Interesting morphological behavior of organic salt choline fenofibrate: Effect of supersaturation and polymeric impurity

Article abstract of DOI:10.1021/cg500283y

Crystal habit of drug molecules can have significant influence on the processing and performance of pharmaceutical products. During the development of Trilipix, a pharmaceutical product used for the treatment of mixed dyslipidemia, several crystal habits were observed for the active ingredient choline fenofibrate. The dissolution and performance of the drug product were not impacted by changes in crystal habit of the active ingredient due to high solubility of the drug. However, the formulation process was impacted by variations in crystal habit of the active ingredient, requiring robust control of the crystal habit. The crystal habit was greatly influenced by supersaturation during crystallization from a mixed solvent system comprising methanol and isopropanol. In addition to supersaturation, trace levels of a polymeric impurity in the starting material fenofibrate had a detrimental effect on the crystal habit. This article discusses the effects of these factors on the crystal habit of choline fenofibrate and the design of a crystallization process to deliver the target crystal habit, most suited to the formulation process. The article also provides preliminary mechanistic insights into the crystal habit of this organic salt using an extension of the spiral growth model for morphology prediction of organic molecular crystals. An attempt is made to explain the effect of supersaturation and impurity on the crystal habit of choline fenofibrate using the concepts of stability of surfaces, building units, periodic bond chain theory, and the spiral growth model.

Full text of DOI:10.1021/cg500283y

Subscriber access provided by University of Texas Libraries
Article
Interesting Morphological Behavior of Organic Salt Choline
Fenofibrate: Effect of Supersaturation and Polymeric Impurity
Shailendra Bordawekar, Zubin Kuvadia, Preshit Dandekar, Samrat Mukherjee, and Michael F. Doherty
Cryst. Growth Des., Just Accepted Manuscript • Publication Date (Web): 19 Jun 2014
Downloaded from http://pubs.acs.org on June 20, 2014
Just Accepted
“Just Accepted” manuscripts have been peer-reviewed and accepted for publication. They are posted
online prior to technical editing, formatting for publication and author proofing. The American Chemical
Society provides “Just Accepted” as a free service to the research community to expedite the
dissemination of scientific material as soon as possible after acceptance. “Just Accepted” manuscripts
appear in full in PDF format accompanied by an HTML abstract. “Just Accepted” manuscripts have been
fully peer reviewed, but should not be considered the official version of record. They are accessible to all
readers and citable by the Digital Object Identifier (DOI®). “Just Accepted” is an optional service offered
to authors. Therefore, the “Just Accepted” Web site may not include all articles that will be published
in the journal. After a manuscript is technically edited and formatted, it will be removed from the “Just
Accepted” Web site and published as an ASAP article. Note that technical editing may introduce minor
changes to the manuscript text and/or graphics which could affect content, and all legal disclaimers
and ethical guidelines that apply to the journal pertain. ACS cannot be held responsible for errors
or consequences arising from the use of information contained in these “Just Accepted” manuscripts.
Crystal Growth & Design is published by the American Chemical Society. 1155
Sixteenth Street N.W., Washington, DC 20036
Published by American Chemical Society. Copyright © American Chemical Society.
However, no copyright claim is made to original U.S. Government works, or works
produced by employees of any Commonwealth realm Crown government in the course
of their duties.

Page 1 of 39
Crystal Growth & Design
1
2
3
4
5
6
Interesting Morphological Behavior of Organic Salt Choline Fenofibrate: Effect of
Supersaturation and Polymeric Impurity
7
8
9
Shailendra Bordawekar1*, Zubin Kuvadia2, Preshit Dandekar2, Samrat Mukherjee1 and Michael
Doherty2
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
39
40
41
42
43
44
45
46
47
48
49
50
51
52
53
54
55
56
57
58
59
60
1Process Research & Development, AbbVie, North Chicago, Illinois 60064, United States
2Department of Chemical Engineering, University of California Santa Barbara, Santa Barbara,
California 93106, United States
Abstract
Crystal habit of drug molecules can have significant influence on the processing and
performance of pharmaceutical products. During the development of Trilipix®, a pharmaceutical
product used for the treatment of mixed dyslipidemia, several crystal habits were observed for
the active ingredient Choline Fenofibrate. The dissolution and performance of the drug product
were not impacted by changes in crystal habit of the active ingredient due to high solubility of
the drug. However, the formulation process was impacted by variations in crystal habit of the
active ingredient, requiring robust control of the crystal habit. The crystal habit was greatly
influenced by supersaturation during crystallization from a mixed solvent system comprising of
methanol and isopropanol. In addition to supersaturation, trace levels of a polymeric impurity in
the starting material Fenofibrate had a detrimental effect on the crystal habit. This article
discusses the effects of these factors on the crystal habit of Choline Fenofibrate and the design of
a crystallization process to deliver the target crystal habit, most suited to the formulation process.
The article also provides preliminary mechanistic insights into the crystal habit of this organic
salt using an extension of the spiral growth model for morphology prediction of organic
molecular crystals. An attempt is made to explain the effect of supersaturation and impurity on
crystal habit of Choline Fenofibrate using the concepts of stability of surfaces, building units,
periodic bond chain theory and the spiral growth model.
(*Corresponding Author : Shailendra Bordawekar, email: shailendra.bordawekar@abbvie.com,
Tel: 847 935 0212)
1
ACS Paragon Plus Environment

