Drug Delivery and Translational Research

https://doi.org/10.1007/s13346-018-00592-0

ORIGINAL ARTICLE

Neuro-fuzzy modeling of ibuprofen-sustained release from tablets based

on different cellulose derivatives
Samia Rebouh 1,2 & Sonia Lefnaoui 1,2 & Mounir Bouhedda 3 & Madiha M. Yahoum 2 & Salah Hanini 1

# Controlled Release Society 2018

Abstract
In the present study, we investigated the drug release behavior from cellulose derivative (CD) matrices in the oral form of tablets.
We used the adaptive neural-fuzzy inference system (ANFIS) to predict the best formulation parameters to get the perfect
sustained drug delivery using ibuprofen (IB) as a model drug. The different formulations were prepared with different CDs,
namely CMC, HEC, HPC, HPMC, and MC. The amount of the active ingredient varied between 20 and 50%. The flow
properties of the powder mixtures were evaluated for their angle of repose, compressibility index, and Hausner ratio, while the
tablets were evaluated for weight uniformity, hardness, friability, drug content, disintegration time, and release ratio. All tablet
formulations presented acceptable pharmacotechnical properties. In general, the results showed that the drug release rate in-
creases with an increase in the loaded drug. Kinetic studies using the Korsmeyer-Peppas equation showed that different drug
release mechanisms were involved in controlling the drug dissolution from tablets. The drug release mechanism was influenced
by the gel layer strength of the CDs formed in the dissolution medium. The mean dissolution time (MDT) was determined and the
highest MDT value was obtained for the HPMC formulations. Moreover, HPMC exhibited release profiles adequate for sustained
release formulations for over 14 h. The intelligent model based on the experimental data was used to predict the effect of the
polymer’s nature, the amount of the active ingredient, and the kinetic release profile and rate (R2 = 0.9999 and RMSE = 5.7 ×
10−3). The ANFIS model developed in this work could accurately model the relationship between IB release behavior and tablet
formulation parameters. The proposed model was able to successfully describe this phenomenon and can be considered an
efficient tool with predictive capabilities that is useful for the designing and testing of new dosage systems based on polymers.

Keywords ANFIS . Cellulose derivatives . Matrix tablets . Modeling . Sustained release

* Samia Rebouh Introduction

samrebouh@yahoo.fr

Sonia Lefnaoui Over the past three decades, research in the field of sustained
lefnaoui_sonia@yahoo.fr drug release has focused on controlling drug release profiles to
ensure an effective serum concentration for an extended peri-
Mounir Bouhedda
bouhedda.m@gmail.com od of time [1–4]. The development of a sustained release
formulation reduces the side effects associated with a massive
Madiha M. Yahoum
madihamelha@yahoo.fr
release of the active principle and reduces the number of med-
ications taken by prolonging the analgesic and anti-
Salah Hanini inflammatory actions. This also makes it possible to overcome
s_hanini2002@yahoo.fr
the problem of active ingredients with low half-lives [5–7].
1
Laboratory of Biomaterials and Transport Phenomena (LBMPT), Before designing a sustained release system, the pathway
University of Medea, Nouveau pôle urbain, Medea University, for the liberation of active ingredients should be chosen ac-
26000 Medea, Algeria cording to several considerations, i.e., the physical and chem-
2
Faculty of Sciences, University of Medea, Nouveau pôle urbain, ical properties of the drug, active ingredient doses, route of
Medea University, 26000 Medea, Algeria administration, type of release system, desired drug effect,
3
Laboratory of Advanced Electronic Systems (LSEA), University of physiological release of the drug from the delivery system,
Medea, Nouveau pôle urbain, Medea University, bioavailability of the drug at the absorption site, and the
26000 Medea, Algeria

pharmacodynamics of the drugs [8–10]. IB is a non-steroidal
anti-inflammatory drug that also has analgesic and antipyretic
properties. It was used as a model compound because it is
highly lipophilic and is rapidly excreted from the body. It is
a small molecule (MW = 206) with a log D (pH = 7.0) value of
1.16, a log P value of 3.722, and its pKa value is 4.4 [5–7].
IB has poor water solubility and is therefore a good candi-
date for formulation into a modified release system because of
its short half-life (approximately 2 h) [11–13]. It is readily
absorbed in the gastrointestinal tract; however, it can cause
irritation of the gastric mucous membrane. Sustained release
of this drug in a hydrophilic matrix would therefore help re-
duce this irritation. To control the release of the active ingre-
dient over time, it was formulated to be contained in a Bdrug
carrier^ and encapsulated, coated, or retained in a matrix from
which it was progressively released [14–18].
Polymer-based matrices have become an integral part of
the pharmaceutical industry and hydrophilic matrices are ca-
pable of sustaining drug release over an extended period
[19–21]. These are compressed matrix tablets of a drug pow-
der mixture based on swellable polymer and other excipients
distributed throughout the matrix. These matrices are mostly
employed for their advantages, e.g., their simple formulation
and the use of low-cost safety polymers [22–24]. Furthermore,
these release systems have the ability to release the drug over
an extended period because of their longer disintegration time.
Most CDs are obtained via etherification of cellulose,
which involves the reaction of the hydroxyl groups of cellu-
lose with methyl and ethyl units [25]. The excipients are based
on cellulose ethers, such as methylcellulose (MC), hydroxy-
propyl methylcellulose (HPMC), hydroxypropyl cellulose
(HPC), hydroxyethyl cellulose (HEC), and sodium carboxy-
methylcellulose (CMC), which are among the most widely
used CDs [26]. These CDs have gained popularity in the for-
mulation of oral hydrophilic matrices due to their swelling
properties and nontoxic nature. Additionally, cellulose ethers
have good compression characteristics, meaning they can be
directly compressed to form sustained release swellable ma-
trices and have a high capacity for drug loading [27–30].
Mathematical models are useful in the prediction of release
kinetics before the release systems are realized. They can also
be used in the calculation of key drug release parameters, such
as the diffusion coefficient. The development of mathematical
models, which requires an understanding of the involved
physicochemical phenomena and the influences of the drug
release kinetics, plays an important role in optimizing the for-
mulation process [31]. Most of the mathematical functions
used to describe the release profile are complex and nonlinear;
however, there is no single accepted approach for determining
release kinetic profiles [32–35].
This has led researchers to exploit artificial intelligent tech-
niques when modeling such complex systems is done by using
input-output data sets. Fuzzy logic, which is a widely used
Drug Deliv. and Transl. Res.
model, allows system modeling using experience and human
knowledge with if–then rules, and the fuzzy parameters must
be turned effectively. On the other hand, artificial neural net-
works (ANNs) can be advantageous because they provide an
adjustment mechanism and can be used to optimize the fuzzy
system [36]. The combination of ANNs with fuzzy systems
creates a new robust model called an adaptive neural-fuzzy
inference system (ANFIS) [37].
Fuzzy set theory has been applied to a wide range of prob-
lems in automatic control, signal treatment, data clustering,
pattern recognition, and event classification. ANFIS has prov-
en its effectiveness in modeling nonlinear functions [38]. A
fuzzy system can learn the characteristics of a data set and then
adjust the characteristics of the system according to a given
error criterion. Drug release behavior that presents nonlinear-
ities in the operating parameters must be described accurately
to get an effective prediction model. To build the model, the
designer must use one of two methods. The first is mathemat-
ical modeling, which involves knowledge of chemistry, biol-
ogy, and other sciences and would be extremely difficult for
the current research. The second requires experimental data to
identify a relationship between outputs and inputs [39]. This
method is used when the modeling process involves extreme-
ly complex physical phenomena or has strong nonlinearities
and dependence on various parameters. It was therefore suit-
able for the current research.
The current investigation dealt with the optimization of
hydrophilic matrices for IB containing gel-forming polymers
based on CDs. The principal benefit of this approach is the
possibility of simulating the effect of the design parameters of
CD-based drug delivery systems on the release profiles. The
study required the use of obtained experimental data to iden-
tify the relationship between outputs and inputs using an
ANFIS.
Materials and methods
Materials
The active ingredient, IB, was obtained from Antibiotical
Laboratories (Saidal industries, Algeria). The CDs were pur-
chased from Colorcon (UK) and included hydroxypropyl
methyl cellulose (HPMC, Premium Methocel K4M, Mw ≈
95,000 g/mol., methoxyl groups = 28.9%, and
hydroxypropoxyl groups = 9.2%), hydroxypropyl cellulose
(HPC, Mw ≈ 1.150,000 g/mol., molar substitution = 3.7),
methyl cellulose (MC, Mw ≈ 41,000 g/mol., degree of substi-
tution = 1.6), hydroxyethyl cellulose (HEC, Mw ≈ 720,000 g/
mol., molar substitution = 2.5), and carboxymethyl cellulose
sodium salt (CMC, Mw ≈ 250,000 g/mol., degree of substitu-
tion = 0.7). Lactose monohydrate (LM) and magnesium stea-
rate (MGS) were supplied by Polypharma-GMBH
Drug Deliv. and Transl. Res.

