Separation and Purification Technology 132 (2014) 126–137
Contents lists available at ScienceDirect
Separation and Purification Technology
journal homepage: www.elsevier.com/locate/seppur
Control the effects caused by noise parameter fluctuations to improve
pharmaceutical process robustness: A case study of design space
development for an ethanol precipitation process
Xingchu Gong, Yao Li, Zhengtai Guo, Haibin Qu ⇑
Pharmaceutical Informatics Institute, College of Pharmaceutical Sciences, Zhejiang University, Hangzhou 310058, China
a r t i c l e
i n f o
Article history:
Received 22 March 2014
Received in revised form 13 May 2014
Accepted 14 May 2014
Available online 22 May 2014
Keywords:
Design space
Process robustness
Ethanol precipitation
Danhong injection
Quality by design
a b s t r a c t
To provide a new method to develop a design space to improve the robustness of drug manufacturing
processes based on the control of noise parameter influences, the ethanol precipitation process of
Danhong injection was investigated as a sample. Water content in the concentrated extract (WCCE),
the concentration of ethanol (CEA), and the amount of ethanol added (AEA) are three adjustable parameters. The effects of refrigeration temperature (RT) was investigated on three levels for simulating its fluctuations. The models between parameters and process critical quality attributes (CQAs) were obtained
using a simplified central composite design with determination coefficients more than 0.84. The decrease
of RT led to lower active ingredient recoveries and higher DMR. The increase of CEA and the decrease of
WCCE caused more precipitation. The decrease of CEA or AEA resulted in higher active ingredient
recoveries. The design space was calculated using an exhaustive search-Monte Carlo method and normal
operation ranges were also obtained. The results of verification experiments agreed well with prediction
results. The proposed method can be used to develop a design space applicable in a larger scale manufacturing process with the negative effects caused by the fluctuations of noise parameters controlled.
Ó 2014 Elsevier B.V. All rights reserved.
1. Introduction
Quality by design (QbD) is a concept based on knowledge management and risk management [1,2]. It helps to develop new drugs
and improve drug quality control [3,4]. There are several steps to
implement QbD concept in drug manufacturing, including critical
quality attribute (CQA) definition, critical process parameter
(CPP) identification, risk assessment, design space development,
Abbreviations: AC, active ingredient content (mg g1); ACR, active ingredient
recovery (%); AEA, the amount of ethanol added (mL g1); ANOVA, analysis of
variance; ARD, average relative deviation (%); CE, concentrated extract; CEA, the
concentration of ethanol (g g1); CPP, critical process parameter; CQA, critical
quality attribute; DM, dry matter content (mg g1); DMR, dry matter removal (%);
DSS, Danshensu; EV, experimental value; FEP, first ethanol precipitation; HSYA,
hydroxysafflor yellow A; LA, lithospermic acid; MAS, mass of supernatant (g); MCE,
mass of concentrated extract (g); NOR, normal operating ranges; PV, predicted
value; QbD, quality by design; RA, rosmarinic acid; RSD, relative standard deviation;
RT, refrigeration temperature (°C); SaB, Salvianolic acid B; SP, supernatant; WCCE,
water content in concentrated extract (%).
⇑ Corresponding author. Address: Room 327, College of Pharmaceutical Sciences,
Zhejiang University, 866 Yuhangtang Road, Hangzhou 310058, China. Tel./fax: +86
571 88208428.
E-mail address: quhb@zju.edu.cn (H. Qu).
http://dx.doi.org/10.1016/j.seppur.2014.05.014
1383-5866/Ó 2014 Elsevier B.V. All rights reserved.
control strategy design, and continual improvement in drug lifecycle [5,6]. Design space development is a key component in the
implementation of QbD [7]. According to the ICH Q8(R2) guideline,
working within the design space will not result in drug quality
changes [5]. To establish a design space, quantitative relationships
between process parameters and CQAs must be obtained [8–13].
Response surface methodology is the most widely applied method
to develop a design space.
Recently, Rozet et al. define a design space as ‘‘a multivariate
domain of input factors ensuring that critically chosen responses
are included within predefined limits with an acceptable level of
probability’’ [7]. Because response surface methodology cannot
give any guarantee that CQAs will attain the defined criteria with
high probability [7], several other methods, such as the Monte-Carlo simulations and Bayesian modeling, are applied to calculate the
probability [7,14–16].
In the production of drugs, some CPPs are difficult to be controlled in a narrow range. For example, because of high control
expenses, refrigeration temperature (RT) of a precipitation process
is a parameter often fluctuates with seasons. While pH value is a
parameter often fluctuates because of the control difficulties
caused by big time lag. These CPPs can be considered as the noise
127
X. Gong et al. / Separation and Purification Technology 132 (2014) 126–137
parameters in drug manufacturing [17]. Contrarily, parameters
that can be easily controlled with acceptable costs are considered
as the adjustable parameters [17]. For a specific process, a design
space should only define the ranges for adjustable parameters. To
increase process robustness, the negative effects caused by the
fluctuations of the noise parameters must be controlled using the
adjustable parameters. However, to the author’s knowledge, there
is no published works focusing on the control of noise parameter
effects in the design space development.
In this work, a design space is developed aiming to increase process robustness by the control of noise parameter effects. The first
ethanol precipitation (FEP) process of Danhong injection is investigated as a sample. Danhong injection is a botanical injection clinically for the treatment of coronary heart disease, angina,
myocardial infarction, and cerebral diseases [18]. The sales of
Danhong injection have reached more than 4 billion RMB per year.
Phenolic acids and flavones are considered as the active ingredients of Danhong injection, such as Danshensu (DSS), rosmarinic
acid (RA), lithospermic acid (LA), Salvianolic acid B (SaB) and
hydroxysafflor yellow A (HSYA) [19]. Danhong injection is made
from Salviae miltiorrhizae Radix et Rhizoma (Danshen) and Carthami
Flos (Honghua) with a series of processes, including water extraction, concentration, ethanol precipitation, and adsorption. The
FEP process deals with the concentrated extract of mixed Danshen
and Honghua. Highly polar impurities in the concentrated extract
are usually removed in the ethanol precipitation process, such as
saccharides, salts, and proteins [20–25]. However, the losses of
active ingredients, including phenolic acids and flavonoids, are also
observed in published works [23,26,27]. Because of the important
impacts on drug safety and efficacy, the development of a design
space for the FEP process will help to improve the quality control
of Danhong injection [3].