Crystal Growth & Design
Page 2 of 39
1
2
3
4
1. Introduction
5
6
7
8
9
Crystallization is one of the most important unit operations in the manufacture of active
pharmaceutical ingredients (APIs). The majority of solid dosage forms of pharmaceutical
products contain crystalline APIs [1]. The physical and chemical properties of APIs dictate the
processing and performance of drug products. While chemical purity of an API primarily
impacts safety of the drug product, physical properties have the potential to impact downstream
processing, as well as bioavailability of the drug product. API manufacturing processes rely
heavily on the crystallization operation to achieve desired chemical purity and physical attributes
of the drug substance. One of the key physical attributes of an API is its crystal habit. API
crystal habit can play a pivotal role in defining many properties of the bulk drug such as surface
area, surface charge, powder flow, bulk density, wettability, etc. One or more of these properties
has the potential to influence the behavior of the drug product formulation process. The ability to
tailor crystal habit continues to be a problem of great interest in the API process development
community. Most pharmaceutical industry practitioners depend heavily on empirical approaches
to obtain desirable crystal habits. These approaches typically involve variations in crystallization
solvent and supersaturation, ripening via heat-cool cycles, application of sonication, etc. [2]
Recent advances in mechanistic models have provided fundamental insights into crystal shape
prediction. It is always beneficial to have sound theoretical understanding to support
experiments. Theoretical understanding of the crystal growth process also allows us to configure
and design possibilities of morphology control and engineering; thus meeting the current need of
introducing the new material or product into the market in the minimum time. Various
modifications to the classical Burton, Cabrera and Frank model [3] including the spiral growth
model for non-centrosymmetric molecules by Kuvadia and Doherty [4] allow for the prediction
of crystal morphology of realistic API molecules and also lay the foundation to design schemes
to control and tailor the crystal morphology in accordance with the application requirements.
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
39
40
41
42
43
44
45
46
47
48
49
50
51
52
53
54
55
56
57
58
59
60
In this article, we discuss the crystallization process development for Choline Fenofibrate, with a
particular emphasis on the crystal habit of this API. Choline Fenofibrate refers to the Choline salt
of Fenofibric acid. Choline Fenofibrate is the API in Trilipix®, a prescription medication used
along with diet to lower triglycerides and raise HDL (good) cholesterol in people at high risk of
heart disease who have abnormal cholesterol levels and are taking a statin medicine to manage
2
ACS Paragon Plus Environment

Page 3 of 39
Crystal Growth & Design
1
2
3
4
5
6
7
8
9
LDL (bad) cholesterol. Trilipix® is available as delayed release capsules for oral administration.
Each delayed release capsule contains enteric coated mini-tablets which are manufactured via a
wet granulation process. During the early phases of development of this product, multiple crystal
habits of the API were observed. The API crystal habit played a critical role in the performance
of the wet granulation process. This will be briefly discussed in the paper, followed by a
description of the crystallization process. We will then discuss the experimental findings relating
API crystal habit to crystallization process parameters in two different solvent systems. We will
also describe the impact of trace levels of a polymeric impurity, encountered in the starting
material on the crystal habit of the API. Finally, we will attempt to build on the existing
knowledge in the crystal growth community about crystal morphology prediction and provide a
fundamental approach for prediction of crystal habit of this organic salt, together with
mechanistic insights into the experimental observations. It is worthwhile to note that organic
salts are a very challenging class of crystals as far as morphology prediction is concerned and in
this article we will make a first pass attempt to gain some understanding in that direction.
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
39
40
41
42
43
44
45
46
47
48
49
50
51
52
53
54
55
56
57
58
59
60
2. Choline Fenofibrate crystal habits and impact on wet granulation process
The chemical structure of Choline Fenofibrate is shown in Figure 1. Choline Fenofibrate is
manufactured from Fenofibrate, which is an API used in Tricor®. The process used to
manufacture Choline Fenofibrate during early stages of development of the product is also
shown in Figure 1.
3
ACS Paragon Plus Environment

Crystal Growth & Design
Page 4 of 39
1
2
3
O
4
5
6
7
8
N
O
Cl
O
OH
O
9
Choline Fenofibrate
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
39
40
41
42
43
44
45
46
47
48
49
50
51
52
53
54
55
56
57
58
59
60
O
O
O
o ne
Ch li
. a .
IPA fl
a
N OH
,
,
1
q
rox
e
,
re ux
h d id
y
N
O
OH
O
Cl
O
Cl
O
Cl
O
OH
e
.
.
IPA M OH
2 A HCl
,
q
O
O
O
t
eno ra e
fib
F
t
r
o ne eno ra e
Ch li
fib
e o
n
c c
F
fib i A id
F
Figure 1. Chemical structure and manufacturing process for Choline Fenofibrate
The first step of the process is the saponification of Fenofibrate using aqueous sodium
hydroxide. The reaction is carried out in isopropanol (IPA) at reflux (82 ºC) for approximately 2
hours to yield >99.5 % conversion of Fenofibrate. Fenofibric acid is isolated directly from the
reaction mixture by acidification with hydrochloric acid. The Fenofibric acid is treated with a
methanol solution of Choline hydroxide to form the Choline salt of Fenofibric acid (Choline
Fenofibrate). The Choline Fenofibrate solution in methanol is distilled under vacuum. During
distillation, IPA is periodically charged to maintain a volume of approximately 12 L/kg of initial
Fenofibrate charge. The distillation replaces methanol (MeOH) with IPA, thereby reducing the
solubility of Choline Fenofibrate and induces crystallization of the API. The solid product is
isolated by filtration. The wet cake is washed with IPA to displace crystallization liquors and
dried under heat and vacuum. The early development lots of Choline Fenofibrate manufactured
by this process showed differences in crystal habit as depicted in Figure 2. It is worth noting that
all of the observed habits belong to the same crystal structure. This was confirmed via
independent measurements of their powder X-ray diffraction patterns and melting points.
4
ACS Paragon Plus Environment