(Germany). All other chemicals were of analytical or pharma- machine (ERWEKA, GmbH, Germany). The compression
ceutical grade and were used without further purification. speed was one tablet per 3 s.

Methods Micromeritic studies

Formulation of the matrix tablets based on CDs Prior to compression, the flowability of the powder blends
was assessed by determining several parameters, including
Tablets weighing 300 mg and containing different levels (20, angle of repose, Hausner’s ratio, and Carr’s index. To achieve
30, 40, and 50%) of the model drug (IB) were prepared by an accurate result, each test was repeated three times and the
direct compression. The compositions of the various formula- average was taken (see Table 3). The angle of repose was
tions employing 30% (w/w) CDs (CMC, MC, HEC, HPC, or determined according to the funnel method, and the mean
HPMC), LM (as diluent), and MGS (2%, w/w, as a lubricant) diameter of the base of the powder mixture cone and the
are summarized in Table 1. Three hundred grams of tableting tangent of angle of repose were determined using Eq. (1) [41].
mixture was prepared for each formulation (components are

h
listed in Table 1). Overall, 20 formulations of the tableting θ ¼ tan−1 ð1Þ
mixtures were prepared. All the ingredients except the lubri- r
cant were sieved through a 450-μm mesh and mixed with IB
where θ is the angle of repose, his the cone height, and r is the
in a cubic mixer (ERWEKA AR400, Germany) at 30 rpm for
cone base radius.
12 min to obtain a uniform mixture. At the end, the MGS was
Hausner’s ratio and Carr’s index were, respectively, calcu-
sieved and added as a lubricant to the powder blend, which
lated using Eqs. (2) and (3) deduced from the USP method [40].
was then mixed for an additional minute.
The powder bed was evaluated for its flowability properties ρtapped
HR ¼ ð2Þ
according to the standard procedures [40]. The powder mix- ρbulk
tures were then directly compressed at low (8 kN) and high ρtapped− ρbulk
(15 kN) compression forces into convex tablets with 10 mm CI ð%Þ ¼  100 ð3Þ
ρtapped
(diameter) concave punches using a single-punch tablet
The bulk density ρbulk was determined as the mass to vol-
Table 1 Formulation of IB-sustained release matrix tablets
ume ratio before compaction. The tapped density ρtapped was
Formula number Ingredient (%) determined by tapping the samples (500 taps) using a tapping
machine (ERWEKA, GmbH, Germany). The tests were per-
IB CD (30%) LM MGS formed according to the USP method [40].
F1 20 CMC 48 2
F2 30 CMC 38 2 Evaluation of matrix tablets
F3 40 CMC 28 2
F4 50 CMC 18 2 The weight uniformity (Electronic balance, Sartorius BP
F5 20 HEC 48 2 121S, Germany), diameter (Digital caliper, Mituyoto, Japan),
F6 30 HEC 38 2 friability (Erweka friabilator, Italy), hardness (Hardness tester
F7 40 HEC 28 2 Sotax HT1, USA), and disintegration time (Erweka ZT 220
F8 50 HEC 18 2 Series Manual Disintegration Tester, Germany) were evaluat-
F9 20 MC 48 2 ed according to the standard procedure [40]. The drug content
F10 30 MC 38 2 was determined spectrophotometrically (Thermospectronic
F11 40 MC 28 2 scientific Helios) at 263 nm. All tests were repeated three
F12 50 MC 18 2
times and the average values were used (see Table 4)
F13 20 HPMC 48 2
F14 30 HPMC 38 2 In vitro dissolution study
F15 40 HPMC 28 2
F16 50 HPMC 18 2 The drug release study was carried out in a paddle dissolution
F17 20 HPC 48 2 tester apparatus II (Dissolution tester DT 620, Erweka) accord-
F18 30 HPC 38 2 ing to the US Pharmacopeia [40]. For the dissolution medium,
F19 40 HPC 28 2 900 ml of 0.1 N HCl was used for the initial 2 h and phosphate
F20 50 HPC 18 2 buffer pH 6.8 was used for the remaining 10 h. Paddle speed
was adjusted to 75 rpm and the temperature of the medium was

maintained at 37 ± 0.5 °C. A sample volume of 5 ml was with-
drawn hourly and analyzed at 263 nm for IB content using a
UV-spectrophotometer. The withdrawn sample was immediate-
ly replaced by an equal volume of fresh buffer solution.
Drug release kinetics
To characterize the drug release behavior of the polymeric
systems and to understand the corresponding mechanism,
the obtained in vitro dissolution data was fitted into a mathe-
matical model. The Korsmeyer-Peppas model [42] was used
to analyze the kinetics and drug release mechanisms of the
dosage form according to Eq. (4):
Mt
¼ Kt n
M∞
where Mt/M∞is the fraction of drug released at time t, Kis the
constant incorporating the structural and geometric character-
istics of the matrix tablets, and n is the release exponent indic-
ative of the drug release mechanism.
To determine the extent and rate of drug release from the
various formulations, the time required for 10%, 50%, and
90% of the drug to be released (t10%,t50%, and t90%, respec-
tively) and the MDT of the formulations were calculated.
t10%,t50%, and t90%were obtained directly from the dissolution
time curves, while MDT was calculated using Eq. (5):
N f(x1, x2, x3).x1, x2, and x3 are the crisp functions in the conse-
∑ t mid  ΔM
i¼1
MDT ¼ N
∑ ΔM
i¼1
where i is the dissolution sample number, N is the dissolution
sample time,tmidis the time at the midpoint between i and i − 1,
and ΔM is the amount of drug dissolved between i and i − 1
[43]. MDT is a measure of the dissolution rate; the higher the
MDT, the slower the release rate.
Neuro-fuzzy modeling
The fuzzy logic was based on representing numbers with lin-
guistic variables to represent uncertainties. The relationship
between an uncertain quantityxand a membership function
μ, which ranges between 0 and 1, is represented by a fuzzy
value [44]. The inference rule is an Bif–then^ rule with the
general form Bif x is A then y is C^, whereAandCare the
linguistic values defined by fuzzy sets in the universes of
discourse of X and Y, respectively. This rule has two parts:
the ‘if’ part is called the antecedent and the ‘then’ part is called
the consequent of a rule.
Equation (6) gives the fuzzy membership function, which
defines the grade of membership ofxinA [44]. The form of this
function is mostly triangular, trapezoidal, bell shaped,
Drug Deliv. and Transl. Res.
Gaussian, or sigmoidal [39]. Figure 1 represents the model
composition of a fuzzy inference system [34].
A ¼ fx; μa ðxÞ=x∈X g ð6Þ
(1) Knowledge base unit: contains several fuzzy if–then
rules that define the membership functions of the fuzzy
sets used in the fuzzy rules.
(2) Decision making unit: performs the inference operations
on the rules.
(3) Fuzzification interface unit: transforms the crisp inputs
into degrees of match with linguistic values.
(4) Defuzzification interface unit: transforms the fuzzy re-
sults of the inference into a crisp output.
ð4Þ
Mamdani and Sugeno [45, 46] proposed two types of fuzzy
inference systems. The output membership functions of the
models are either constant or linear functions of the input
variables. In the Mamdani models, however, the output mem-
bership functions are fuzzy sets. Takagi–Sugeno (TS) is found
to be more suitable for adaptive modeling in combination with
ANNs Eq. (7) [47].
if x1 is A1 and if x2 is A2 and x3 is A3 then y
¼ f ðx 1 ; x2 ; x3 Þ ð7Þ
where A1,A2, andA3are the fuzzy sets in the antecedent fory =
quent. Usually, the function f is a polynomial of the input
ð5Þ variables x and y, but it can be any function that can appropri-
ately describe the output of the model within the fuzzy region
specified by the antecedent of the rule. When f(x, y)is a first-
order polynomial, the resulting fuzzy inference system is
called a first-order Sugeno fuzzy model (see Fig. 2), which
was originally proposed in references [45, 46]. When f is a
constant, a zero-order Sugeno fuzzy model is obtained [39].
ANFIS architecture
The ANFIS architecture is shown in Fig. 3, where the circular
nodes represent fixed operations and square nodes represent
functions with parameters to be learnt. ANFIS has five layers
where the function can be explained as follows [48].
Layer 1: in this layer where the fuzzification process takes
place, every node is adaptive. The outputs of this
layer form the membership values of the premise
part. The output of each node is given using Eq. 8.
O1;i ðxÞ ¼ μAi ðxÞ for i ¼ 1; 2
ð8Þ
O1;i ðyÞ ¼ μBi−2 ðyÞ for i ¼ 3; 4
Drug Deliv. and Transl. Res.

