Applications of NIR Spectroscopy in Pharmaceutical Analysis

1
Applications of NIR Spectroscopy in
Pharmaceutical Analysis

• Rodolfo J. Romañach
• UPR-Mayagüez

2
Pharmaceutical Applications NIR

1. Blend Uniformity.
2. Chemometrics The Guiding Discipline.
3. Identification of Raw Materials.
4. Drug Content in Tablets.
5. Comparing NIR HPLC (or other method) the
   three possible errors.
6. Wet Granulation (Fluid Bed Drying)
7. Understanding Powder Voiding.
8. Chemical Imaging

3
Interest in Blend Uniformity

• CGMPs require control procedures to monitor and
  validate the adequacy of mixing to assure
  uniformity and homogeneity
• Problems in Blend Uniformity Analysis (BUA) using
  sample thieves.
• Optimization of Blending is needed due to meet
  the strict requirements on drug distribution on
  final dosage form.

4
Blending Optimization (R. Hwang, Pharm. Tech.
March 1998, 158-170)
5
Blend Uniformity Profile Obtained with NIR
Spectroscopy
S.Sekulic, H.W. Ward, D.R. Brannegan, et.al.
On-Line Monitoring of Powder Blend Homogeneity
by Near-Infrared Spectroscopy, Analytical
Chemistry, 1996, 68, 509-513.
Monitoring of V-blender by a spectroscopic method
through axis originally patented. A disclaimer to
all claims of patent was filed in April 2003.
6
Notes on Standard Deviation Graph

• Average standard deviation over the many
  wavelengths collected.
• The standard deviation of spectra is a spectrum.
• Rise at the end occurs after addition of
  magnesium stearate for lubrication.

S.Sekulic, H.W. Ward, D.R. Brannegan, et.al.
On-Line Monitoring of Powder Blend Homogeneity
by Near-Infrared Spectroscopy Analytical
Chemistry, 1996, 68, 509-513.
7
Spectra Obtained
Components of Blend
Experiments Obtained During Blending
8
Moving block standard deviation
9
Blend Uniformity within the PAT Context

• Blending process will specify a range of mixing
  times.
• Idea is that you mix until it is homogeneous
  rather than for a specific time point.
• Will require some research to establish adequate
  manner to establish end-point.
• Will you base your end-point on the homogeneity
  of API, or all excipients?
• Will you evaluate the homogeneity of the blend
  from a chemical point of view, or from chemical
  physical points of view.

10
Recent Research

• Compared three approaches to establishing end
  point monitoring changes in drug content,
  monitoring moving block std dev., and root mean
  standard nominal value (RMSNV).
• RMSNV calculated from predictions of API,
  lactose, and microcrystalline cellulose.
• Authors found that moving block method was not
  sensitive to changes in MCC and lactose since
  their spectra are similar.
• Authors found RMSNV to be the most robust of the
  methods.

Zhenqi Shi, Robert P. Cogdill, Steve M. Short,
Carl A. Anderson, Process characterization of
powder blending by near-infrared spectroscopy
Blend end-points and beyond, Journal of
Pharmaceutical and Biomedical Analysis, 2008, 47,
738 745.
11
Theory of Analytical Chemistry

• A guiding theory of analytical chemistry can be
  used to specify what information can be extracted
  from the data produced by an analytical
  instrument or method. In addition, it can be used
  to optimize existing analytical tools and direct
  researchers to construct more powerful analytical
  tools.

K.S. Booksh and B.R. Kowalski, Theory of
Analytical Chemistry, 1994, 66(15), 782 A 791A.
12
Chemometrics-Historical Perspective

• In 1969, Jurs, Kowalski and Isenhour published a
  series of papers in Analytical Chemistry on the
  application learning machines to interpret mass
  and infrared spectra. Work focused on
  interpretation of data.
• Growing belief that some of the new mathematical
  methods or theories, such as pattern recognition,
  information theory, operations research, etc.,
  are relevant to the basic aims of analytical
  chemistry, such as the evaluation, optimization,
  selection, classification, combination and
  assignment of procedures, in short all those
  processes involved in determining exactly which
  analytical procedure or programme should be used

D.L. Massart, A. Dijkstra and L. Kaufman,
Evaluation and Optimization of Laboratory Methods
and Analytical Procedures. Elsevier. Amsterdan,
1978 as quoted in Handbook of Chemometrics and
Qualimetrics Part A. D.L. Massart, et. Al.
Elsevier, 1997, page 12.
13
Chemometrics-Historical Perspective II

• June 1972 Wold uses the word chemometrics in
  paper published in Swedish journal. Calls his
  group Forkningsgruppen for Kemometri, and
  Kowalski calls his group the Laboratory of
  Chemometrics.
• Wold and Kowalski form the Chemometrics society
  in 1974.
• Early 1980s establishment of Center for Process
  Analytical Chemistry (CPAC) at the University of
  Washington, Seattle.
• Partial least squares introduced by S. Wold and
  co-workers in 1983.
• Journal of Chemometrics (Wiley) and Chemometrics
  and Intelligent Laboratory Systems (Elsevier)
  launched in 1986.