Page 5 of 39
Crystal Growth & Design
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
39
40
41
42
43
44
45
46
47
48
49
50
51
52
53
54
55
56
57
58
59
60
Lot - 2
Lot - 1
Lot - 3
Figure 2. Crystal habits of Choline Fenofibrate obtained during early development
The optical microscope images in Figure 2 suggest that Lot-1 crystals exhibit a plate-like
morphology with very small thickness. We will refer to this habit as “Thin Plates”. The crystals
in Lot-2 appear thicker than those in Lot-1. We will refer to the crystal habit of Lot-2 as “Thick
Plates”. The crystals in Lot-3 also exhibit significant thickness like those in Lot-2. However,
they appear more elongated compared to Lot-2 crystals. We will refer to the Lot-3 habit as
“Prisms”. Our initial hypothesis was that the variations in the crystal habit may be caused by
evaporative crystallization. During the evaporation process, crystallization could occur at varying
degrees of supersaturation across the different API lots. We will further discuss the validation of
this hypothesis in the next section.
It was important to understand the impact of these crystal habits on the wet granulation process.
Thus, we evaluated nine different lots of API in the wet granulation. These nine lots covered the
5
ACS Paragon Plus Environment

Crystal Growth & Design
Page 6 of 39
1
2
3
4
5
6
7
8
9
three crystal habits described above, each in three different particle size ranges. The different
particle sizes for each habit were obtained by milling large particle size API lots under different
milling conditions. The resultant particle sizes ranged from D(v, 0.9) of about 200 microns at the
large end, D(v, 0.9) of about 70 microns in the mid range, to D(v, 0.9) of about 15 microns at the
small end, as measured by laser light scattering using a Malvern Mastersizer instrument. D(v,
0.9) represents the particle diameter at which 90% of the volume distribution has a lower particle
size than this value. The studies conducted using these API lots illustrated clear differences in
their performance in the wet granulation process. The API constitutes greater than 50% of the
granulation blend and thereby dominates performance of the process. The large particle size
prisms and thick plates exhibited very good API flow. However, they produced weak and fluffy
granules that did not produce mini-tablets of acceptable hardness. The very small particle size
lots for all crystal habits produced granules with acceptable hardness. However, these lots
suffered from very poor flow. The medium sized lots of prisms and thick plates showed good
flow and they produced granules with acceptable hardness resulting in mini-tablets with optimal
properties. However, the medium sized thin plates showed suboptimal flow. This lot of thin
plates also required a larger amount of water and longer mixing time in the high shear granulator
to reach the granulation end point. Despite this, the granules resulting from this lot were weak
and fluffy, resulting in poor quality mini-tablets. Based on these results, the thin plate habit was
not desirable for the formulation process. Prisms and thick plates in the medium particle size
range were identified as the API property targets for the formulation process. Experience had
shown that fairly large crystals of each habit could be obtained from the API crystallization and
the desired particle size range could be easily accomplished through a milling operation. Thus,
the API development objective was to design a robust crystallization process to consistently
deliver large particle size prisms or thick plates which could then be milled to the desired particle
size.
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
39
40
41
42
43
44
45
46
47
48
49
50
51
52
53
54
55
56
57
58
59
60
3. Crystallization of prisms and thick plates
Since the earlier API process involved evaporative crystallization from a mixed solvent system
comprising of MeOH and IPA, further efforts to control the crystal habit were initiated using the
same solvent system. To test the hypothesis about the role of supersaturation on crystal habit,
6
ACS Paragon Plus Environment

Page 7 of 39
Crystal Growth & Design
1
2
3
4
5
6
7
8
9
controlled crystallization studies were conducted. A prerequisite to controlled crystallization
experiments is determining the solubility of the API under conditions relevant to the process.
Solubility studies were conducted at varying ratios of methanol and IPA at different
temperatures. The solubility data are presented in Figure 3.
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
39
40
41
42
43
44
45
46
47
48
49
50
51
52
53
54
55
56
57
58
59
60
50
45
60 C
40
25 C
35
50 C
30
25
20
15
10
5
0
0
10
20
30
40
50
60
70
80
90
100
Volume% MeOH in MeOH-IPAmixture
Figure 3. Solubility of Choline Fenofibrate in MeOH-IPA as a function of temperature
It is evident from Figure 3 that MeOH offers good solubility for the API and IPA is well suited
as an antisolvent. Additional solubility values at intermediate temperatures were calculated
using the van’t Hoff equation [5]. Solubility at intermediate solvent compositions not
experimentally measured, was estimated via interpolation of the above data. Using these data, a
crystallization protocol was developed as follows. The Choline Fenofibrate isolated from the
evaporative crystallization was dissolved in a certain ratio of MeOH/IPA by heating to a defined
temperature. Additional IPA was charged to the solution to generate a supersaturated solution.
The supersaturated solution was seeded with 1-5 wt.% (relative to the total amount of API in
solution) of milled Choline Fenofibrate seed (Thin plates, D(v, 0.9) of 66 microns). Additional
IPA was added at the seeding temperature at a controlled rate to crystallize more product,
7
ACS Paragon Plus Environment

Crystal Growth & Design
Page 8 of 39
1
2
3
4
5
6
following which the slurry was cooled to 20 ºC. The API was isolated by filtration and dried
under vacuum at 50 ºC.
7
8
9
The solution supersaturation at seeding is a function of the API concentration, MeOH/IPA ratio
and temperature. Table 1 lists some of the key experiments that demonstrate the impact of
supersaturation on crystal habit. The supersaturation ratio (C/C*) is defined as the ratio of API
concentration in solution (C) to the API solubility at those conditions (C*). Different
supersaturation ratios at the seeding point were obtained through various combinations of API
concentration, MeOH/IPA ratio and temperature. In each of these experiments (except 10 and
11), addition of seeds was followed by IPA addition over 30 min at the seeding temperature to
reduce the MeOH/IPA ratio to 1/4. Since the API solubility decreases as the MeOH/IPA ratio
decreases, addition of IPA post seeding was necessary for yield recovery. Following the IPA
addition, the slurry was cooled to 20 °C at 10 °C/hr in order to recover more product. The crystal
habit of isolated API is also reported in Table 1. This habit matched the habit seen after seeding
and prior to IPA addition.
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
39
40
41
42
43
44
45
46
47
48
49
50
51
52
53
54
55
56
57
58
59
60
8
ACS Paragon Plus Environment