Fig. 1 Fuzzy inference system

Therefore,O1, i(x)and O1, i(y) are essentially the member- Layer 4: The nodes in this layer operate as a function block
ship grades forxandy, respectively. whose parameters are adaptive and variable. They are
The membership functions could be triangular, trapezoidal, the input values and the output of this layer forms the
or any other function, but for illustration purposes, we will use TSK outputs. The nodes in this layer are adaptive and
the bell-shaped function given using Eq. 9: perform the consequent of the rules (Eq. (12)):
1
μ A ð xÞ ¼  2bi ð9Þ
 x−ci 
1 þ  ai 
O4;i ¼ wi f i ¼ wi ðpi x þ qi y þ ri Þ for i ¼ 1; 2 ð12Þ
where pi, qi, and riare the parameters to be learnt. These are the
The parameters in this layer (pi, qi, ri) are to be determined
premise parameters.
and are referred to as the consequent parameters.
Layer 2: The nodes in this layer are fixed. Each node output
Layer 5: This layer sums all the incoming signals and pro-
represents a weight of the rule. Every node in this
duces the output. There is a single node here that
layer is fixed. This is where the t-norm is used to
computes the overall output (Eq. (13)):
BAND^ membership grades. The following prod-
uct is an example (Eq. (10)):

∑ wi f i
O2;i ¼ wi ¼ μAi ðxÞμBi ðyÞ for i ¼ 1; 2 ð10Þ O5 ¼ ∑ wi f i ¼ i
ð13Þ
i ∑ wi
i

For ANFIS learning, there are a number of possible ap-

Layer 3: The nodes are also fixed in this layer. The ratio of proaches, but we discuss the hybrid-learning algorithm pro-
the ith rule’s firing weight to the sum of all rules’ posed by Jang et al. [49], which uses a combination of back-
weights is computed for the corresponding node. It propagation and least squares estimation (LSE). It can be
contains fixed nodes that calculate the ratio of the shown that for the network described, if the premise parameters
firing strengths of the rules (Eq. (11)): are fixed, the output is linear in the consequent parameters. Two
sets of parameters are considered: the set of premise parameters
(S1) that contains the parameters for all the chosen input mem-
wi
O3;i ¼ wi ¼ for i ¼ 1; 2 ð11Þ bership functions and the set of consequent parameters (S2) that
w1 þ w2 contains the parameters of the output membership function.

Fig. 2 First-order Sugeno fuzzy

model

Fig. 3 ANFIS architecture
Consequently, ANFIS uses a two-pass learning algorithm.
The forward pass, where S1 is unmodified and S2 is computed
using an LSE algorithm, and the backward pass, where S2 is
unmodified and S1 is computed using a back-propagation
algorithm. Therefore, the hybrid learning algorithm uses a
combination of two algorithms to adapt the parameters in the
network [39]. The obtained ANFIS was validated with the
data remaining for the test’s purpose. The root mean square
error (RMSE) given by Eq. (14) was used as an evaluation
criterion for the model’s performance [37].
sffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffi
1 N
RMSE ¼ ∑ ðy −^ ^yÞ2 ð14Þ
N i¼1 i
where N is the number of the tested data, yi is the ith desired
output, and ^yi is the ith predicted output from the model.
ANFIS modeling setup
As illustrated in Fig. 4, the developed ANFIS has three inputs
that represent the operating parameters (CD nature, active in-
gredient amount, and the release time) and one output corre-
sponding to the release rate. Ninety percent of the experimen-
tal data values were used for ANFIS identification and the
remaining data were used to test the obtained ANFIS archi-
tecture. When using the MATLAB’s ANFIS toolbox, several
approaches were tested by making changes to the type and
number of input membership functions and the type of the
output membership function. The details of the best obtained
results in the learning phase are given in Table 2. Twenty sets
of parameters were tested with experimental data that were not
Fig. 4 ANFIS-developed
architecture
Drug Deliv. and Transl. Res.
used for the ANFIS architecture identification to validate the
obtained ANFIS model [39].
Results and discussion
Pre-compressional characteristics of powder mixtures
Twenty different formulation tablets were developed by using
CDs as matrix carriers. Direct compression was used, which
eliminates the granulation steps in the tablet formulation and
requires the use of excipients that have very good flow and
compaction properties. In addition, in the studied formulations
(see Table 3), the cellulose ethers (30%, w/w) were kept con-
stant and the effect of the IB concentration variation was mea-
sured. The CD ratio of 30% was chosen after the preliminary
experiments as this proportion was the most appropriate for
the aim of our investigation.
The Carr Compressibility Index and Hausner ratio are two
measures that can be used to predict the capacity of a given
powder sample to be compressed. They are also understood to
reflect the importance of interparticulate interactions. The
flowability of the powder is excellent if CI < 10, good if 11
< CI < 15, fair if 16 < CI <20, passable if 21 < CI < 25, and
poor if CI > 25. The Hausner ratio represents the inter-
particle friction state, and the closer the ratio to is 1, the better
the flowability. In general, the worst flowability is observed
when the ratio is greater than 1.25 [40, 50]. In the present
study, Carr’s index was found to range between 10.25 ± 0.25
and 19.73 ± 0.69. The Hausner’s ratio of all formulations
ranged from 1.09 ± 0.01 to 1.39 ± 0.02.
Drug Deliv. and Transl. Res.

Table 2 Best obtained results on

the learning phase Input Input membership Output membership Iterations RMSE R2
functions function

Type Number

Cellulose derivative Bell shape 4 Linear 570 5.7 × 10−3 0.9999

Active ingredient Bell shape 3
amount
Time Bell shape 4

The angle of repose has been defined as an indirect mea-
sure of particle size, shape, porosity, cohesion, fluidity, surface
area, and bulk. It provides an indication of the ease at which
the powder will flow. A small angle of repose (less than 40°)
indicates free-flowing powder, whereas large angles (more
than 50°) indicate poor flow properties [40, 51]. The angle
of repose ranged from 27.29 ± 0.45° to 37.96 ± 0.64°.
According to the obtained results, there was a decrease in
the angle of repose of the HPMC-based formulations com-
pared to the other ethers of cellulose used. These results indi-
cate that all formulations have a free-flowing powder with
‘good’ to ‘fair’ compressibility.
Post-compression evaluation
The matrix tablets were prepared on a single-punch tablet
machine and evaluated for post-compressional characteristics
according to the requirements of the pharmacopeia [40]. The
results of the physical and chemical characterization of the IB
tablets are given in Table 4. The tablets’ average weight was
between 300.12 ± 2.27 and 303.25 ± 3.25 mg. As the limit of
the percentage deviation was ± 7.5% of the average weight of
the tablets, all the randomly selected 20 tablets of the studied
formulations were within the deviation provided. In addition,
the diameter of the tablets ranged from 10.01 ± 0.04 to 11.41
± 0.01 mm. Tablet hardness was between 2.99 ± 1.93 and
7.68 ± 1.25 kg/cm2 and the friability was less than 1%, indi-
cating good mechanical integrity. This meant that the tablets
had sufficient strength and hardness and would remain intact
during drug release studies.
On the other hand, it was noticed that the hardness of the
tablets varied between formulations even though the CD con-
centrations were the same. This was mainly because of the
drug concentration, which was further proved with the de-
creased hardness found with an increase in the concentration
of IB. The formulations with MC had a very minimum

Table 3 Flow properties of

powder mixtures Formula number Carr’s index (%)a Hausner’s ratio a Angles of repose (°θ) a Flow properties