Handbook of Chemometrics and Qualimetrics Part
A. D.L. Massart, et. Al. Elsevier, 1997, page 12.
14

• Chemometrics is a chemical discipline that
  uses mathematics, statistics and formal logic
• to design or select optimal performance
  experimental procedures.
• To provide maximum relevant chemical information
  by analyzing chemical data.
• To obtain knowledge about chemical systems

Design Learn (Model) Use
Handbook of Chemometrics and Qualimetrics Part
A. D.L. Massart, et. Al. Elsevier, 1997, page 1.
15
The Chemometric Approach

• First, establish a purpose
• Simple Understanding Exploratory Analysis (look
  at the data, carefully examine it, get a simple
  visual idea about the main relationships between
  samples. Chromatogram may contain hundreds of
  peaks but the human eye cannot tell those that
  vary the most from sample to sample) Learn
• Property Prediction regression modeling. Compare
  spectra or patterns. Model
• Automate Routine Predictions if modeling
  succeeds.
• Use

Infometrix, Chemometrics Training Course, 2004,
R.G. Brereton, Chemometrics Data Analysis for the
Laboratory and Chemical Plant, Wiley, 2003, page
183.
16
NIR in ID - Examples
Pattern recognition methods are used to compare
NIR spectra and measure similarity. Cannot do
visual comparison of spectra because of the wide
overlapping bands.

• Pharmaceutical excipients, API, and other
  compounds may be identified by NIR. Blanco
  developed library at Menarini España for over 125
  excipients, that is still in use. M. Blanco and
  M. A. Romero, Near-infrared libraries in the
  pharmaceutical industry a solution for identity,
  The Analyst, 2001, 126, 22122217
• ID of tablet formulations inside blisters.
  Distinguish between formulation with 66 active
  ingredient and two placebos. P.K. Aldridge, R.F.
  Mushinsky, M.M. Andino, and C.L. Evans, Appl.
  Spectrosc., 1994, 48, 1272 1276.
• Identification of counterfeit drugs using NIR.
  S.H. Frasson Scafi, C. Pasquini, Analyst, 2001,
  126, 2218 2224.

17
NIR - Identification

• Possibility of identifying all compounds within a
  pharmaceutical plant.
• Method should be able to distinguish between
  polymorphs and materials that are chemically
  identical but may have different particle sizes.
• Should be able to distinguish between different
  manufacturers or different grades of a product.

M. Blanco and M. Alcalà, Process Analytical
Technology, Blackwell Publishing, K. Bakeev. Ed.
18
ID of Raw Materials

• Developed spectral library with about 125
  spectral excipients. A total of over 3000 spectra
• Found that in most cases the correlation
  coefficient was sufficient to discriminate
  between the compounds.
• At least 3 batches per substance were used with
  triplicate spectra, to have at least 9 spectra
  for each class.
• Suggest idea of cascading identification.
  Identifying compound versus general library, and
  if inconclusive then a second smaller library is
  used.
• Still in place, still working.

M. Blanco and M. A. Romero, Near-infrared
libraries in the pharmaceutical industry a
solution for identity, The Analyst, 2001, 126,
22122217
19
Steps for Develop ID Method - Blanco

• Recording of NIR spectra using a set of known
  (Certificate of Analysis, mid-IR, etc.) samples
  from several batches of materials.
• Constructing the spectral library (involves a
  number of steps).
• Constructing sub-cascading libraries (that
  include mutually related materials.
• External Validation (check system with validation
  set).

CASCADING LIBRARIES
Identification of Product within large library
Define a characteristic that differentiates
materials such as particle size or moisture.
Smaller spectral variability (differences in
impurity, origin of manufacture, etc)
M. Blanco and M. Alcalà, Process Analytical
Technology, Blackwell Publishing, K. Bakeev. Ed.
20
Constructing the Spectral Library Blanco (Step
2)

• Choose pattern recognition method to use (vector
  correlation, distance, etc.).
• Choose spectral pretreatment method, range of
  wavelengths, threshold.
• Evaluate threshold, wavelength range, whether to
  keep spectral pretreatment.

21
Spectra of Sucrose Samples Blanco
22
Typical Correlation Values Obtained in Crossed
Identification of Different Types of Sucrose.
Blanco
Sub-library needed in this case.
23
Situation Analysis

• NIR could easily analyze 30 40 tablets per hour
  with current existing technology.
• A number of formulations are being approved with
  potent drug (low drug content).
• A 0.2 mg aggregate of drug is not important in a
  100 mg unit dose, but is important in a 0.1mg
  unit dose.

• USP requires analysis of 10 30 tablets in a
  batch that may include 3,000,000 or more tablets.
  
• Sometimes one tablet is outside of USP limits,
  and an investigation is required.

C. Peroza Meza, M.A. Santos, and R.J. Romañach,
Quantitation of drug content in a low dosage
formulation by Transmission Near Infrared
Spectroscopy, 2006, 7(1), Article 29
(http//www.aapspharmscitech.org).
24
Application

• Calibration set with 110 tablets of 0.5, 0.7,
  and 1.0 (w/w) ibuprofen. Average tablet weight
  was about 250 mg.
• Transmittance Spectra obtained.
• 48 independent tablets in validation set.
• Used UV method as reference method for drug
  content.