Page 9 of 39
Crystal Growth & Design
1
2
3
4
Table 1. Crystallization Experiments using MeOH-IPA
5
6
7
8
9
Experiment #
API
Concentration1
MeOH/IPA
Ratio1
Seeding
Supersaturation
Ratio1
Product
Temperature
Morphology
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
39
40
41
42
43
44
45
46
47
48
49
50
51
52
53
54
55
56
57
58
59
60
(wt.%)
(C/C*)
1.7
(°C)
50
25
50
50
50
50
50
50
22
22
25
1
2
20.1
9.5
57/43
58/42
59/41
57/43
50/50
80/20
80/20
80/20
40/60
75/25
75/25
Thick Plates
Thick Plates
Thick Plates
Thick Plates
Thick Plates
Thick Plates
Thick Plates
Thick Plates
Prisms
1.5
3
20.7
19.5
17.2
38
1.6
4
1.6
5
1.7
6
1.50
1.42
1.35
2.5
7
36
8
34
9
7.2
10
11
12.6
12.3
1.16
1.1
Thin Plates
Thin Plates
1Value at Seeding Conditions.
The experiments in Table 1 indicate that crystals with predominantly thick plate habit are
obtained at a supersaturation ratio in the range of 1.3-1.7. Various combinations of the seeding
temperature, MeOH/IPA ratio and API concentration were used to generate supersaturation in
this range and all of those experiments produced thick plates. Higher levels of supersaturation
(C/C* > 2) produced crystals with predominantly prism-like morphology and lower levels of
9
ACS Paragon Plus Environment

Crystal Growth & Design
Page 10 of 39
1
2
3
4
5
6
7
8
supersaturation (C/C* ~ 1.1) resulted in predominantly thin-plate morphology. Representative
scanning electron micrographs (SEM) and optical microscope images of the three habits are
shown in Figure 4.
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
39
40
41
42
43
44
45
46
47
48
49
50
51
52
53
54
55
56
57
58
59
60
PRISMS
THICK PLATES
THIN PLATES
Figure 4. SEM micrographs and Optical Microscope images of Choline Fenofibrate habits
These experimental results validated our hypothesis about the role of supersaturation in defining
the crystal habit of Choline Fenofibrate crystallized from MeOH-IPA solvent system. In
addition, they provided the framework to design the crystallization process to deliver API with
thick plate habit. The manufacturing process was designed as follows. The first isolation of
Choline Fenofibrate followed the evaporative crystallization process from MeOH-IPA described
earlier. The isolated API was dissolved in 80/20 MeOH/IPA at 60 °C at a concentration of 34 %
by weight of the solution. The solution was gradually cooled to 50 °C to generate a
supersaturation ratio of about 1.4 and seeded with 1 wt.% seed (relative to the total amount of
API in solution). The seeds were added as a slurry in IPA. After seeding, a preheated charge of
IPA was gradually added maintaining a temperature of 50 °C, to obtain a MeOH/IPA solvent
10
ACS Paragon Plus Environment

Page 11 of 39
Crystal Growth & Design
1
2
3
4
5
6
7
8
9
ratio of 30/70. The slurry was cooled to 20 °C over 2 hours. The MeOH content of the slurry
was further reduced through a constant volume distillation under vacuum at ~20 °C. The
constant volume was maintained by charging IPA during the distillation. The API was isolated
by filtration and dried under vacuum at approximately 50 °C. These process parameters were
chosen for optimal throughput and yield while maintaining good operational robustness. The
distillation operation after the cooling step reduces the MeOH/IPA ratio and thereby offers a
higher yield of the product without any negative impact on the thick plate crystal habit.
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
39
40
41
42
43
44
45
46
47
48
49
50
51
52
53
54
55
56
57
58
59
60
Other crystallization parameters were also studied for their effect on crystal habit. The effect of
cooling rate was investigated by varying the cooling rate after IPA addition at 50 °C. Cooling
rates of 5 °C/h and 15 °C/h both produced the target thick plate morphology. Also, crystallization
experiments were conducted without seeding to determine the impact on crystal morphology. In
the absence of seeds, the solution nucleates either during the IPA addition at 50 °C or during
cooling to 20 °C. Thus, nucleation occurs at a higher supersaturation compared to the seeded
process and results in prism morphology, which is also acceptable for the formulation process.
Although these studies indicated robustness of the process to deliver the target crystal habit,
seeding and controlled cooling at a predefined rate were incorporated in the manufacturing
process, for consistency.
The Choline Fenofibrate manufacturing process was transferred to two manufacturing sites. One
of these sites preferred not to use MeOH in their plant. In order to accommodate this need, other
solvents were screened for use in the reaction and crystallization processes. The solvent system
for salt formation reaction was switched from MeOH to IPA-water and the crystallization solvent
was switched to ethanol (EtOH). We will briefly discuss the crystallization process using EtOH
and the effect of supersaturation on crystal habit in this solvent. The first isolation of Choline
Fenofibrate in this modified process occurs from a mixture of IPA-water (predominantly IPA
with low concentration of water) and the filtered cake is washed with IPA. In order to streamline
the unit operations in the process, the wet cake from the first isolation is carried directly into the
crystallization process without drying. Similar to the earlier manufacturing process, the second
crystallization using EtOH in this case, serves to control the crystal habit of the API. Since the
product from the first isolation is not dried, it carries some amount of IPA into the next
11
ACS Paragon Plus Environment