F1 11.86 ± 0.23 1.15 ± 0.02 30.25 ± 0.32 Good

F2 12.75 ± 0.56 1.16 ± 0.02 30.64 ± 0.31 Good
F3 14.26 ± 0.41 1.17 ± 0.03 31.12 ± 0.45 Good
F4 15.52 ± 0.62 1.18 ± 0.02 31.21 ± 0.54 Good
F5 17.73 ± 0.42 1.21 ± 0.03 32.84 ± 0.69 Fair
F6 18.18 ± 0.63 1.27 ± 0.01 33.95 ± 0.14 Fair
F7 18.86 ± 0.12 1.30 ± 0.02 34.59 ± 0.98 Fair
F8 19.75 ± 0.87 1.33 ± 0.02 35.84 ± 0.45 Fair
F9 18.76 ± 0.89 1.35 ± 0.01 35.42 ± 0.47 Fair
F10 19.12 ± 0.47 1.36 ± 0.04 36.75 ± 0.64 Fair
F11 19.52 ± 0.41 1.37 ± 0.02 37.81 ± 0.75 Fair
F12 19.73 ± 0.69 1.39 ± 0.02 37.96 ± 0.64 Fair
F13 10.25 ± 0.25 1.09 ± 0.01 27.29 ± 0.45 Good
F14 10.82 ± 0.52 1.11 ± 0.03 27.45 ± 0.54 Good
F15 11.73 ± 0.64 1.15 ± 0.04 29.54 ± 0.65 Good
F16 12.28 ± 0.68 1.16 ± 0.01 30.50 ± 0.75 Good
F17 15.26 ± 0.78 1.16 ± 0.04 30.23 ± 0.88 Fair
F18 15.51 ± 0.24 1.17 ± 0.04 30.65 ± 0.45 Fair
F19 16.40 ± 0.58 1.18 ± 0.02 31.56 ± 0.63 Fair
F20 17.92 ± 0.47 1.19 ± 0.03 32.15 ± 0.87 Fair
a
All results represent mean ± standard deviation, n = 3
 Drug Deliv. and Transl. Res.

Table 4 Evaluation of post-compressional properties of IB matrix tablet

Formulations Average weight (mg)a Hardness (kg/cm2)b Friability (%)b Diameter (mm)a Disintegration time (h)c Drug content (%)b

F1 301.65 ± 2.47 6.21 ± 2.25 0.58 ± 0.02 10.18 ± 0.02 11.70 ± 0.25 100.06 ± 0.24
F2 299.42 ± 3.25 6.17 ± 2.87 0.51 ± 0.03 10.21 ± 0.04 11.22 ± 0.32 101.41 ± 0.14
F3 301.25 ± 2.45 5.82 ± 2.05 0.42 ± 0.01 10.25 ± 0.01 10.95 ± 0.52 100.47 ± 0.21
F4 300.12 ± 2.27 5.27 ± 1.45 0.35 ± 0.01 10.04 ± 0.04 10.23 ± 0.41 100.01 ± 0.11
F5 301.58 ± 3.01 4.18 ± 2.01 0.28 ± 0.03 10.02 ± 0.02 7.95 ± 0.32 100.08 ± 0.17
F6 301.50 ± 2.87 4.02 ± 1.47 0.25 ± 0.02 10.15 ± 0.01 7.40 ± 0.24 100.17 ± 0.12
F7 299.52 ± 2.15 3.98 ± 1.85 0.22 ± 0.02 10.23 ± 0.01 7.22 ± 0.31 100.21 ± 0.32
F8 300.13 ± 2.22 3.81 ± 2.16 0.20 ± 0.01 11.09 ± 0.05 7.12 ± 0.23 100.07 ± 0.22
F9 301.58 ± 1.09 3.49 ± 1.75 0.20 ± 0.04 10.51 ± 0.02 7.65 ± 0.17 100.14 ± 0.10
F10 302.15 ± 2.28 3.20 ± 2.60 0.16 ± 0.03 11.41 ± 0.04 7.41 ± 0.20 99.99 ± 0.12
F11 303.01 ± 2.27 3.12 ± 2.94 0.12 ± 0.02 10.04 ± 0.02 6.57 ± 0.34 100.12 ± 0.41
F12 300.35 ± 2.61 2.99 ± 1.93 0.11 ± 0.01 10.42 ± 0.02 6.54 ± 0.42 100.22 ± 0.27
F13 301.22 ± 1.51 7.56 ± 1.25 0.85 ± 0.02 11.25 ± 0.02 15.55 ± 0.12 101.00 ± 0.14
F14 302.85 ± 2.76 7.21 ± 1.87 0.79 ± 0.03 10.35 ± 0.04 14.25 ± 0.18 100.01 ± 0.14
F15 301.15 ± 2.27 7.08 ± 1.45 0.70 ± 0.01 11.41 ± 0.01 13.90 ± 0.27 101.02 ± 0.10
F16 300.58 ± 3.01 6.98 ± 2.45 0.69 ± 0.01 10.74 ± 0.03 13.40 ± 0.22 99.97 ± 0.57
F17 302.55 ± 2.25 5.12 ± 1.01 0.42 ± 0.03 10.01 ± 0.04 8.92 ± 0.27 101.07 ± 0.25
F18 302.12 ± 3.20 5.01 ± 2.27 0.38 ± 0.02 10.12 ± 0.02 8.62 ± 0.22 101.14 ± 0.14
F19 301.18 ± 3.15 4.80 ± 2.15 0.34 ± 0.02 10.41 ± 0.01 8.24 ± 0.14 99.97 ± 0.21
F20 300.58 ± 2.12 4.74 ± 1.86 0.31 ± 0.01 10.09 ± 0.04 7.84 ± 0.21 100.16 ± 0.47

All results represent mean ± standard deviation

a
n = 20
b
n = 10
c
n=6

hardness than the formulations with HPMC and CMC, which
may have been due to the fibrous nature of the materials. The
above results are in accordance with the pharmacopeial spec-
ification standards [40].
Drug content and disintegration time
The drug content of all formulations (F1-F20) was within 90–
110%, indicating content uniformity (see Table 4). The disin-
tegration test was carried out according to the general proce-
dure by placing six tablets in the six baskets of the disintegra-
tion apparatus, which contained a pH 1.2 acidic buffer, for 2 h
at 37 ± 2 °C. According to the standards [40], sustained re-
lease tablets should not disintegrate for 2 h in the tested me-
dium (acidic buffer, pH 1.2). All the tablets remained intact
beyond 2 h. Moreover, the disintegration time value ranged
from 6.54 ± 0.42 to 15.55 ± 0.12 h (see Table 4). The MC-
based tablets had the lowest disintegration times, and it was
clearly observed that there was a significant difference in the
disintegration time with an increase in the concentration of the
drugs. It was observed from the results that the increase in the
concentration of IB decreased the disintegration time.
Therefore, the HPMC and CMC, which were considered to
retard the drug release, could also increase the hardness and
increase the disintegration time [52].
In vitro dissolution studies
The main goal of a sustained release tablet is the release of the
amount of drug necessary to maintain an effective drug plas-
ma concentration [53–55]. To achieve this effect, the tablet
should be formulated to release the drug with prolonged ki-
netics. All the compressed tablets (F1-F20) were studied for
12 h (see Fig. 7), and their dissolution profiles were analyzed
to study the effect of varying concentrations of IB and the
effect CD polymers on the kinetics of IB release. A general
observation made during the dissolution testing of the matrix
tablets was the surface hydration of the matrix that resulted in
its swelling and the consequent formation of a gel layer [54].
In fact, the hydration progressed from the surface to the core
of the tablet over time, but the hydration rate was different for
CD-based matrices. Up on sufficient hydration, the gel layer
slowly dissolved and eroded away, exposing a new gel layer.
This mechanism is commonly observed with swellable
sustained release tablets [10, 56, 57].
Drug Deliv. and Transl. Res.
Release kinetics
The use of different CDs as tablet matrices contributed to
obtaining different kinetic releases of IB from the tablet matrix.
It can be seen that the sustained drug release from the tablets
was maintained and was close to 100% in all formulations at
times ranging from 7 to 12 h. As indicated in Fig. 5, the tablets
containing MC (F9, F10, F11, and F12) and HEC (F5, F6, F7,
and F8) showed an initial burst release during the first hour that
ranged from 23.46 ± 0.81 to 24.96 ± 1.24% for the HEC-based
formulations and between 28.12 ± 1.36 and 33.67 ± 1.5% for
the MC-based formulations. This behavior may have been re-
lated to surface erosion with early disintegration of the matrix
tablets before they formed the gel layer [52]. That being said, no
burst release effect was observed in the HPC, CMC, and
HPMC-based formulations. Indeed, less than 20% of the drug
(a)
(c)
Fig. 5 Release profiles of the IB derivative cellulose-based tablets with different active ingredient amount: a 20%, b 30%, c 40%, d 50%
was released in 1 h for HPC, CMC, and HPMC: 19.77% ±
1.32%, 12.76 ± 2.41%, and 7.99 ± 1.85%, respectively.
Figure 6 shows the effect of the loaded drug on the time
required for 10%, 50%, and 90% drug release for each formu-
lation. The above results indicated that for all tablet matrices,
increasing the concentration of IB in the matrix led to faster
drug release. The above results were confirmed by the t10%,
t50%, and t90% values (time required for 10%, 50%, and 90%
drug release, respectively), which increase when the loaded
drug in the different matrix compositions increases for all
formulations (see Fig. 6). This observation was in accordance
with previous studies that have underscored the importance of
drug models and their concentrations on the release of the
drug from the matrix [56, 58–60].
It has been reported that poorly soluble drugs like IB,
which is practically water-insoluble with a solubility value
(b)
(d)
 Drug Deliv. and Transl. Res.