The ethanol precipitation process is an easily operating process
without using toxic solvent [28,29]. However, the mechanism of
the ethanol precipitation process is usually very complicated when
dealing with the water extract of medicinal plants. Active ingredients may lose because of several different reasons, such as precipitation, degradation, or encapsulation [30]. Therefore response
surface methodology, such as central composite design or BoxBehnken design, was often applied to investigate the ethanol precipitation process [31]. Simple mathematical models can also be
obtained with response surface methodology [32,33].
In this work, a risk assessment was carried out to obtain the
CQAs of the FEP process. Experimental design was applied to establish the quantitative relationships among the adjustable parameters, the noise parameter, and the CQAs of the FEP process. The
probability to attain the defined criteria was calculated using the
Monte-Carlo method. A design space was obtained and verified
using the adjustable parameters to control the effects caused by
noise parameter fluctuations.
2. Materials and methods
acetonitrile was purchased from Merck (Darmstadt, Germany).
The HPLC-grade formic acid was obtained from ROE SCIENTIFIC
INC. The HPLC-grade ammonium formate was purchased from Alfa
Aesar China (Tianjin) Co., Ltd. All materials were used as received
without any further purification.
2.2. Design of experiments
In this work, four parameters of refrigeration temperature,
concentration of ethanol added (CEA), water content of concentrated extract (WCCE), and amount of ethanol added (AEA), were
investigated with a simplified central composite design. Table 1
shows the coded and uncoded values of parameters. The effects
of RT was investigated on three levels to simulate its fluctuations.
The central point was repeated for 3 times. Other points were
repeated for twice. Therefore a total of 47 experiments were carried out. The run orders are listed in Table 2. The ranges of the four
parameters were set based on production experiences. The experimental conditions for the verification of the design space are listed
in Table 3.
2.3. Experimental procedure
2.3.1. Preparation of concentrated extract
Reflux extraction was carried out to extract mixed 6 kg of
Danshen and 2 kg of Honghua with 80 L of distilled water for 1 h.
The extract then was filtered and collected. The extraction was
repeated twice. Two water extracts then were combined and concentrated under reduced pressure. After concentration, the water
content of the obtained concentrated extract was 26.8% ± 0.32%.
The concentrated extracts with higher water content were
obtained by dilution with deionized water.
2.3.2. Ethanol precipitation
The ethanol precipitation experiments were carried out in run
orders. Ethanol solution of designed concentration was added into
20 g of a concentrated extract in a conical flask under magnetic
stirring with a flow rate of 5 mL/min. After the addition of ethanol,
the flasks then were kept in a low-temperature thermostat bath
(THD-1008 W, Ningbo Tianheng Instrument Factory) for 20 h. After
that, the supernatants were collected and weighed. The contents of
DSS, LA, RA, SaB, and HSYA and dry matter in the supernatants
were determined.
2.4. Analytical methods
The concentrations of DSS, HSYA, RA, LA, and SaB were determined by high performance liquid chromatography analysis
according to the method published in previous work [34]. The
method is briefly described as follows. The HPLC system HP 1100
series (Agilent Technologies, Waldbronn, Germany) was equipped
with a Chemstation software (Agilent Technologies). The separations were carried out at 30 °C on a ZORBAX Eclipse Plus C18
2.1. Materials and chemicals
Danshen was purchased from Nepstar Drugstore (Zhejiang,
China), and Honghua was purchased from Daily Healthy Drugstore
(Zhejiang, China). Ethanol (>99%) was obtained from Tianjin
Damao Chemical Reagent Factory (China). Standard substances
including DSS, RA, and LA were purchased from Winherb Medical
S&T Development Co., Ltd. (Shanghai, China). SaB was obtained
from Chengdu Biopurify Phytochemicals Ltd. (Sichuan, China).
HSYA was purchased from Aladdin Industrial Inc. (Shanghai,
China). Deionized water was produced using a Mill-Q academic
water purification system (Milford, MA, USA). The HPLC-grade
Table 1
Coded and uncoded values of parameters for central composite design.
Coded
values
1.5
1
0
1
1.5
Uncoded values
X1 (WCCE, %)
X2 (CEA,
g g1)
X3 (AEA,
mL g1)
40.0
43.0
49.0
55.0
58.0
0.89
0.90
0.92
0.94
0.95
1.1
1.3
1.7
2.1
2.3
X4 (RT, °C)
5.0
15.0
25.0
128
X. Gong et al. / Separation and Purification Technology 132 (2014) 126–137
Table 2
Simplified central composite design and experimental results.
Run order Parameters
DMR, Y1 (%) ACRDSS, Y2 (%) ACRHSYA, Y3 (%) ACRRA, Y4 (%) ACRLA, Y5 (%) ACRSaB, Y6 (%)
X1 (WCCE, %) X2 (CEA, g g1) X3 (AEA, mL g1) X4 (RT, °C)
1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24
25
26
27
28
29
30
31
32
33
34
35
36
37
38
49
40
41
42
43
44
45
46
47
43.0
43.0
40.0
55.0
49.0
55.0
55.0
55.0
55.0
49.0
43.0
55.0
43.0
49.0
55.0
55.0
58.0
43.0
43.0
49.0
43.0
49.0
55.0
58.0
49.0
55.0
49.0
49.0
43.0
55.0
55.0
55.0
43.0
55.0
43.0
43.0
43.0
55.0
43.0
43.0
49.0
43.0
55.0
43.0
40.0
49.0
49.0
0.90
0.94
0.92
0.90
0.89
0.94
0.94
0.90
0.90
0.92
0.94
0.90
0.90
0.92
0.94
0.90
0.92
0.90
0.90
0.95
0.94
0.92
0.94
0.92
0.92
0.90
0.95
0.92
0.90
0.94
0.90
0.94
0.94
0.90
0.94
0.90
0.94
0.94
0.90
0.90
0.92
0.94
0.94
0.94
0.92
0.89
0.92
1.3
2.1
1.7
2.1
1.7
2.1
1.3
1.3
1.3
1.1
1.3
1.3
2.1
1.7
1.3
2.1
1.7
1.3
1.3
1.7
1.3
2.3
1.3
1.7
1.1
1.3
1.7
2.3
2.1
1.3
2.1
2.1
1.3
2.1
2.1
2.1
2.1
2.1
1.3
2.1
1.7
2.1
2.1
1.3
1.7
1.7
1.7
25.0
25.0
15.0
5.0
15.0
25.0
25.0
5.0
25.0
15.0
5.0
5.0
25.0
15.0
25.0
5.0
15.0
25.0
5.0
15.0
25.0
15.0
5.0
15.0
15.0
25.0
15.0
15.0
5.0
5.0
25.0
5.0
5.0
25.0
5.0
5.0
25.0
25.0
5.0
25.0
15.0
5.0
5.0
25.0
15.0
15.0
15.0
70.6
80.3
81.6
63.3
66.9
65.6
53.6
53.3
42.4
63.9
81.0
54.0
76.5
70.6
52.2
63.8
54.8
74.5
77.8
74.6
78.2
61.9
59.8
54.8
65.0
43.3
74.6
73.1
78.1
59.6
58.5
70.5
81.2
58.7
81.7
77.8
80.1
65.6
76.8
75.7
71.0
81.1
68.6
77.9
81.6
67.2
70.5
column (100 mm 4.6 mm, 1.8 lm) with a formic acid500 mmol L1 ammonium formate–water solution (0.5:10:90,
v/v/v) as mobile phase A and an acetonitrile–formic acid solution
(100:0.5, v/v) as mobile phase B in a gradient mode at a flow rate
of 0.5 mL min1. Detection wavelengths were 280 nm for DSS,
RA, LA, SaB, and 380 nm for HSYA. The injection volume was 10
lL. The linear gradient program was set as follows: 0–10 min, 2–
9% B; 10–13 min, 9–10% B; 13–20 min, 10–17% B; 20–37 min,
17–20% B; 37–47 min, 20–25% B; 47–50 min, 25–80% B. The programmed wavelength was set at 280 nm from 0 min to 15.9 min
and at 380 nm from 15.9 min to 17.9 min. The wavelength then
was set at 280 nm from 17.9 min to 50 min. Dry matter contents
of the concentrated extract and the supernatants were determined
using hot air drying at 105 °C for 3 h [35].