Crystal Growth & Design
Page 12 of 39
1
2
3
4
5
6
crystallization. In light of this, we measured the solubility of Choline Fenofibrate in EtOH-IPA
solvent mixture at various temperatures. The solubility data are shown in Figure 5.
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
39
40
41
42
43
44
45
46
47
48
49
50
51
52
53
54
55
56
57
58
59
60
20
18
16
14
12
10
8
25C
50C
70C
6
4
2
0
0
20
40
60
80
100
Wt.% EtOH in EtOH-IPA mixture
Figure 5. Solubility of Choline Fenofibrate in EtOH-IPA as a function of temperature
Additional solubility values at intermediate temperatures were calculated using the van’t Hoff
equation. Solubility at intermediate solvent compositions not experimentally measured, was
estimated via interpolation of the above data. Using these data, a crystallization protocol was
developed as follows. The Choline Fenofibrate isolated from the evaporative crystallization was
dissolved in EtOH by heating to 75 ºC. The solution was cooled to a defined temperature to
generate supersaturation. The supersaturated solution was seeded with 1-5 wt.% (relative to the
total amount of API in solution) of finely milled Choline Fenofibrate seed. After seeding, the
12
ACS Paragon Plus Environment

Page 13 of 39
Crystal Growth & Design
1
2
3
4
5
6
crystallized slurry was cooled at a controlled rate to 0 ºC to recover more product. The API was
isolated by filtration and dried using heat and vacuum.
7
8
9
The solution supersaturation at seeding is a function of the API concentration, EtOH/IPA ratio
and temperature. Typical amounts of IPA brought in by the wet Choline Fenofibrate product
from the first isolation resulted in less than 5 wt.% IPA relative to EtOH in the final
crystallization process. The effect of supersaturation at seeding, on the API crystal morphology
was investigated by conducting crystallization experiments at different levels of supersaturation.
These experiments are summarized in Table 2. The supersaturation at seeding was varied by
changing the API concentration in solution, the EtOH/IPA solvent ratio and the seeding
temperature. Through various combinations of these three parameters, a range of supersaturation
ratios from 1.07 to 1.89 was investigated. After seeding, the crystallization slurry was cooled to
0 °C and filtered. None of the experiments resulted in thin plate morphology. Unlike the
MeOH-IPA crystallization process, thick plate morphology can be obtained at low
supersaturation ratios (C/C*<1.1) in the ethanol crystallization. This suggests that solvent
interactions play a significant role in determining the crystal habit of Choline Fenofibrate. The
effect of low supersaturation on crystal morphology is much less pronounced when EtOH is used
as the crystallization solvent. This allows for a wider range of operating conditions for the EtOH
crystallization process. Crystallization without seeding results in nucleation at a lower
temperature during cooling. Since the supersaturation at the lower nucleation temperature is
higher, the resultant API exhibits prism morphology. Other experiments indicated that varying
the cooling rate during crystallization did not have an adverse effect on the crystal habit.
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
39
40
41
42
43
44
45
46
47
48
49
50
51
52
53
54
55
56
57
58
59
60
13
ACS Paragon Plus Environment

Crystal Growth & Design
Page 14 of 39
1
2
3
4
Table 2. Crystallization Experiments using EtOH-IPA
5
6
7
8
9
Seeding
Supersaturation
Ratio1
API
Concentration1
EtOH/IPA
Ratio1
Product
Temperature
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
39
40
41
42
43
44
45
46
47
48
49
50
51
52
53
54
55
56
57
58
59
60
Experiment #
Morphology
(wt.%)
(C/C*)
1.53
1.23
1.18
1.28
1.52
1.15
1.45
1.15
1.07
1.89
(°C)
60
1
2
18
14
95/5
97.5/2.5
100/0
Thick Plates
Thick Plates
Thick Plates
Thick Plates
Thick Plates
Thick Plates
Thick Plates
Thick Plates
Thick Plates
Prisms
58
3
14
58
4
14
95.1/4.9
97.5/2.5
98/2
58
5
17.3
13.2
14
58
6
58
7
99.5/0.5
97.5/2.5
95/5
52.5
60
8
14
9
18
70
10
15
95/5
50
1Value at Seeding Conditions.
The API solubility is not a strong function of solvent composition in the EtOH-IPA system at 0
ºC. As a result, there is no significant increase in the crystallization yield by going to a higher
IPA/EtOH ratio. However, the volume efficiency of the process is negatively impacted at higher
IPA/EtOH ratios. Hence the crystallization process in EtOH-IPA was designed to be a cooling
crystallization process. The manufacturing process is operated at an API concentration of 14 wt.
% in a 97.5/2.5 EtOH/IPA solution. The seeding temperature is 58 °C, which corresponds to a
supersaturation ratio of about 1.2. About 1 wt.% seed (relative to the total amount of API in
solution) is used. The seeds are added as a slurry in IPA. Following seeding, the slurry is cooled
14
ACS Paragon Plus Environment

