pre-cleaned, cut edges; Thermo Scientific, Gerhard Menzel B.V. & Co. KG, Braunschweig, Germany) were degreased by washing them with a dishwasher liquid, then thoroughly rinsed with hot tap water, and placed in 4 consecutive purified water baths. Each slide was wiped dry with a laboratory wiper (KIMTECH science, Kimberly-Clark Professional, Roswell, Canada) just before droplet deposition. 3 μl droplets of the tested pharmaceutical preparation were deposited on the slides in two parallel rows, 7 droplets per row, by the use of a micro-pipette of 20 µl capacity (Eppendorf Research Plus, Eppendorf, Hamburg, Germany).
Evaporation took place in an incubator (KBF 720, cooled incubator with controlled humidity system, WTB Binder Labortechnik GmbH, Tuttingen, Germany) with an inner plexi-glass-chamber with a semi-permeable cover placed on a vibration absorbing basis. The microscope slides with droplets were placed in the inner-chamber and left for evaporation in 26 °C and 44%rH for 1 hour. The slide distribution inside the chamber followed a quasi-randomization design in order to provide a uniform arrangement of the samples within the rows (Fig. 2). In case of Echinacea 10 −2 , Baptisia 10 −3 , Luffa 10 −4 , and Spongia 10 −6 , the 100X images included the whole structure formed inside the droplet; whereas, in case of Baptisia 10 −4 , only selected parts of the structure were included, chosen by the experimenter on the basis of density and intensity of forms. computerized pattern evaluation. Image analysis was performed with the software ImageJ (v. 1.50b) 27 with the plug-ins GLCM Texture 28 and Frac-Lac 16 . All 100× images were subjected to a background extraction by means of the sliding paraboloid with rolling ball radius set at 50 pixels ensuring same background throughout the image database. Consecutively the images were analyzed (i) for their grey-level distribution, (ii) after conversion into 8-bit type, by running the GLCM algorithm (considering distances between pixel pairs of 4 pixels and angles of 90°), for their texture (parameter entropy), and (iii) after conversion into binary, by means of Frac-Lac's DLC tool with odd sizes scaling method and size limits for the grid caliber series of minimum 4 and maximum 40 pixels, for the size of the structures (parameter foreground pixels), complexity (parameter local connected fractal dimension), and characterization of the gaps between the structure elements (parameter lacunarity). After conversion into binary, 68 Echinacea 10 −2 images could not be used due to a too dense ramification-network, and were excluded from fractality analysis. Whereas, in case of Baptisia 10 −3 and Luffa 10 −4 , fractal analysis was performed on images reduced in size to 500 × 375 pixel. Statistical analysis. The data deriving from the computerized image analysis were analyzed by means of a two-way analysis of variance (CoStat, v. 6.311) (CoHort Software, Monterey, USA) at alpha = 0.05 with independent factors number of succussion strokes (N S ) and day or row and column. An interaction term between the independent factors was included in the statistical model in order to assess stability and reproducibility. Distribution of data was checked by visual inspection. Slight deviations from normality were irrelevant due to the central limit theorem. Data-sets with larger deviations from normality were logarithmically transformed (log10); in total 18 data sets were transformed (Echinacea 10 −2 main/control study: FP, LAC/FP, LAC; Baptisia 10 −3 : FP, LAC/FP, LAC; Baptisia 10 −4 : GLD, FP/GLD, FP; Luffa 10 −4 : FP/FP; Spongia 10 −6 : GLD, FP/GLD, FP). Global significance was determined with F-tests. Pairwise mean comparison was performed two-tailed with the protected Fisher's least significant difference test (pairwise comparisons were evaluated only if the global F-test was significant at p < 0.05). This procedure gives a good safeguard against type I as well as type II errors, and thus balances well between false-positive and false-negative conclusions 29 . Results of the transformed data sets were back-transformed for presentation.

Data availability
The datasets generated and analyzed during the current study are available from the corresponding author on reasonable request.