Fig. 6 Effect of the formulation 18

t10% CMC t10% HEC t10% MC t10% HPMC t10% HPC
composition on the 10%, 50%, 16
t50% CMC t50% HEC t50% MC t50% HPMC t50% HPC
and 90% drug release
14 t90% CMC t90% HEC t90% MC t90% HPMC t90% HPC

Release Time (hours)

12
10
8
6
4
2
0
IB 20% IB 30% IB 40% IB 50%

of 0.06 mg/mL, are mainly released by erosion, while water-
soluble drugs are released by a diffusion mechanism [61–63].
Consequently, the fastest drug release was found in the tablets
most loaded with IB for all CD-based formulations. The re-
lease rate decreased with a decrease in IB concentration load-
ed into the CD matrices [64] (Fig. 7).
Moreover, drug release from polymer matrices is influ-
enced by the molecular size of the polymer and various studies
have reported a decline in the rate of drug release with an
increase in the polymer viscosity and its molecular weight
[59, 65–67]. In turn, IB release from HPC and HEC was not
as immediate as with MC. From the release profiles of these
matrices, it is seen that 50% of the drug was released between
2.68 and 3.90 h, which was the lowest release compared with
the MC matrices. This could be attributed to the dissolution of
the drug mainly located on the surface of the matrix and by the
fact that early on in the dissolution (when the pH of the me-
dium was in the region of 1.2), the gel thickness was rather
low, resulting in a limited diffusion pathway for the drug mol-
ecules to diffuse out of. Different authors have reported iden-
tical data for other drugs [3, 68–70].
For cellulose ethers, the degree of substitution plays an in-
fluential role in its water solubility. MC, HEC, and HPC are
characterized by lower hydrophilicity compared with CMC and
HPMC. This difference in hydrophilicity explains the lower
rates of the absorption of water by their matrices, which conse-
quently led to the rapid release in the first hours of the tests
compared to CMC and HPMC [71–73]. From these observa-
tions, F1, F2, F3, and F4 (based on CMC matrices) achieved a
drug release of more than 97.49% within 12 h, whereas F13,
F14, F15, and F16 (based on HPMC matrices) were only able
to achieve a maximum of 88.95% drug release within 12 h. The
dissolution profiles for the latter periods of these matrices
showed significantly slower rates of release that were probably
due to an increase in the length of the diffusional path of the
drug over time and the stabilization of the gel barrier.
From the release profiles, it is evident that the release rate of
IB was greater in the CMC tablets compared with the HPMC
tablets, which exhibited the lowest release. The delay in release
in the F1, F2, F3, and F4 formulations based on CMC was
important because of the binding property and the strength of
the gelatinous layer formed around the molecule and the coat-
ing property of CMC over the particle surface. CMC is a poly-
electrolyte ionic CD that is sensitive to changes in pH [74]
related to hydration and its chains untangle to form a viscous
gel layer on the surface. Erosion of the gel has been found to be
one of the main mechanisms by which this polymer releases
drugs [75–77]. The presence of anionic polymers, such as
CMC, had a beneficial effect on the viscosity and gave an
almost linear release of IB over a 12-h period. As the carboxyl

Fig. 7 a 3D presentation given

the maximum release rate of IB as
a function of the polymer nature
and the active ingredient
concentration. b Comparison
between the effects of the active
ingredient amount on the different
CD release rates

(a) (b)
Drug Deliv. and Transl. Res.
groups composed of CMC molecules are anionic, they have a
tendency to swell slowly in a pKa value of ≈ 4.3. This results in
an increase in the gel viscosity that explains the better control
these polymers have on the release of the drug [65, 78, 79].
The drug release from the HPMC tablets (F13, F14, F15,
and F16) was not related to the Mw of the CD. During the drug
release testing, the surfaces of the HPMC tablets cracked slight-
ly when exposed to the acidic medium, causing faster water
uptake and swelling of the tablets. The increased swelling of
the HPMC matrices led to a longer diffusion path for the drug
molecules, which resulted in a slower drug release from the
HPMC tablets [23, 80]. The hydration rate of this cellulosic
polymer is related to its hydroxypropyl substitute percentage.
HPMC-K4M, which was used in this study, contains the
greatest amount of these groups and produces a strongly
(a)
(c)
Fig. 8 Relative errors between experimental and predicted data: a 20%, b 30%, c 40%, d 50%
viscose gel that plays an important role in the drug release in
addition to its pH-independent solubility. Although gel strength
is controlled by polymer viscosity and concentration, polymer
chemistry also plays a significant role. The obtained results
suggest that the chemistry of HPMC leads to a stronger gel
compared to other cellulosics. As a result, the drug release rates
were longer with HPMC than with equivalent levels of MC,
HEC, HPC, and CMC. These properties make it suitable for use
as an excipient in sustained release formulations [1, 81–85].
ANFIS modeling
The obtained experimental data were divided into two sepa-
rate sets: the training data set and the checking data set. The
training data set was used to train the ANFIS, whereas the
(b)
(d)
 Drug Deliv. and Transl. Res.

checking data set was used to verify the accuracy and the fact, the ANFIS not only predicted the kinetics release but also
effectiveness of the trained ANFIS model for the adaptation the effect of other parameters, which were the active ingredi-
of learning content. The optimal ANFIS model was chosen ent concentration and the polymers’ nature. In addition, a
based on the number of membership functions assigned to squared coefficient (R2) close to 1.0 (0.9999) indicated that
each structure and the RMSE values for different epoch num- the optimal ANFIS model was reached.
bers. Therefore, for the training of the ANFIS system, four To validate the obtained ANFIS model, it was also tested
aspects were overfitting, number of membership functions, by using a new set of data that has never been used in the
type of membership functions, and training options. The learning phase modeling. The new set of data belonged to
MATLAB 7.10.1. software was used to calculate the predicted 20 different formulations with variables and random inputs.
data for each experiment. The inputs used for the ANFIS Table 5 gives the prediction results and the corresponding
training were the three independent variables: the polymer calculated errors. As a result, a good similarity between the
nature, the active ingredient concentration, and the release experimental and the predicted sets of data were obtained with
time. These three variables have a direct effect on the IB re- an RMSE value of 5.7 × 10−3. This result shows the capability
lease rate, which constitutes the output variable. of the synthesized ANFIS model of predicting the drug release
Once the system finished the learning phase, we correlated profile from the formulation. This model can predict the be-
the experimental and predicted values. Consequently, the havior during the release experiment, which saves time and
ANFIS model was evaluated based on its performance for the materials.
training and checking sets (see Fig. 8). A comparison between
the ANFIS’s model prediction and the experiment data indicat- Mechanism of drug release (Korsmeyer-Peppas model)
ed the good performance of the former. Moreover, the perfor-
mance of the ANFIS model was evaluated using relative error The drug release mechanism of matrix tablets is complex be-
measurements, which revealed the optimal setting necessary for cause although some processes can be classified as diffusion
better predictability. Errors were calculated for each single run or erosion controlled, many others can be controlled by both at
and the results given in Fig. 8 show that the ANFIS model is, in the same time [34]. A large number of mathematical models
general, accurate and acceptable. have been developed to describe the drug release profiles of
Correlation plots were constructed for the predicted versus
experimental values of drug release for all trial formulations. Table 5 Comparison testing between experimental and predicted
The point cloud observed near the bisector in Fig. 9 shows the values
narrowness between the experimental values influenced by Time (h) Polymer IB Experimental ANFIS Relative
several parameters (concentration of the active ingredient, amount (%) predicted error (%)
CD, and time) and those predicted by the ANFIS model. In
4 1 50 42.63 42.73 − 0.10
5 1 20 43.95 44.07 − 0.12
6 1 30 54.07 54.14 − 0.07
8 1 40 72.19 71.70 0.49
1 2 30 24.12 23.90 0.22
2 2 50 40.85 40.88 − 0.03
7 2 40 98.24 97.79 0.45
11 2 20 100.00 99.61 0.39
1 3 40 31.88 31.72 0.15
3 3 30 57.98 57.77 0.21
6 3 20 85.41 85.58 − 0.17
6 3 50 95.18 94.63 0.54
5 4 50 38.05 37.85 0.19
6 4 20 37.75 37.82 − 0.07
6 4 30 43.81 43.94 − 0.13
8 4 40 58.94 58.25 0.69
1 5 30 58.94 58.25 0.69
5 5 20 62.87 62.79 0.08
5 5 40 66.32 66.13 0.19
8 5 50 96.03 95.68 0.35
Fig. 9 Point cloud representing the ANFIS predicted release amount as a 4 1 50 42.63 42.73 − 0.10
function of the experimental data
Drug Deliv. and Transl. Res.