Page 15 of 39
Crystal Growth & Design
1
2
3
4
5
6
to 0 °C over 4 h (linear cooling rate of 15 °C/h) and filtered. The product is dried under vacuum
at approximately 50 °C with a nitrogen purge.
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
39
40
41
42
43
44
45
46
47
48
49
50
51
52
53
54
55
56
57
58
59
60
4. Effect of impurity on crystal habit of Choline Fenofibrate
During development of the manufacturing process for the API, low levels of a polymeric
impurity were encountered in several lots of the starting material Fenofibrate. This impurity had
an adverse effect on the crystal habit of Choline Fenofibrate. These observations will be
discussed in this section.
Isopropyl 2-bromo-2-methylpropionate is a starting material in the Fenofibrate synthesis. A
polymeric impurity, poly-isopropylmethacrylate is formed in the Fenofibrate synthesis, from this
starting material. The Fenofibrate process involves crude product isolation, followed by a
recrystallization, which ensures removal of this polymeric impurity to undetectable levels. In
order to increase the throughput of the Fenofibrate process, the manufacturer was considering
removal of the Fenofibrate recrystallization. Fenofibrate lots made from the single isolation
process contained significant levels of the polymeric impurity (up to 0.5 wt.%). Although the
impurity levels were reduced during the first isolation of Choline Fenofibrate (evaporative
process), the impurity was not completely removed. When these lots of API were carried forward
through the second crystallization for habit control, thin plates were obtained. The polymer
impurity inhibits growth of the most dominant crystal face of API, resulting in the formation of
thin-plates. Several lots of Choline Fenofibrate from evaporative crystallization, with varying
levels of polymer were crystallized using the MeOH-IPA process to determine the sensitivity of
the morphology to polymer concentration. The polymer concentrations in these lots were
determined by gravimetric analysis. This technique gives a conservative number for the polymer
concentration since it measures the total mass of insoluble matter in the sample. The presence of
polymer impurity was confirmed by IR spectroscopy. A few key crystallization experiments with
Choline Fenofibrate lots containing polymer impurity are summarized in Table 3.
15
ACS Paragon Plus Environment

Crystal Growth & Design
Page 16 of 39
1
2
3
4
Table 3. Effect of Polymer Impurity on API Crystal Habit
5
6
7
8
9
Expt. #
Polymer
Concentration Concentration1
API
MeOH/IPA Seeding Supersaturation Product
Ratio1
Temp.
Ratio1
Morphology
in Choline
Fenofibrate
isolated from
evaporative
crystallization
(wt.%)
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
39
40
41
42
43
44
45
46
47
48
49
50
51
52
53
54
55
56
57
58
59
60
(C/C*)
(°C)
(wt.%)
0.4
1
2
3
36
36
36
80/20
80/20
80/20
50
50
50
1.42
1.42
1.42
Thin Plates
Thin Plates
Thin Plates
0.3
0.1
1Value at Seeding Conditions.
These experiments indicate that the API crystal morphology is adversely affected by the polymer
impurity even at a very low concentration (0.1 wt.% relative to Choline Fenofibrate). This
phenomenon was confirmed by conducting crystallization experiments under identical
conditions, using polymer free lots of Choline Fenofibrate spiked with a poly-
isopropylmethacrylate standard with an average molecular weight of 100,000. The polymer
standard was spiked at two different levels, 1.5 wt.% and 0.2 wt.% (relative to Choline
Fenofibrate). Both experiments resulted in API with thin-plate morphology. A comparison of
the API habits obtained from identical crystallization experiments, with and without the polymer
impurity, is shown in Figure 6.
16
ACS Paragon Plus Environment

Page 17 of 39
Crystal Growth & Design
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
39
40
41
42
43
44
45
46
47
48
49
50
51
52
53
54
55
56
57
58
59
60
API crystallized in the presence of 0.2 wt. % polymer
API crystallized in the absence of polymer
Figure 6. Effect of Polymer Impurity on API Crystal Morphology
Similar to the MeOH-IPA crystallization process, the polymer impurity had an adverse effect on
the API crystal morphology for the EtOH based crystallization. Even crystallization experiments
conducted at higher supersaturation using either process produced API with thin plate habit when
the polymer impurity was present. We will provide some fundamental insight into the effect of
this polymer impurity on crystal habit later in the paper.
As a result of these findings, the proposed modifications to the Fenofibrate process were
abandoned. The recrystallization in the Fenofibrate process ensures removal of the polymer
impurity, thereby eliminating risk to Choline Fenofibrate crystal morphology.
5. Experimental insights into the crystal growth mechanism of Choline Fenofibrate
The effect of supersaturation on crystal habit of Choline Fenofibrate crystallized using the
MeOH-IPA process prompted us to probe the growth mechanism of the most dominant face of
the crystal. A possible explanation for change in habit with supersaturation is a change in the
growth mechanism of the most dominant face. A detailed discussion of the crystal lattice and
17
ACS Paragon Plus Environment

Crystal Growth & Design
Page 18 of 39
1
2
3
4
5
6
insights into the morphology are included in the next section of the paper. In this section, we
present some experimental evidence for the growth mechanism.
7
8
9
We obtained AFM images of API crystals grown from pure methanol at a very low
supersaturation. These images are shown in Figure 7. The images were obtained from different
crystals grown in the same batch crystallizer. Image A shows steps with uniform spacing, which
may be indicative of spiral growth originating from a screw dislocation [3]. Images B and C
show 2-D nuclei on the dominant face of the crystal. The spacing between steps originating from
2-D nuclei depends on the nucleation rate and the step velocity, and may not be uniform [6]. It is
likely that both spiral growth and 2-D nucleation mechanisms may be at play, either due to
spatial variations in supersaturation in the batch crystallizer, or because the supersaturation is
near the transition region between the two mechanisms (Figure 16).
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
39
40
41
42
43
44
45
46
47
48
49
50
51
52
53
54
55
56
57
58
59
60
18
ACS Paragon Plus Environment

Page 19 of 39
Crystal Growth & Design
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
39
40
41
42
43
44
45
46
47
48
49
50
51
52
53
54
55
56
57
58
59
60
A
B
C
Figure 7. AFM images of most dominant face of API crystal obtained from MeOH at very low
supersaturation
19
ACS Paragon Plus Environment