2.5. Data processing
Eqs. (1)–(3) were used to calculate WCCE, dry matter removal
(DMR), and active ingredient recovery (ACR), respectively.
WCCE ¼ ð1 DMCE Þ 100%
ð1Þ
47.4
40.8
35.4
59.3
51.1
56.8
62.9
61.3
69.4
49.8
37.7
61.7
45.1
49.5
65.1
59.2
64.5
42.0
40.5
46.9
38.7
51.3
58.6
64.2
50.1
70.0
45.5
50.6
42.7
57.9
63.9
51.6
35.9
64.6
39.1
44.3
40.1
57.5
40.8
45.5
50.2
39.2
54.9
39.0
36.8
53.7
50.9
41.7
29.6
27.5
58.8
50.6
54.9
65.7
64.5
74.7
51.3
30.1
64.7
36.0
45.0
67.5
56.9
66.1
38.6
35.4
41.3
32.7
44.1
60.5
66.5
51.7
73.3
40.7
43.7
34.4
59.7
64.3
49.3
29.1
63.8
27.5
35.1
30.2
54.1
36.2
36.9
45.9
29.2
51.5
33.4
27.5
52.5
46.5
63.3
49.9
55.9
75.7
69.5
71.8
77.7
76.3
80.1
68.7
59.6
76.5
65.9
65.8
76.1
73.3
76.2
61.3
60.5
65.2
60.9
66.8
70.6
79.1
70.2
81.1
69.9
69.1
67.6
79.2
81.6
73.0
63.2
83.6
59.3
64.0
60.7
72.7
63.6
71.4
75.6
64.2
77.5
63.8
62.5
72.1
71.8
42.5
22.3
27.5
53.8
49.3
48.7
63.4
62.7
70.2
52.5
32.7
63.4
35.7
42.1
62.7
52.1
61.6
38.1
36.3
39.7
35.2
39.8
54.7
63.1
52.8
70.5
39.7
35.9
34.3
58.8
57.6
46.5
30.6
58.4
26.8
34.8
29.0
49.2
36.5
37.5
45.8
28.4
47.1
34.5
29.1
51.3
46.6
45.1
25.1
31.0
56.9
52.9
52.1
65.7
64.8
70.7
54.5
35.9
65.4
39.3
45.6
64.4
55.0
63.7
41.8
40.5
43.9
39.3
43.5
57.1
65.4
56.2
71.2
42.7
38.2
37.5
60.2
57.8
48.2
33.4
58.3
30.2
38.5
32.0
51.5
40.1
41.2
49.5
31.6
49.6
37.3
31.4
53.4
49.1
where DM is the dry matter content and subscript CE refers to the
concentrated extract.
DMR ¼
1
DMSP MAS
100%
DMCE MCE
ð2Þ
where MAS and MCE are the mass of supernatant and concentrated
extract, respectively. Subscript SP refers to the supernatant.
ACR ¼
ACSP MAS
100%
ACCE MCE
ð3Þ
where AC refers to the active ingredient content.The experimental
data were analyzed using the Design-Expert 8.0.6 software (StateEase Inc., MN, USA) to obtain response surface models. As the refrigeration temperature was investigated on three levels, the quadratic
term of the refrigeration temperature was not included in modeling.
The basic form of the mathematical model was described as follows,
Y ¼ a0 þ a1 X 1 þ a2 X 2 þ a3 X 3 þ a4 X 4 þ a5 X 1 X 2 þ a6 X 1 X 3
þ a7 X 1 X 4 þ a8 X 2 X 3 þ a9 X 2 X 4 þ a10 X 3 X 4 þ a11 X 21 þ a12 X 22
þ a13 X 23
ð4Þ
129
X. Gong et al. / Separation and Purification Technology 132 (2014) 126–137
Table 3
Verification experiment conditions and the probability to attain CQA criteria.
No.
Concentrated extract amount (g)
WCCE (%)
CEA (g/g)
AEA (mL/g)
RT (°C)
Probability
Within design space
V1
V2
V3
V4
V5
V6
V7
V8
V9
200.0
20.0
20.0
20.0
20.0
20.0
20.0
20.0
20.0
54.0
55.0
55.0
55.0
52.0
52.0
52.0
49.0
49.0
0.92
0.94
0.94
0.94
0.91
0.91
0.91
0.91
0.91
1.9
1.9
1.9
1.9
1.7
1.7
1.7
1.8
1.8
5.0
5.0
15.0
25.0
5.0
15.0
25.0
5.0
25.0
0.980
0.968
0.968
0.968
0.970
0.970
0.970
0.000
0.000
Yes
Yes
Yes
Yes
Yes
Yes
Yes
No
No
Table 4
Risk assessment to identify main process influences.