Table 6 Korsmeyer-Peppas model parameters and 0.89, this indicates a superposition of both phenomena
Formula Korsmeyer-Peppas model Drug release (anomalous transport) [1, 88].
number mechanism The corresponding drug dissolution data of all formulations
K (% h−n) n R2 (F1-F20) were fitted for the Korsmeyer-Peppas equation and
the results were evaluated with the determination coefficient
F1 10.160 0.91 0.998 Case-II transport
(R2) values. For the overall results, the R2 values were over the
F2 10.760 0.90 0.998 Case-II transport
range of 0.990–0.998. The obtained data (see Table 6) indi-
F3 11.509 0.88 0.998 Anomalous transport
cated that the drug release of the tablets fitted well into the
F4 12.685 0.87 0.998 Anomalous transport
studied model and confirmed that the drug release mechanism
F5 22.678 0.71 0.991 Anomalous transport
was not controlled by the diffusion process, but by matrix
F6 23.978 0.70 0.998 Anomalous transport swelling and matrix erosion.
F7 25.126 0.68 0.997 Anomalous transport Different values for the exponent were obtained corre-
F8 25.841 0.67 0.996 Anomalous transport sponding to the dominating drug release mechanism. The re-
F9 27.981 0.62 0.998 Anomalous transport lease exponent Bn^ value is greater than 0.5, which indicates a
F10 29.289 0.61 0.998 Anomalous transport coupling of the diffusion and erosion mechanism, namely
F11 31.672 0.60 0.998 Anomalous transport anomalous diffusion. These results indicate that the drug re-
F12 33.315 0.58 0.998 Anomalous transport lease is controlled by more than one process in all cases of
F13 6.179 1.01 0.998 Case-II transport tablet formulations studied.
F14 7.100 0.99 0.990 Case-II transport The ANOVA-based comparisons were carried out with the
F15 7.513 0.98 0.998 Case-II transport percentage of IB dissolved at tx% corresponding to the time
F16 7.986 0.97 0.998 Case-II transport necessary to release a determined percentage of drug, using a
F17 16.610 0.81 0.998 Anomalous transport univariate one-way ANOVA followed by a Dunnett’s multiple
F18 18.069 0.79 0.998 Anomalous transport comparison test, where the differences were considered sig-
F19 18.680 0.78 0.998 Anomalous transport nificant if p < 0.05. Dissolution profiles were compared with
F20 19.541 0.76 0.998 Anomalous transport the (t10%, t50%, and t90%) values and significant differences
were found (p < 0.05) at time points, between the CD formu-
lations [31, 89, 90].
matrices [1, 32, 86, 87]. In fact, the simple and more widely By comparing the value of the exponent n ranging from
used model is the one derived by Korsmeyer-Peppas [42, 86], 0.58 to 1.01 for the different types of used CDs in tablet
which has been used to evaluate the drug release of polymeric formulations (corresponding to F12 and F13, respectively),
matrices. This model is frequently used when the drug release it is clear that the lower the value of the exponent n, the faster
mechanism is unknown or when there is more than one release the drug release kinetics. Indeed, the release time of 10%
mechanism involved. Consequently, the analysis of experi- (0.126 h for F12 and 1.611 h for F13), 50% (2.014 h for F12
mental data according to Korsmeyer-Peppas and the interpre- and 7.926 h for F13), and 90% (5.548 h for F12 and 14.183 h
tation of the corresponding release exponent values (n) leads for F13) of IB indicates a slower release of the active ingredi-
to a better understanding of the balance between these mech- ent for values of n approaching 1.0.
anisms. For the tablets, when n is less than 0.45, the drug Mean dissolution time (MDT) value is used to characterize
release is controlled by diffusion, and when n is greater than drug release rate from a dosage form and indicates the drug
0.89, the drug release is controlled by the matrix erosion (case release retarding efficiency of polymer. According to Fig. 10,
II transport). Therefore, as the values of n were between 0.45 tablets prepared with HPMC and CMC showed higher MDT