Crystal Growth & Design
Page 20 of 39
1
2
3
4
6. Mechanistic insights into the crystal habit of Choline Fenofibrate
5
6
7
8
9
Although mechanistic/predictive studies on crystal morphology of organic salts are scarce in the
open literature, it is a very important research focus for pharmaceutical companies. The classical
attachment energy method that is most commonly used for morphology prediction is not
mechanistic and does not give reliable predictions for certain systems such as needles. It also
cannot explain the effect of supersaturation on crystal habits. The method developed for non-
centrosymmetric growth units by Kuvadia & Doherty [4] applies conceptually to salt
compounds. In this work, our primary aim is to identify accurately the morphologically relevant
slow growing crystal faces of Choline Fenofibrate and based on this prediction, understand the
fundamental reasons behind the interesting morphological behavior displayed by this system as
described in the previous sections.
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
39
40
41
42
43
44
45
46
47
48
49
50
51
52
53
54
55
56
57
58
59
60
6.1 Crystal structure of Choline Fenofibrate
Choline Fenofibrate crystallizes in the space group Pbca (orthorhombic) with the lattice
parameters a=12.009 Å; b= 12.275 Å and c =28.281 Å as identified by single crystal X-ray
diffraction measurements. A single unit cell consists of 8 negatively-charged Fenofibrate ions
and 8 positively charged Choline counter-ions, having 8 symmetry operators, as shown in
Figures 8 a) and 8 b). Hence, the overall stoichiometry of the packing structure is 1:1 with
respect to Choline and Fenofibrate and a total of 16 ions comprise one unit cell that can be
translated to generate packing in all directions. The first step in the method for predicting the
crystal morphology is to use a suitable force-field to estimate the solid-solid intermolecular
interactions for this organic salt crystal. We first calculate the partial charge on each atom of the
Fenofibrate and the Choline ions within the unit cell using GAUSSIAN 03 [7] and then use the
AMBER force field [8,9] and HABIT 95 to calculate the intermolecular interactions within the
crystal lattice. The calculated value of lattice energy using this approach is -125 kcal/mol which
is in the range of values expected for organic salts; higher than the typical range for organic
molecular crystals because of the high electrostatic component of the intermolecular interactions.
This value of lattice energy is the converged value for a suitably large number of unit cells. (This
is especially important to verify for ionic systems and provides confidence in the force field
chosen).
20
ACS Paragon Plus Environment

Page 21 of 39
Crystal Growth & Design
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
39
40
41
42
43
44
45
46
47
48
49
50
51
52
53
54
55
56
57
58
59
60
a)
21
ACS Paragon Plus Environment

Crystal Growth & Design
Page 22 of 39
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
39
40
41
42
43
44
45
46
47
48
49
50
51
52
53
54
55
56
57
58
59
60
b)
Figure 8: Crystal packing structure of Choline Fenofibrate viewed perpendicular to the b axis.
a) 4 unit cells shown, coloring scheme used to show different elements (gray =Carbon,
cyan=Hydrogen, violet=Nitrogen, Red=Oxygen). b) Coloring based on symmetric equivalence,
green represents Fenofibrate ions and blue represents Choline ions.
6.2 Morphology prediction of Choline Fenofibrate
In order to obtain theoretical understanding of the crystal morphology, we need to identify the
slow growing morphologically relevant faces based on the bonding structure. Since the
compound is a salt (ionic compound), certain additional rules must be imposed to identify stable
crystal slices (surfaces) in each direction based on the electrostatic dipole moment. According to
Tasker [10], a stable slice must be stoichiometric in composition and also should not have any
dipole moment perpendicular to the slice. Additionally, for the particular slice to be slow
growing, the slice must contain periodic bond chains in at least two different directions [11].
Moving perpendicular to the c axis, there are two layers of Fenofibrate ions followed by two
layers of Choline ions stacked along the {002} direction repeating to form the packing structure.
We use the rules of stability of ionic crystal surfaces [10] to determine which is the stable slice in
the {002} direction. Stable surfaces have either equal anions and cations on each plane (Tasker
type 1) or charged individual planes with no net dipole moment perpendicular to the surface
(Tasker type 2) [10,11]. The {002} planes are of type 2 and there are two possible
22
ACS Paragon Plus Environment

Page 23 of 39
Crystal Growth & Design
1
2
3
4
5
6
7
8
9
configurations, as shown in Figure 9. We select the slice with the least repulsive energy within
the slice to be the most stable configuration (the repulsive interaction between the two Choline
layers is significantly more than the repulsive interaction between the two Fenofibrate layers as
the Choline ions are more closely packed in the unit cell due to their smaller size). The most
stable slice configuration for the {002} planes is shown in Figure 10b). All the other
morphologically relevant planes, {022}, {111}, {020} and {200} are Tasker type 1 surfaces.
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
39
40
41
42
43
44
45
46
47
48
49
50
51
52
53
54
55
56
57
58
59
60
23
ACS Paragon Plus Environment

Crystal Growth & Design
Page 24 of 39
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
39
40
41
42
43
44
45
46
47
48
49
50
51
52
53
54
55
56
57
58
59
60
a)
b)
Figure 9a) and b). Packing structure representation showing {002} slices and repeat units in
terms of charged layers, viewed perpendicular to b axis. Two configurations which give zero
dipole moment perpendicular to the surface. The molecules are represented by their centers of
mass, shown by solid circles of different colors, and the lines represent intermolecular
interactions.
24
ACS Paragon Plus Environment

Page 25 of 39
Crystal Growth & Design
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
39
40
41
42
43
44
45
46
47
48
49
50
51
52
53
54
55
56
57
58
59
60
a)
b)
Figure 10 a) and b). Two configurations of the {002} slice based on slice shifting with zero
dipole moment. b) shows the most stable slice with least repulsive interactions within the slice.
Figure 11. Intermolecular interactions and the correct choice of a building unit for Choline
Fenofibrate crystal.
25
ACS Paragon Plus Environment