Drug quality attributes
Extraction
First ethanol
precipitation
Second ethanol
precipitation
Activated carbon
adsorption
Water
precipitation
pH
adjustment
Sterilization
Color
pH value
Active ingredient
contents
Fingerprint similarity
Dry matter content
Abnormal toxicity
Residue on ignition
Insoluble particles
Bacterial content
++
++
+
+
++
+
+
++
+
+
+
+
++
+
+
++
++
+
+
+
+
+
++
++, +, and refer to high risk, moderate risk, and low risk, respectively.
where a0 is a constant, a1–a13 are regression coefficients, and Y is a
CQA. All the variables were coded before modeling. Stepwise
regression was applied to remove insignificant terms. The significance levels for adding terms and remove terms were both set to
0.05. Terms removed were considered to be insignificant. Analysis
of variance (ANOVA) was performed to obtain the p values of the
remaining terms.The design space was calculated using an exhaustive search-Monte Carlo method, a self-written program of Matlab
(Version 7.11, MathWorks, USA). In the exhaustive search, the calculation step sizes for WCCE, AEA, CEA, and RT were 0.02, 0.02,
0.02, and 0.2, respectively. For all the experimental results, the values of relative standard deviation (RSD) of the active ingredient
contents or dry matter content in supernatant were considered to
be the same with the RSD values of the center point in Monte-Carlo
simulation. Exhaustive search was repeated 10,000 times to calculate the probability. The acceptable probability for the design space
was set as 0.90.
Table 5
Criteria of CQAs.
CQAs
Lower control limit (%)
Upper control limit (%)
DMR
ACRDSS
ACRHSYA
ACRRA
ACRLA
ACRSaB
58.0
50.0
45.0
60.0
40.0
48.0
80.0
70.0
70.0
82.0
70.0
70.0
3. Results and discussion
relates to both the fingerprint similarity and dry matter content
of the injection. Therefore the removal of dry matter and the recoveries of active ingredients are identified to be the CQAs of the FEP
process. There are a total of six CQAs of DMR, DSS recovery
(ACRDSS), HSYA recovery (ACRHSYA), RA recovery (ACRRA), LA recovery (ACRLA), and SaB recovery (ACRSaB). According to the industrial
experiences and published results [36], the criteria of CQAs are
shown in Table 5.
3.1. CQA identification
3.2. CPP selection
Danhong injection is a botanical injection produced with multiunit operations. Therefore the CQAs of the FEP process should be
identified with a global perspective. The criteria for Danhong injection contain color, pH value, active ingredient contents, fingerprint
similarity, dry matter content, abnormal toxicity, residue on ignition, insoluble particles, bacterial content, heavy metals, and harmful elements. Heavy metals and harmful elements are generally
controlled in raw material quality testing. The risks of different
processes on other Danhong injection criteria were identified
based on experiences. The risk assessment results are listed in
Table 4. The FEP process affects the criteria of active ingredient
contents, fingerprint similarity, and dry matter content of Danhong
injection. Higher active ingredient recoveries help to attain the criteria of active ingredient contents. The removal of dry matter
Many parameters may affect the supernatant composition of
the ethanol precipitation process. An Ishikawa diagram analysis
was carried out to find out the potential CPPs, as seen in Fig. 1.
The influences from environment, material attributes, equipment,
ethanol addition, and refrigeration are included.
Zhang et al. also summarized possible mechanisms for the losses
of active components in the ethanol precipitation process, which are
precipitation, reaction, and adsorption [30]. In the ethanol precipitation process, the solubilities of the active ingredients and the impurities are mainly affected by refrigeration temperature and the
solvent composition in the supernatant. The solvent in the supernatant is mainly composed of water and ethanol. Water in the supernatant can be a reactant for the hydrolysis reactions. Refrigeration
temperature can affect the reaction rate. Therefore, RT, the water
130
X. Gong et al. / Separation and Purification Technology 132 (2014) 126–137
Fig. 1. Ishikawa diagram analysis for the FEP process of Danhong injection.
content and the ethanol content in the supernatant are important
parameters for an ethanol precipitation process. Because CEA,
AEA, and WCCE all affect the solvent composition of the supernatant
[42], they are probably important parameters.
Recently, Yan et al. selected significant process parameters in
the ethanol precipitation process of Danhong injection using a fractional factorial design [36]. WCCE, CEA, AEA, and RT are found to be
the most important four parameters that affects supernatant quality [36], which is in agreement with the mechanism of ethanol precipitation. Zhang et al. using the Failure Mode Effects Analysis
method identified three parameters, which are concentrated
extract density, AEA, and RT, as the CPPs of an ethanol precipitation
process [30]. Concentrated extract density is usually controlled in
the concentration process by evaporating the water in the extract.
Considering both the mechanism and the literature results, WCCE,
CEA, AEA, and RT are selected as the CPPs in this work. Because of
the high control expenses, normally RT is a noise parameter in production process. The effects caused by its fluctuations should be
mitigated by the other three CPPs.
3.3. Modeling and analysis
The values of active ingredient contents in the dry matter of the
concentrated extract were 3.20 mg g1, 4.55 mg g1, 2.93 mg g1,
4.65 mg g1, and 56.99 mg g1 for DSS, HSYA, RA, LA, and SaB,
respectively. However, because the sum of five active ingredient
contents were only 72.32 mg g1, dry matter was mainly composed of impurities. SaB content was much higher than any other
active ingredient content in the concentrated extract, which is consistent with the literature [27,37].
The experimental results of six CQAs are listed in Table 2. The
losses of all the active ingredients were observed, which may be
caused by chemical transformation or precipitation [26]. After
regression, Eqs. (5)–(10) were applied to calculate DMR (Y1),
ACRDSS (Y2), ACRHSYA (Y3), ACRRA (Y4), ACRLA (Y5), and ACRSaB (Y6),
respectively.
Y 3 ¼ 46:62 þ 13:77X 1 3:47X 2 2:88X 3 þ 2:33X 4
1:83X 1 X 3 þ 0:95X 1 X 4 0:63X 3 X 4 þ 0:73X 23
Y 4 ¼ 69:45 þ 6:90X 1 1:84X 2 1:33X 2 X 3 1:37X 2 X 4
1:72X 1 X 3 þ 0:85X 1 X 4 0:83X 3 X 4
1:62X 1 X 3 þ 0:66X 1 X 4 0:88X 3 X 4
ð10Þ
2
The determination coefficients (R ) and the p values for regression coefficients are listed in Table 6. For DMR, ACRDSS, ACRHSYA,
ACRLA, and ACRSaB, the determination coefficients were higher than
0.84, which means satisfactory correlation results were obtained.