Fig. 10 Influence of CD’s nature

used in the formulation on the
disintegration time of tablets

value, ranging between 10.23 and 15.55 h, in comparison to
tablets prepared with MC, HEC, and HPC (ranged between
6.54 and 8.92 h). These results can be attributed to the water
diffusion property into the core of tablets, which retarded dif-
ferently the drug release from the matrices [83].
Conclusion
In this study, various types of CD have been used in formula-
tion of hydrophilic polymeric matrices such as HPMC, CMC,
HPC, HEC, and MC. Indeed, tablet dosage forms of IB have
been successfully formulated. These cellulose ethers can be
used as polymeric matrices due to their hydrophilic nature and
ability to form gel in aqueous media. The highest release rate
among studied cellulose ethers corresponds to HEC matrices,
but the most sustained release is ensured by HPMC due to its
excellent swelling properties, good compressibility, and fast
hydration in contact with water.
A neuro-fuzzy architecture was successfully used to estimate
the percentage release profile where the values of R2 and the
mean square error were 0.9999 and 5.7 × 10−3, respectively.
Consequently, the new architecture gives better modelization
than the Korsmeyer-Peppas model. Furthermore, it enhances
and gives a better understanding of the drug system delivery.
This work is demonstrative and can be extended to use neuro-
fuzzy models for predicting the drug release profile for a variety
of active ingredients from different formulation types.
Compliance with ethical standards
Conflicts of interest The authors declare that they have no conflict of
interest.
References
1. Siepmann J, Peppas NA. Modeling of drug release from delivery
systems based on hydroxypropyl methylcellulose (HPMC). Adv
Drug Deliv Rev. 2001;48(2–3):139–57.
2. Wen H, Park K. Oral controlled release formulation design and drug
delivery: theory to practice. 1st ed. Hoboken: Wiley; 2010.
3. Khlibsuwan R, Pongjanyakul T. Chitosan-clay matrix tablets for
sustained-release drug delivery: effect of chitosan molecular weight
and lubricant. J Drug Deliv Sci Technol. 2016;35:303–13.
4. Shoaib M, et al. Sustained drug delivery of doxorubicin as a func-
tion of pH, releasing media, and NCO contents in polyurethane urea
elastomers. J Drug Deliv Sci Technol. 2017;39:277–82.
5. Krishnaiah YSR. Pharmaceutical Technologies for Enhancing Oral
Bioavailability of poorly soluble drugs. J Bioequivalence
Bioavailab. 2010;2:2.
6. Savjani KT, Gajjar AK, Savjani JK. Drug solubility: importance
and enhancement techniques. ISRN Pharm. 2012;2012:195727.
7. Abbaspour MR, Sadeghi F, Afrasiabi Garekani H. Design and study
of ibuprofen disintegrating sustained-release tablets comprising
coated pellets. Eur J Pharm Biopharm. 2008;68(3):747–59.
Drug Deliv. and Transl. Res.
8. Fredenberg S, Wahlgren M, Reslow M, Axelsson A. The mecha-
nisms of drug release in poly(lactic-co-glycolic acid)-based drug
delivery systems–a review. Int J Pharm. 2011;415(1–2):34–52.
9. Mohamed FAA, Roberts M, Seton L, Ford JL, Levina M, Rajabi-
Siahboomi AR. Film-coated matrix mini-tablets for the extended
release of a water-soluble drug. Drug Dev Ind Pharm. 2015;41(4):
623–30.
10. Lal N, Dubey J, Gaur P, Verma N, Verma A. Chitosan based in situ
forming polyelectrolyte complexes: a potential sustained drug de-
livery polymeric carrier for high dose drugs. Mater Sci Eng C.
2017;79:491–8.
11. Ferreira TR, Lopes LC, Ferreira TR, Lopes LC. Analysis of anal-
gesic, antipyretic, and nonsteroidal anti-inflammatory drug use in
pediatric prescriptions. J Pediatr. 2016;92(1):81–7.
12. Guerra-Ponce WL, et al. In vitro evaluation of sustained released
matrix tablets containing ibuprofen: a model poorly water-soluble
drug. Braz J Pharm Sci. 2016;52(4):751–9.
13. Brayfield A, editor. Martindale: the complete drug reference, 38th
Revised edition. London: Pharmaceutical Press; 2014.
14. Schultz P, Tho I, Kleinebudde P. A new multiparticulate delayed
release system. Part II : Coating formulation and properties of free
films. J Control Release. 1997;47(2):191–9.
15. Siepmann F, Hoffmann A, Leclercq B, Carlin B, Siepmann J. How
to adjust desired drug release patterns from ethylcellulose-coated
dosage forms. J Control Release. 2007;119(2):182–9.
16. Muschert S, Siepmann F, Leclercq B, Carlin B, Siepmann J. Drug
release mechanisms from ethylcellulose: PVA-PEG graft
copolymer-coated pellets. Eur J Pharm Biopharm. 2009;72(1):
130–7.
17. Gaber DM, Nafee N, Abdallah OY. Mini-tablets versus pellets as
promising multiparticulate modified release delivery systems for
highly soluble drugs. Int J Pharm. 2015;488(1–2):86–94.
18. Yang Q, Ma Y, Zhu J. Sustained drug release from electrostatic
powder coated tablets with ultrafine ethylcellulose powders. Adv
Powder Technol. 2016;27(5):2145–52.
19. Siepmann J, Kranz H, Peppas NA, Bodmeier R. Calculation of the
required size and shape of hydroxypropyl methylcellulose matrices
to achieve desired drug release profiles. Int J Pharm. 2000;201(2):
151–64.
20. Qiu Y, Zhang G. Chapter 21 - development of modified-release
solid Oral dosage forms. In: Developing solid Oral dosage forms.
San Diego: Academic Press; 2009. p. 501–17.
21. Ratnaparkhi MP, Gupta JP. Sustained release oral drug delivery
system-an overview. Int J Pharma Res Rev. 2013;2:11–21.
22. Priyadarshini R, Nandi G, Changder A, Chowdhury S, Chakraborty
S, Ghosh LK. Gastroretentive extended release of metformin from
methacrylamide-g-gellan and tamarind seed gum composite matrix.
Carbohydr Polym. 2016;137:100–10.
23. Maderuelo C, Zarzuelo A, Lanao JM. Critical factors in the release
of drugs from sustained release hydrophilic matrices. J Control
Release. 2011;154(1):2–19.
24. Wen X, Nokhodchi A, Rajabi-Siahboomi A. Oral extended release
hydrophilic matrices: formulation and design. In: Wen H, Park K,
editors. Oral controlled release formulation design and drug deliv-
ery. New Jersey: John Wiley & Sons, Inc.; 2010. p. 89–100.
25. Zhou X, et al. Hydroxyethyl Pachyman as a novel excipient for
sustained-release matrix tablets. Carbohydr Polym. 2016;154:1–7.
26. Ghori MU, Ginting G, Smith AM, Conway BR. Simultaneous
quantification of drug release and erosion from hypromellose hy-
drophilic matrices. Int J Pharm. 2014;465(1):405–12.
27. Sood S, Gupta VK, Agarwal S, Dev K, Pathania D. Controlled
release of antibiotic amoxicillin drug using carboxymethyl cellu-
lose-cl-poly(lactic acid-co-itaconic acid) hydrogel. Int J Biol
Macromol. 2017;101:612–20.
Drug Deliv. and Transl. Res.
28. Ghori MU, Šupuk E, Conway BR. Tribo-electrification and powder
adhesion studies in the development of polymeric hydrophilic drug
matrices. Materials. 2015;8(4):1482–98.
29. Rajabi-Siahboomi AR, Nokhodchi A, Rubinstein MH. Compaction
behaviour of hydrophilic cellulose ether polymers. Pharm Technol.
1998;22(10):32–40.
30. Nokhodchi A, Ford JL, Rowe PH, Rubinstein MH. The effects of
compression rate and force on the compaction properties of differ-
ent viscosity grades of hydroxypropylmethylcellulose 2208. Int J
Pharm. 1996;129(1):21–31.
31. Costa P, Sousa Lobo JM. Modeling and comparison of dissolution
profiles. Eur J Pharm Sci. 2001;13(2):123–33.
32. Siepmann J, Siepmann F. Mathematical modeling of drug delivery.
Int J Pharm. 2008;364(2):328–43.
33. Dash S, Murthy PN, Nath L, Chowdhury P. Kinetic modeling on
drug release from controlled drug delivery systems. Acta Pol
Pharm. 2010;67(3):217–23.
34. Salome Amarachi C, Onunkwo G, Onyishi I. Kinetics and mecha-
nisms of drug release from swellable and non swellable matrices: a
review. Res J Pharm, Biol Chem Sci. 2013;4:97–103.
35. Manga RD, Jha PK. Mathematical models for controlled drug re-
lease through pH-responsive polymeric hydrogels. J Pharm Sci.
2017;106(2):629–38.
36. Rebouh S, Bouhedda M, Hanini S, Djellal A. ‘Neural modeling
adsorption of copper, chromium, nickel, and Lead from aqueous
solution by natural wastes’, in Progress in clean energy, Volume 1.
Cham: Springer; 2015. p. 341–56.
37. Yoneyama J, Nishikawa M, Katayama H, Ichikawa e A. Design of
output feedback controllers for Takagi–Sugeno fuzzy systems.
Fuzzy Sets Syst. 2001;121(1):127–48.
38. M. Chadli and P. Borne, Multiple models approach in automation:
Takagi-Sugeno fuzzy systems. Wiley-ISTE, 2012.
39. Rebouh S, Bouhedda M, Hanini S. Neuro-fuzzy modeling of cu(II)
and Cr(VI) adsorption from aqueous solution by wheat straw.
Desalin Water Treat. 2016;57(14):6515–30.
40. U. S. P. Convention. U.S. Pharmacopeia National Formulary 2016:
USP 39 NF 34, Supplement edition. Rockville: United States
Pharmacopeial; 2016.
41. Kaleemullah M, et al. Development and evaluation of Ketoprofen
sustained release matrix tablet using Hibiscus rosa-sinensis leaves
mucilage. Saudi Pharm J. 2017;25(5):770–9.
42. Korsmeyer RW, Gurny R, Doelker E, Buri P, Peppas NA.
Mechanisms of solute release from porous hydrophilic polymers.
Int J Pharm. 1983;15(1):25–35.
43. Gohel MC, Panchal MK. Novel use of similarity factors f2 and Sd
for the development of diltiazem HCl modified-release tablets using
a 3(2) factorial design. Drug Dev Ind Pharm. 2002;28(1):77–87.
44. Zadeh LA. Fuzzy sets. Inf Control. 1965;8(3):338–53.
45. Mamdani EH, Assilian S. An experiment in linguistic synthesis
with a fuzzy logic controller. Int J Man Mach Stud. 1975;7(1):1–13.
46. Takagi T, Sugeno M. Fuzzy identification of systems and its appli-
cations to modeling and control. IEEE Trans Syst Man Cybern.
1985;SMC-15(1):116–32.
47. Huang Z, Hahn J. Fuzzy modeling of signal transduction networks.
Chem Eng Sci. 2009;64(9):2044–56.
48. Patel J, Gianchandani R. ANFIS control for robotic manipulators:
adaptive neuro fuzzy inference Systems for Intelligent Control.
Saarbrucken: LAP LAMBERT Academic Publishing; 2011.
49. Jang J-SR, Sun C-T, Mizutani E. Neuro-fuzzy and soft computing:
a computational approach to learning and machine intelligence, 1
edition. Upper Saddle River: Pearson; 1997.
50. Suñé Neģre JM et al. ‘SeDeM diagram: an expert system for pre-
formation, characterization and optimization of tablets obtained by
direct compression’. In: Aguilar JE, editor. Formulation tools for
pharmaceutical development. Cambridge: WoodHead Publishing;
2013. pp. 109–135.
51. Lefnaoui S, Moulai-Mostefa N. Synthesis and evaluation of the
structural and physicochemical properties of carboxymethyl
pregelatinized starch as a pharmaceutical excipient. Saudi Pharm
J. 2015;23(6):698–711.
52. Shergill M, Patel M, Khan S, Bashir A, McConville C.
Development and characterisation of sustained release solid disper-
sion oral tablets containing the poorly water soluble drug disulfi-
ram. Int J Pharm. 2016;497(1):3–11.
53. Porfire A, Filip C, Tomuta I. High-throughput NIR-chemometric
methods for chemical and pharmaceutical characterization of
sustained release tablets. J Pharm Biomed Anal. 2017;138:1–13.
54. Saurí J, et al. Quality by design approach to understand the physi-
cochemical phenomena involved in controlled release of captopril
SR matrix tablets. Int J Pharm. 2014;477(1–2):431–41.
55. Wang X, Yan H. Methotrexate-loaded porous polymeric adsorbents
as oral sustained release formulations. Mater Sci Eng C. 2017;78:
598–602.
56. Vijayabhaskar K, Venkateswarlu K, Naik SBT, Jyothi RK, Vani
GN, Chandrasekhar KB. Preparation and in-vitro evaluation of ra-
nitidine Mucoadhesive microspheres for prolonged gastric reten-
tion. Future J Pharm Sci. 2016;10(2):1–12.
57. Agrawal P. Significance of polymers in drug delivery system. J
Pharmacovigil. 2014;1:3.
58. Ghori MU, Conway BR. Hydrophilic matrices for Oral control drug
delivery. Am J Pharmacol Sci. 2015;3(5):103–9.
59. Ravi PR, Kotreka UK, Saha RN. Controlled release matrix tablets
of zidovudine: effect of formulation variables on the in vitro drug
release kinetics. AAPS PharmSciTech. 2008;9(1):302–13.
60. Reynolds TD, Mitchell SA, Balwinski KM. Investigation of the
effect of tablet surface area/volume on drug release from
hydroxypropylmethylcellulose controlled-release matrix tablets.
Drug Dev Ind Pharm. 2002;28(4):457–66.
61. Chaudhari SP, Dugar RP. Application of surfactants in solid disper-
sion technology for improving solubility of poorly water soluble
drugs. J Drug Deliv Sci Technol. 2017;41:68–77.
62. Ghori MU, Supuk E, Conway BR. Tribo-electric charging and ad-
hesion of cellulose ethers and their mixtures with flurbiprofen. Eur J
Pharm Sci. 2014;65:1–8.
63. Tiwari SB, Murthy TK, Pai MR, Mehta PR, Chowdary PB.
Controlled release formulation of tramadol hydrochloride using
hy d r o p h i l i c a n d h y d r o p h o b i c ma t r i x s y s t e m . A A PS
PharmSciTech. 2003;4(3):E31.
64. Khan J, et al. Comparative study on the effect of hydrophilic and
hydrophobic polymers on the dissolution rate of a poorly water
soluble drug. Int J Pharm Anal Res. 2014;3:291–300.
65. Kambham V, Kothapalli Bonnoth C. Development of stavudine
sustained release tablets: in-vitro studies. Future J Pharm Sci.
2016;2(2):37–42.
66. Gao P, Skoug JW, Nixon PR, Ju TR, Stemm NL, Sung KC.
Swelling of hydroxypropyl methylcellulose matrix tablets. 2.
Mechanistic study of the influence of formulation variables on ma-
trix performance and drug release. J Pharm Sci. 1996;85(7):732–
40.
67. Al-Kahtani AA, Sherigara BS. Controlled release of diclofenac
sodium through acrylamide grafted hydroxyethyl cellulose and so-
dium alginate. Carbohydr Polym. 2014;104:151–7.
68. Dai Q, Kadla JF. Effect of nanofillers on carboxymethyl cellulose/
hydroxyethyl cellulose hydrogels. J Appl Polym Sci. 2009;114(3):
1664–9.
69. Sinha Roy D, Rohera BD. Comparative evaluation of rate of hy-
dration and matrix erosion of HEC and HPC and study of drug
release from their matrices. Eur J Pharm Sci. 2002;16(3):193–9.
70. Huang H, et al. Compression-coated tablets of glipizide using
hydroxypropylcellulose for zero-order release: in vitro and in vivo
evaluation. Int J Pharm. 2013;446(1–2):211–8.