Crystal Growth & Design
Page 26 of 39
1
2
3
4
5
6
7
8
9
The periodic bond chains (PBCs) in Choline Fenofibrate crystal should satisfy all the properties
of PBCs given by Hartman and Perdok [12]. Specifically, the PBCs should have stoichiometric
composition and there should be no net dipole moment perpendicular to the PBC direction. The
concept of a building unit has been proposed in the ionic crystal growth literature [13] that is
used to identify the periodic bond chains in any ionic crystal. This concept was also used here to
identify a building unit for the PBCs in Choline Fenofibrate.
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
39
40
41
42
43
44
45
46
47
48
49
50
51
52
53
54
55
56
57
58
59
60
A building unit is a stoichiometric arrangement of atoms in the solid-state such that mere
translation of this arrangement in the symmetry-related directions in the lattice yields all the
periodic bond chains in the lattice. The building unit should be stoichiometric and should have a
zero dipole moment, only then can a single building unit generate PBCs in all directions in
keeping with the slice stability rules. We select the building unit whose dipole moment is zero, as
shown in Figure 11. It should be noted that the selected building unit contains an inversion centre
within it, so that the entire bonding structure is transformed to a centrosymmetric one which is
simpler to deal with.
It is important to note that the building units are not the exact growth units that actually attach to
kink sites and new methods are being developed to enable a detailed mechanistic relationship
between building units of periodic bond chains, growth units and kink densities and kink rates
[14]. In this work, the building unit kink site approximation is made to get workable estimates
for the crystal morphology of this complicated system.
Once the building unit is identified, all the PBCs that lie within the slice of any face of the crystal
can be determined and PBC network diagrams can be drawn for each of the important faces of
the crystal. The periodic bond chains for the slow growing F faces are shown in Figure 12. This
is followed by the calculation of the PBC energies. In the case of organic salts, the presence of
ions in the solid state makes it necessary to consider the long-range nature of the electrostatic
interactions. Once a PBC direction has been identified, the interactions of a central building unit
with all the other building units in that PBC direction were summed up to give the overall
strength of intermolecular interactions along that PBC direction. The calculation of the PBC
energies along with the PBC network diagram for each crystal face allows for an accurate
prediction of the prominent edges on the growth spiral and the shape of the spiral.
26
ACS Paragon Plus Environment

Page 27 of 39
Crystal Growth & Design
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
39
40
41
42
43
44
45
46
47
48
49
50
51
52
53
54
55
56
57
58
59
60
Figure 12. The periodic bond chain representation for different F faces of Choline Fenofibrate
using the building unit described in Figure 11.
Thus, the periodic bond chain analysis using building units helps in predicting the set of slowest
growing faces in Choline Fenofibrate and this can be combined with preliminary application of
the spiral growth model [4] to predict the relative growth rates of each of those faces (Table 4)
and estimate the steady-state crystal morphology of the material grown under low
supersaturation conditions.
27
ACS Paragon Plus Environment

Crystal Growth & Design
Page 28 of 39
1
2
3
4
5
6
7
Table 4. Spiral growth model calculations on all the F-faces of Choline Fenofibrate
8
9
Bond
Kink
Step velocity
Interplanar
Relative Growth
rate (R)
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
39
40
41
42
43
44
45
46
47
48
49
50
51
52
53
54
55
56
57
58
59
60
Face
chain
direction
density (dimensionless) Spacing dhkl (Å)
[110]
[-110]
[100]
[011]
[110]
[101]
[100]
[10-1]
[101]
[011]
[01-1]
[011]
[01-1]
5.8E-06
5.8E-06
0.00184
0.0003
5.1E-05
5.1E-05
0.01123
0.00264
9.1E-05
0.14265
0.0257
{002}
14.14
8.21
6.13
1
{111}
{020}
5.8E-06
0.01554
0.00184
0.01554
0.01554
0.0003
4.8
0.18651
0.18651
0.00362
0.02753
0.00465
0.00465
551.3
{022}
{200}
4.635
4.635
39.0
27.1
0.00184
0.0003
0.0003
We estimate the morphology of Choline Fenofibrate using the spiral growth mechanistic model
to be plate-like (low to moderate thickness), with the {002} face being morphologically the most
dominant face (Figure 13) followed by the {111} family. (All other faces grow out of the
predicted final steady-state shape). This morphology is similar to that predicted using the
attachment energy model (Figure S1a). However, it is well known that the spiral growth model
provides a better mechanistic understanding of crystal growth of molecular crystals [4]. It is
important to note once more that the building unit and periodic bond chain-spiral growth method
used here is our first attempt to explain the growth mechanism of organic salts crystallized from
28
ACS Paragon Plus Environment

Page 29 of 39
Crystal Growth & Design
1
2
3
4
5
6
7
8
9
solution. The exact relation between kink sites and building units remains unknown at this stage.
Also, this morphology estimation is carried out in vacuum. Despite the simple method adopted, it
is one of the higher fidelity methods which correctly captures the behavior of F faces and
identifies the morphologically relevant faces and their relative importance to morphology. There
are ongoing efforts to develop the capability to understand the relationship between building
units and actual kink site growth units, and also estimate the effect of mixed solvents on crystal
morphology for these types of systems.
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
39
40
41
42
43
44
45
46
47
48
49
50
51
52
53
54
55
56
57
58
59
60
Figure 13. Predicted morphology of Choline Fenofibrate crystal using the spiral growth model
at low supersaturation.
6.3 Change of morphology with impurity
Even a small amount of additive causes significant modification to the crystal habit. Since a
morphology change can only be brought about if the additive affects the slow growing F faces
29
ACS Paragon Plus Environment