ANOVA was applied to determine the impacts of WCCE, CEA,
AEA, and RT on all the CQAs, respectively. As seen in Table 6, the
linear terms of WCCE and CEA are significant process parameters
for all the CQAs. The linear terms of AEA and RT are insignificant
for ACRRA. For DMR, ACRDSS, ACRHSYA, ACRLA, and ACRSaB, the interaction terms of X1X3, X1X4 and X3X4 are significant. For ACRRA, the
Table 6
Determination coefficients and p values obtained with ANOVA method.
CQAs
DMR
ACRDSS
ACRHSYA
ACRRA
ACRLA
ACRSaB
r2
r 2adj
0.9698
0.9624
0.9843
0.9815
0.9947
0.9936
0.8462
0.8315
0.9850
0.9823
0.9837
0.9807
<0.0001a
<0.0001a
<0.0001a
<0.0001a
0.0440
<0.0001a
0.0113
<0.0001a
<0.0001a
0.0174
<0.0001a
<0.0001a
<0.0001a
<0.0001a
<0.0001a
<0.0001a
0.0005a
<0.0001a
<0.0001a
<0.0001a
<0.0001a
<0.0001a
<0.0001a
<0.0001a
<0.0001a
<0.0001a
0.0009a
<0.0001a
<0.0001a
<0.0001a
0.0068a
<0.0001a
0.0301
0.0082a
0.0049a
p value
X1
X2
X3
X4
X1X2
X1X3
X1X4
X2X3
X2X4
X3X4
X23
ð5Þ
ð6Þ
ð9Þ
Y 6 ¼ 48:06 þ 11:22X 1 3:42X 2 4:19X 3 þ 1:49X 4
Y 2 ¼ 50:73 þ 9:77X 1 2:45X 2 0:52X 3 þ 2:01X 4 1:68X 1 X 3
þ 0:86X 1 X 4 0:50X 3 X 4
ð8Þ
Y 5 ¼ 45:36 þ 11:86X 1 3:55X 2 4:28X 3 þ 1:74X 4
Y 1 ¼ 69:86 9:68X 1 þ 2:79X 2 þ 2:90X 3 2:34X 4 þ 0:76X 1 X 2
þ 2:59X 1 X 3 0:97X 1 X 4 þ 0:84X 3 X 4 1:70X 23
ð7Þ
a
0.0195
0.0164
0.0264
0.0018a
0.0349
p value less than 0.01.
0.0028a
0.00123a
131
X. Gong et al. / Separation and Purification Technology 132 (2014) 126–137
o
o
(a) AEA = 1.7 mL·g-1, RT = 15.0 C
(b) CEA = 0.92 g·g-1, RT = 15.0 C
(c) CEA = 0.92 g·g-1, AEA = 1.7 mL·g-1
(d) WCCE = 49.0%, CEA = 0.92 g·g-1
Fig. 2. Contour plots of parameter interactions on DMR.
o
(a) CEA = 0.92 g·g-1, RT = 15.0 C
(b) CEA = 0.92 g·g-1, AEA = 1.7 mL·g-1
(c) WCCE = 49.0%, CEA = 0.92 g·g-1
Fig. 3. Contour plots of parameter interactions on the recovery of DSS.
132
X. Gong et al. / Separation and Purification Technology 132 (2014) 126–137
o
(a) CEA = 0.92 g·g-1, RT = 15.0 C
(b) CEA = 0.92 g·g-1, AEA = 1.7 mL·g-1
(c) WCCE = 49.0%, CEA = 0.92 g·g-1
Fig. 4. Contour plots of parameter interactions on the recovery of HSYA.
o
(a) WCCE = 49.0 %, RT = 15.0 C
(b) WCCE = 49.0%, CEA = 0.92 g·g-1
Fig. 5. Contour plots of parameter interactions on the recovery of RA.
interaction terms of X2X3 and X2X4 are significant. There are interaction terms of adjustable parameters and RT significant for all the
CQAs. Quadratic terms of AEA are significant for DMR and ACRHSYA.
In Table 2, DMR values varied from 42.4% to 81.6%. The removed
dry matter was composed of impurities and active ingredients [27].
The response surfaces of DMR are illustrated in Fig. 2. DMR
increases as WCCE decreases or CEA increases. Saccharides were
found to be the main components of dry matter of the concentrated extract from Danshen [27,37]. There are several different
saccharides existing in Danshen, such as glucose, fructose, and
sucrose [38]. These saccharides are easily extracted when water
is used as the extractant. Because saccharides will partly degrade
and form 5-(hydroxymethyl)-2-furancarboxaldehyde [39], which
is recognized as a potential cytotoxic, genotoxic, and tumorigenic
agent [40], the removal of saccharides helps to improve drug
safety. Saccharide solubility increases as equilibrium temperature
increases or water content in mixed water–ethanol solution
increases [20,22,25,41]. Therefore, in Fig. 2(d), the increase of RT
results in lower DMR. The increase of CEA, increase of AEA, and
decrease of WCCE led to higher ethanol content in the supernatant.
Accordingly, higher DMR can be obtained. These observations are
in agreement with the literature [42].
In Table 2, the loss of DSS is no less than 30%. The response surfaces of DSS recovery are illustrated in Fig. 3. DSS recovery
increases as WCCE increases or CEA decreases. With the decreases
of AEA, the recovery of DSS slightly increases. As seen in Fig. 3(d),
X. Gong et al. / Separation and Purification Technology 132 (2014) 126–137
o
(a) CEA = 0.92 g·g-1, RT = 15.0 C
(b) CEA = 0.92 g·g-1, AEA = 1.7 mL·g-1
(c) WCCE = 49.0%, CEA = 0.92 g·g-1
Fig. 6. Contour plots of parameter interactions on the recovery of LA.
o
(a) CEA = 0.92 g·g-1, RT = 15.0 C
(b) CEA = 0.92 g·g-1, AEA = 1.7 mL·g-1
(c) WCCE = 49.0%, CEA = 0.92 g·g-1
Fig. 7. Contour plots of parameter interactions on the recovery of SaB.
133
134
X. Gong et al. / Separation and Purification Technology 132 (2014) 126–137
the increase of RT results in a higher DSS recovery. DSS is one of the
hydrolyzates of SaB [43,44]. In previous work, DSS loss was found
to be caused by both precipitation and reactions in the ethanol precipitation process [26].
In Table 2, HSYA recovery varied from 27.5% to 74.7%. HSYA is a
hydrolyzate of anhydrosafflor yellow B [45]. HSYA can also hydrolyze and form p-coumaric acid [45]. As seen in Fig. 4, the recovery
of HSYA increases as WCCE increases or CEA decreases. With the
increases of AEA, the recovery of HSYA decreases. In Fig. 4(d), the
increase of RT leads to a higher HSYA recovery.