71. Lee H-J, Kim J-Y, Park S-H, Rhee Y-S, Park C-W, Park E-S.
Controlled-release oral dosage forms containing nimodipine solid
dispersion and hydrophilic carriers. J Drug Deliv Sci Technol.
2017;37:28–37.
72. Ferrero C, Massuelle D, Doelker E. Towards elucidation of the drug
release mechanism from compressed hydrophilic matrices made of
cellulose ethers. II. Evaluation of a possible swelling-controlled
drug release mechanism using dimensionless analysis. J Control
Release. 2010;141(2):223–33.
73. Larsson M, Hjärtstam J, Berndtsson J, Stading M, Larsson A. Effect
of ethanol on the water permeability of controlled release films
composed of ethyl cellulose and hydroxypropyl cellulose. Eur J
Pharm Biopharm. 2010;76(3):428–32.
74. Rimmer S. Biomedical hydrogels: biochemistry, manufacture and
medical applications. 1st ed. Cambridge: Woodhead publishing;
2011.
75. Lamoudi L, Chaumeil JC, Daoud K. Swelling, erosion and drug
release characteristics of sodium diclofenac from heterogeneous
matrix tablets. J Drug Deliv Sci Technol. 2016;31:93–100.
76. Saeedi M, Akbari J, Enayatifard R, Morteza-Semnani K, Tahernia
M, Valizadeh H. In situ cross-linking of Polyanionic polymers to
sustain the drug release from theophylline tablets. Iran J Pharm Res.
2009;8(4):241–9.
77. Conti S, et al. Matrices containing NaCMC and HPMC: 1.
Dissolution performance characterization. Int J Pharm.
2007;333(1):136–42.
78. Dogsa I, Tomšič M, Orehek J, Benigar E, Jamnik A, Stopar D.
Amorphous supramolecular structure of carboxymethyl cellulose
in aqueous solution at different pH values as determined by rheol-
ogy, small angle X-ray and light scattering. Carbohydr Polym.
2014;111:492–504.
79. Mohammadi G, et al. The effect of inorganic cations Ca2+ and
Al3+ on the release rate of propranolol hydrochloride from sodium
carboxymethylcellulose matrices. Daru. 2009;17:131–8.
Drug Deliv. and Transl. Res.
80. Nair AB, Vyas H, Kumar A. Controlled release matrix uncoated
tablets of enalapril maleate using hpmc alone. J Basic Clin Pharm.
2010;1(2):71–5.
81. Parhi R. Development and optimization of pluronic® F127 and
HPMC based thermosensitive gel for the skin delivery of metopro-
lol succinate. J Drug Deliv Sci Technol. 2016;36:23–33.
82. Pani NR, Nath LK. Development of controlled release tablet by
optimizing HPMC: consideration of theoretical release and RSM.
Carbohydr Polym. 2014;104:238–45.
83. Novak SD, Kuhelj V, Vrečer F, Baumgartner S. The influence of
HPMC viscosity as FRC parameter on the release of low soluble
drug from hydrophylic matrix tablets. Pharm Dev Technol.
2013;18(2):343–7.
84. Nokhodchi A, Raja S, Patel P, Asare-Addo K. The role of Oral
controlled release matrix tablets in drug delivery systems.
BioImpacts. 2012;2(4):175–87.
85. Rahman MM, et al. Evaluation of various grades of
Hydroxypropylmethylcellulose matrix systems as Oral sustained
release drug delivery systems. J Pharm Sci Res. 2011;03:930–8.
86. Siepmann J, Streubel A, Peppas NA. Understanding and predicting
drug delivery from hydrophilic matrix tablets using the Bsequential
layer^ model. Pharm Res. 2002;19(3):306–14.
87. Siepmann J, Peppas NA. Hydrophilic matrices for controlled drug
delivery: an improved mathematical model to predict the resulting
drug release kinetics (the Bsequential layer^ model). Pharm Res.
2000;17(10):1290–8.
88. Peppas NA. Analysis of Fickian and non-Fickian drug release from
polymers. Pharm Acta Helv. 1985;60(4):110–1.
89. Medina JR, Cortes M, Romo E. Comparison of the USP apparatus 2
and 4 for testing the in vitro release performance of ibuprofen ge-
neric suspensions. Int J Appl Pharm. 2017;9(4):90–5.
90. Hettiarachchi TW, Wickramaratne M, Sudeshika T, Niyangoda D,
Sakeena MHF, Herath H. Comparative in-vitro evaluation of met-
formin HCl and paracetamol tablets commercially available in
Kandy district, Sri Lanka. Int J Pharm Sci. 2015;7(2):520–4.