The recovery of RA was in the range of 49.9–83.6%. The loss of
RA is smaller than that of any other active ingredient. As seen in
Fig. 5, the recovery of RA increases as WCCE increases, or CEA
decreases.
LA is also one of the hydrolyzates of SaB [43,44]. In Table 2, LA
recovery varies from 22.3% to 70.5%. Fig. 6 shows the response surfaces of LA recovery. The recovery of LA increases as WCCE
increases, CEA decreases, or RT increases. With the decrease of
AEA, the recovery of LA increases.
In Table 2, SaB recovery varied from 25.1% to 71.2%. Fig. 7 is the
response surface of SaB recovery. In Fig. 7, the recovery of SaB
decreases as WCCE decreases, RT decreases, or CEA increases. With
the decrease of AEA, the recovery of SaB also increases. In previous
work, SaB loss was caused by both precipitation and reactions in
the ethanol precipitation process [26].
It can be found that higher WCCE and lower CEA both led to
higher recoveries of DSS, LA, SaB, and RA. A possible explanation
is that phenolic acids were extracted in its salt form, and salt solubilities increased as the water content in the supernatant
increased [26].
(a) CEA = 0.91 g·g -1
3.4. Design space development and verification
In the calculation of the design space, the range of RT was
allowed to change from 5 to 25 °C to simulate its fluctuations.
The design space is illustrated in Fig. 8. The design space is an
irregular polyhedron. In order to estimate the accuracy, control
ability of noise parameter effects, and the applicability of the
design space, several verification experiments were carried out.
The conditions are chosen based on probability to attain CQA criteria, as seen in Table 3. The results are listed in Fig. 9. Average relative deviation (ARD) values were calculated using Eq. (11).
ARD ¼
jEV PVj
100%
EV
(b) WCCE = 52.0%
ð11Þ
where PV and EV represent the predicted value and the experimental value, respectively.
In Experiment V1, the amount of concentrated extract used was
200 g, which is 10-fold of that used in the simplified central composite experiments. The ARD values of all the CQAs are less than
8%, which means that predicted results agreed well with experimental results. In Fig. 9, all the CQAs of Experiment V1 attain the
criteria in Table 5, which indicates that design space is applicable
in a larger scale ethanol precipitation process. The ARD values
are less than 10% for Experiments V2–V7. Because all the CQAs of
Experiments V2–V7 attain the criteria in Table 5, it can be concluded that the negative effects caused by RT fluctuation can be
controlled by operating the other three adjustable CPPs in the
design space.
The conditions of Experiments V8–V9 are not in the design
space. The predicted results show good agreement with experimental results because the ARD values are less than 5%. However,
ACRDSS, ACRHSYA, and ACRSaB in Experiments V8 are lower than criteria. These results indicate that a noise parameter cannot be controlled when operating adjustable parameters outside the design
space.
(c) AEA = 1.9mL·g-1
Fig. 8. The design space for Danhong ethanol precipitation (Regions in the white
line is the design space; Color bar refers to the probability to attain CQA criteria; h,
V1; s, V2, V3, V4; 4, V5, V6, V7; , V8, V9). (For interpretation of the references to
colour in this figure legend, the reader is referred to the web version of this article.)
Because the design space is an irregular polyhedron, it will be
difficult to manoeuvre practically for an operator. Therefore a multidimensional rectangle, known as the normal operating ranges
(NOR), with probability more than 0.80 to attain CQA criteria is
135
85
75
80
70
75
65
ACR DSS (%)
DMR (%)
X. Gong et al. / Separation and Purification Technology 132 (2014) 126–137
70
65
60
V1
V2
V3
V4
V5
V6
V7
V8
45
V9
80
ACR RA (%)
65
60
55
V5
V6
V7
45
60
V5
V6
V7
V8
55
V9
V1
V2
V3
V4
V5
V6
V7
Experiment No.
Experiment No.
(c) HSYA recovery
(d) RA recovery
75
75
70
70
65
65
60
55
50
V9
V8
V9
V8
V9
70
65
V4
V8
75
50
60
55
50
45
45
40
35
V4
(b) DSS recovery
85
V3
V3
(a) DMR
70
V2
V2
Experiment No.
90
V1
V1
Experiment No.
75
40
ACRLA(%)
55
50
ACRSaB (%)
ACRHSYA(%)
55
60
V1
V2
V3
V4
V5
V6
V7
V8
V9
40
V1
V2
V3
V4
V5
V6
V7
Experiment No.
Experiment No.
(e) LA recovery
(f) SaB recovery
Fig. 9. The results of verification experiments (j, predicted results; s, experimental results; solid line, CQA criteria).
required [46]. This is a part of defining the control strategy. After
calculation, ranges of WCCE of 54.1–55.0%, CEA of 0.932–
0.937 g g1, and AEA of 1.78–1.90 mL g1 are recommended with
a probability more than 0.908 to attain CQA criteria when refrigeration temperature varies from 5 °C to 25 °C.
4. Conclusion
In this work, a design space based on probability was obtained
using the adjustable parameters to control the negative effects
caused by the fluctuations of a noise parameter. The FEP process
dealing with the water extract of the mixed Danshen and Honghua
was investigated as a sample. Six CQAs of DMR, ACRDSS, ACRRA,
ACRLA, ACRHSYA, and ACRSaB were identified using the risk assessment. The CPPs were identified using an Ishikawa diagram. WCCE,
CEA, and AEA are adjustable parameters. RT was investigated on
three levels for simulating its fluctuations. Experiments according
to a simplified central composite design were carried out and the
quantitative relationships between the CPPs and the CQAs were
obtained. The adjusted determination coefficients for all the models are higher than 0.83. The increase of CEA and the decrease of
WCCE will both cause more precipitation. The increase of WCCE
and RT will increase the recoveries of active ingredients. The
increase of CEA and AEA will lead to lower active ingredient recoveries. The design space was calculated and verified. NOR of WCCE
of 54.1–55.0%, CEA of 0.932–0.937 g g1, and AEA of 1.78–
1.90 mL g1 are recommended with probability more than 0.908
136
X. Gong et al. / Separation and Purification Technology 132 (2014) 126–137
to attain the CQA criteria. In the verification experiments,
predicted results agreed well with experimental results, which
means that the models are accuracy and have the potential application in a larger scale ethanol precipitation process. Experimental
results with adjustable parameters operated outside the design
space cannot attain all the criteria of CQAs. The proposed method
can be used to develop a design space with the negative effects
caused by the fluctuations of the noise parameters controlled to
increase process robustness.
[17]
[18]
[19]
[20]
Acknowledgements
[21]
The authors like to acknowledge the supports of China
International Science and Technology Cooperation Project (No.
2010DFB33630) and the Project Supported by Zhejiang Provincial
Natural Science Foundation of China Foundation (LQ12H29004)
on this work.
References
[22]
[23]
[24]
[25]
[1] L.X. Yu, Pharmaceutical quality by design: product and process development,
understanding, and control, Pharm. Res. Dordr. 25 (2008) 781–791.
[2] E. Korakianiti, D. Rekkas, Statistical thinking and knowledge management for
quality-driven design and manufacturing in pharmaceuticals, Pharm. Res.
Dordr. 28 (2011) 1465–1479.
[3] L. Zhang, B. Yan, X. Gong, L.X. Yu, H. Qu, Application of quality by design to the
process development of botanical drug products: a case study, AAPS Pharm.
Sci. Technol. 14 (2013) 277–286.
[4] J.M. Merritt, S.K. Viswanath, G.A. Stephenson, Implementing quality by design
in pharmaceutical salt selection: a modeling approach to understanding
disproportionation, Pharm. Res. Dordr. 30 (2013) 203–217.
[5] International
Conferenceon
Harmonization
(ICH).
Pharmaceutical
development.
Q8(R2),
<http://www.ich.org/fileadmin/Public_Web_Site/
ICH_Products/Guidelines/Quality/Q8_R1/Step4/Q8_R2_Guideline.pdf> (2009.
04).
[6] P.T. Ingvarsson, M. Yang, H. Mulvad, H.M. Nielsen, J. Rantanen, C. Foged,
Engineering of an inhalable DDA/TDB liposomal adjuvant: a quality-by-design
approach towards optimization of the spray drying process, Pharm. Res.
(2013).
[7] E. Rozet, P. Lebrun, B. Debrus, B. Boulanger, P. Hubert, Design spaces for
analytical methods, TrAC Trend Anal. Chem. 42 (2013) 157–167.
[8] X. Li, L. Zhao, M. Li, Y. Feng, D. Xu, K. Ruan, Applications and characteristics of
mathematical models to study compression process of pharmaceutical
powder, Chinese J. New Drug 21 (2012) 1362–1366.
[9] C.P. Jiang, L. Flansburg, S. Ghose, P. Jorjorian, A.A. Shukla, Defining process
design space for a hydrophobic interaction chromatography (HIC) purification
step: application of quality by design (QbD) principles, Biotechnol. Bioeng. 107
(2010) 985–997.
[10] M. Looby, N. Ibarra, J.J. Pierce, K. Buckley, E. O’Donovan, M. Heenan, E. Moran,
S.S. Farid, F. Baganz, Application of quality by design principles to the
development and technology transfer of a major process improvement for the
manufacture of a recombinant protein, Biotechnol. Progr. 27 (2011) 1718–
1729.
[11] Z. Cimarosti, F. Bravo, P. Stonestreet, F. Tinazzi, O. Vecchi, G. Camurri,
Application of quality by design principles to support development of a
control strategy for the control of genotoxic impurities in the manufacturing
process of a drug substance, Org. Process Res. Dev. 14 (2010) 993–998.
[12] S. Thirunahari, P.S. Chow, R.B.H. Tan, Quality by design (QbD)-based
crystallization process development for the polymorphic drug tolbutamide,
Cryst. Growth Des. 11 (2011) 3027–3038.
[13] L.N. Mockus, T.W. Paul, N.A. Pease, N.J. Harper, P.K. Basu, E.A. Oslos, G.A. Sacha,
W.Y. Kuu, L.M. Hardwick, J.J. Karty, M.J. Pikal, E. Hee, M.A. Khan, S.L. Nail,
Quality by design in formulation and process development for a freeze-dried,
small molecule parenteral product: a case study, Pharm. Dev. Technol. 16
(2011) 549–576.
[14] J.K. Mbinze, P. Lebrun, B. Debrus, A. Dispas, N. Kalenda, J.M.T. Mbay, T.
Schofield, B. Boulanger, E. Rozet, P. Hubert, R.D. Marini, Application of an
innovative design space optimization strategy to the development of liquid
chromatographic methods to combat potentially counterfeit nonsteroidal
anti-inflammatory drugs, J. Chromatogr. A 1263 (2012) 113–124.
[15] B. Debrus, P. Lebrun, A. Ceccato, G. Caliaro, E. Rozet, I. Nistor, R. Oprean, F.J.
Ruperez, C. Barbas, B. Boulanger, P. Hubert, Application of new methodologies
based on design of experiments, independent component analysis and design
space for robust optimization in liquid chromatography, Anal. Chim. Acta 691
(2011) 33–42.
[16] B. Debrus, P. Lebrun, J.M. Kindenge, F. Lecomte, A. Ceccato, G. Caliaro, J.M.T.
Mbay, B. Boulanger, R.D. Marini, E. Rozet, P. Hubert, Innovative highperformance liquid chromatography method development for the screening
[26]
[27]
[28]
[29]
[30]
[31]
[32]
[33]
[34]
[35]
[36]
[37]
[38]
[39]
[40]
[41]
[42]
[43]
of 19 antimalarial drugs based on a generic approach, using design of
experiments, independent component analysis and design space, J.
Chromatogr. A 1218 (2011) 5205–5215.
D.C. Montgomery, Design and Analysis of Experiments, John Wiley & Sons Inc.,
New York, 2005.
Y. He, H.T. Wan, Y.G. Du, X.D. Bie, T. Zhao, W. Fu, P.K. Xing, Protective effect of
Danhong injection on cerebral ischemia-reperfusion injury in rats, J.
Ethnopharmacol. 144 (2012) 387–394.
H.T. Liu, Y.F. Wang, O. Olaleye, Y. Zhu, X.M. Gao, L.Y. Kang, T. Zhao,
Characterization of in vivo antioxidant constituents and dual-standard
quality assessment of Danhong injection, Biomed. Chromatogr. 27 (2013)
655–663.
X.C. Gong, C. Wang, L. Zhang, H.B. Qu, Solubility of xylose, mannose, maltose
monohydrate, and trehalose dihydrate in ethanol–water solutions, J. Chem.
Eng. Data 57 (2012) 3264–3269.
X.C. Gong, S.S. Wang, H.B. Qu, Solid-liquid equilibria of D-glucose, D-fructose
and sucrose in the mixture of ethanol and water from 273.2 K to 293.2 K,
Chinese J. Chem. Eng. 19 (2011) 217–222.
L. Zhang, X.C. Gong, Y.F. Wang, H.B. Qu, Solubilities of protocatechuic aldehyde,
caffeic acid, D-galactose, and D-raffinose pentahydrate in ethanol-water
solutions, J. Chem. Eng. Data 57 (2012) 2018–2022.
G.Y. Koh, G. Chou, Z. Liu, Purification of a water extract of Chinese sweet tea
plant (Rubus suavissimus S. Lee) by alcohol precipitation, J. Agric. Food Chem.
57 (2009) 5000–5006.
M. Boulet, M. Britten, F. Lamarche, Dispersion of food proteins in water–
alcohol mixed dispersants, Food Chem. 74 (2001) 69–74.
A. Bouchard, G.W. Hofland, G.J. Witkamp, Properties of sugar, polyol, and
polysaccharide water–ethanol solutions, J. Chem. Eng. Data 52 (2007) 1838–
1842.
X.C. Gong, S.S. Wang, Y. Li, H.B. Qu, Separation characteristics of ethanol
precipitation for the purification of the water extract of medicinal plants, Sep.
Purif. Technol. 107 (2013) 273–280.
X.C. Gong, S.S. Wang, H.B. Qu, Comparison of two separation technologies
applied in the manufacture of botanical injections: second ethanol
precipitation and solvent extraction, Ind. Eng. Chem. Res. 50 (2011) 7542–
7548.
P.A.G. Soares, A.F.M. Vaz, M.T.S. Correia, A. Pessoa, M.G. Carneiro-Da-Cunha,
Purification of bromelain from pineapple wastes by ethanol precipitation, Sep.
Purif. Technol. 98 (2012) 389–395.
S. Golunski, V. Astolfi, N. Carniel, D. de Oliveira, M. Di Luccio, M.A. Mazutti, H.
Treichel, Ethanol precipitation and ultrafiltration of inulinases from
Kluyveromyces marxianus, Sep. Purif. Technol. 78 (2011) 261–265.
L. Zhang, B.J. Yan, X.C. Gong, L.X. Yu, H.B. Qu, Application of quality by design to
the process development of botanical drug products: a case study, AAPS
PharmSciTech 14 (2013) 277–286.
L. Zhang, X.C. Gong, H.B. Qu, Optimizing the alcohol precipitation of danshen
by response surface methodology, Sep. Sci. Technol. 48 (2013) 977–983.
D.Y. Zhang, Y.G. Zu, Y.J. Fu, M. Luo, C.B. Gu, W. Wang, X.H. Yao, Negative
pressure cavitation extraction and antioxidant activity of biochanin A and
genistein from the leaves of Dalbergia odorifera T. Chen, Sep. Purif. Technol. 83
(2011) 91–99.
R. Tabaraki, E. Heidarizadi, A. Benvidi, Optimization of ultrasonic-assisted
extraction of pomegranate (Punica granatum L.) peel antioxidants by response
surface methodology, Sep. Purif. Technol. 98 (2012) 16–23.
Y. Li, Z. Guo, X. Gong, H. Qu, Simultaneous determination of danshensu,
hydroxysafflor yellow A, rosmarinic acid, lithospermic acid, salvianolic acid B
in water extract of mixed Salviae Miltiorrhizae Radix et Rhizoma and Carthami
Flos by HPLC, China J. Chinese Mater. Med. 38 (2013) 1653–1656.
Chinese Pharmacopoeia Commission, Chinese Pharmacopoeia, Chinese
Medical Science and Technology Press, Beijing, 2010.
B. Yan, Z. Guo, H. Qu, B. Zhao, T. Zhao, An approach to determine critical
process parameters for ethanol precipitation process of Danhong injection,
China J. Chinese Mater. Med. 38 (2013) 1672–1675.
X.C. Gong, A.Y. Yan, H.B. Qu, Optimization for the ethanol precipitation process
of botanical injection: indicator selection and factor influences, Sep. Sci.
Technol. 49 (2014) 619–626.
H. Li, F. Song, Z. Zheng, Z. Liu, S. Liu, Characterization of saccharides and
phenolic acids in the Chinese herb Tanshen by ESI-FT-ICR-MS and HPLC, J.
Mass Spectrom.: JMS 43 (2008) 1545–1552.
P.M. Falcone, D. Tagliazucchi, E. Verzelloni, P. Giudici, Sugar conversion
induced by the application of heat to grape must, J. Agric. Food Chem. 58
(2010) 8680–8691.
I. Severin, C. Dumont, A. Jondeau-Cabaton, V. Graillot, M.C. Chagnon, Genotoxic
activities of the food contaminant 5-hydroxymethylfurfural using different
in vitro bioassays, Toxicol. Lett. 192 (2010) 189–194.
E.A. Macedo, A.M. Peres, Thermodynamics of ternary mixtures containing
sugars. SLE of D-fructose in pure and mixed solvents. Comparison between
modified UNIQUAC and modified UNIFAC, Ind. Eng. Chem. Res. 40 (2001)
4633–4640.
X. Gong, B. Yan, H. Qu, Correlations of three important technological
parameters in first ethanol precipitation of Danshen, Zhongguo Zhong Yao
Za Zhi 35 (2010) 3274–3277.
Y.X. Guo, D.J. Zhang, H. Wang, Z.L. Xiu, L.X. Wang, H.B. Xiao, Hydrolytic kinetics
of lithospermic acid B extracted from roots of Salvia miltiorrhiza, J.
Pharmaceut. Biomed. 43 (2007) 435–439.
X. Gong et al. / Separation and Purification Technology 132 (2014) 126–137
[44] J.Y. Pan, X.C. Gong, H.B. Qu, Quantitative H-1 NMR method for hydrolytic
kinetic investigation of salvianolic acid B, J. Pharmaceut. Biomed. 85 (2013)
28–32.
[45] L. Fan, R. Pu, H.Y. Zhao, X. Liu, C. Ma, B.R. Wang, D.A. Guo, Stability and
degradation of hydroxysafflor yellow A and anhydrosafflor yellow B in the
137
Safflower injection studied by HPLC-DAD-ESI-MSn, J. Chinese Pharm. Sci. 20
(2011) 47–56.
[46] C. Castagnoli, M. Yahyah, Z. Cimarosti, J.J. Peterson, Application of quality by
design principles for the definition of a robust crystallization process for
casopitant mesylate, Org. Process Res. Dev. 14 (2010) 1415–1427.