Calibration is the mathematical and statistical
process of extracting information, usually
analyte concentration, from the instrument signal
(Booksh and Kowalski).
25
Exploratory Analysis. Want to relate changes in
the spectra to changes in drug concentration or
other property.
Variation Implies Information !!
26
Partial Least Squares (PLS) MODEL

• The PLS algorithm will decompose the X and Y
  data, obtaining the maximum covariance between
  the X and Y variables.
• Develop a set of vectors that relate the changes
  in the spectra (variation patterns) to the
  changes in the drug content.

27
PLS Overlapping Bands
28
Evaluation of pretreatment and spectral area for
quantitative method
29
Part II - Evaluation of pretreatment and spectral
area for quantitative method
30
Validation Set Tablets
31
Drug Content Prediction
32
Calibration with Error

• Three types of error
• Random error in the reference laboratory values
• Random error in the optical data
• Systematic error in the relationship between
  these two. (e.g. Differences in the sample size
  of the two methods sampling error)

From Principles and Practice of Spectroscopic
Calibration H. Mark, John Wiley Sons, 1991,
p. 17
33
Comparison of Sampling Volume
Systematic error in the relationship between the
optical (NIR) method and the reference method.
First 2 mm sampled by NIR beam vs.
Entire Sampled Analyzed by HPLC or UV method
34
Fiber Probe and Sample Holder
Top NIR talc spectrum, middle
ibuprofen-lactose powder mixture,
bottom-ibuprofen-lactose over thin layer of talc.
Talc estimated to be 1.8 2.0 mm below powder
mixture
M. Popo, S. Romero-Torres, C. Conde and RJ
Romañach, AAPS PharmSciTech 2002 3 (3) article
24 (http//www.aapspharmscitech.org/).
35
Effect of Scale of Scrutiny
Provided a statistical proof to show that the
coefficient of variance (CV) of sampling portions
is greater than or equal to the CV of sampling
whole units.
T. Li, A.D. Donner, C.Y. Choi, G.P. Frunzi, and
K.R. Morris, A Statistical Support for Using
Spectroscopic Methods to Validate the Content
Uniformity of Solid Dosage Forms, J. Pharm.
Sci., 2003, 92, 1526 1530.
36
Fluid Bed Granulation Drying

• Developed calibration models to monitor KF and
  LOD during drying of granulations.
• Initial work in 65L 300 L were performed with
  fiber optic probe inserted directly into the bowl
  without sample cup, but obtained spectra of low
  signal to noise ratio.
• Overcame problem with device to collect
  stationary sample (Fig. 2)

R.L. Green, G. Thurau, N.C. Pixley, A.
Mateos, R. A. Reed, and J.P. Higgins, In-Line
Monitoring of Moisture Content in Fluid Bed
Dryers Using Near-IR Spectroscopy with
Consideration of Sampling Effects on Method
Accuracy, Anal. Chem. 2005, 77, 4515-4522.
37
Fluid Bed Granulation Drying
38
FBD Application

• Collected nearly 750 in-line NIR spectra and
  pulled 119 samples for KF reference analysis,
  ranging from 3.65 - 25.24 (w/w)
• Graph shows measurements every 70 s. Obtained
  RMSE of 0.7 from 3.6-10, and 0.2 in range of
  3.6 5.5.
• Observed a number of spikes that appear to be
  related to moist material discharged from
  filters. Indicate that this was never observed
  before monitoring with NIR.

Method accuracy depended on calibration range.
RMSEC, RMSECV, RMSEP were smaller as the high end
of calibration range was decreased. Residual
errors were higher at higher moistures. Indicate
It was determined that discrepancy in sampling
location plays a role in the magnitude of
prediction residuals and therefore apparent
method accuracy. Modified sample cup to allow
retrieving sample with thief (Fig 2c).
39
FDB Application

• After modifying the sample cup considered that
  the new cup performs better but some sampling
  error was still occurring.
• Established two calibration models, one with a
  wide calibration range, and a second with a 4
  5 (w/w) range based on endpoint for the
  formulation.

40
Advantage of PCA Scores in Understanding a
Granulation Process
J. Rantanen, H. Wikström, R. Turner, and L.S.
Taylor, Use of In-Line Near-Infrared
Spectroscopy in Combination with Chemometrics for
Improved Understanding of Pharmaceutical
Processes, Anal. Chem., 2005, 77, 556 563.
41
M.J. Barajas, A. Rodriguez Cassiani, W. Vargas,
C. Conde, J. Ropero, J. Figueroa, and R.J.
Romañach, A Near Infrared Spectroscopic Method
for Real Time Monitoring of Pharmaceutical
Powders during Voiding, Applied Spectroscopy,
2007, 61(5), 490 496.
42
NIR Spectra obtained During Voiding. Left spectra
obtained, and top after subsctracting
Applied Spectroscopy, 2007, 61(5), 490 496.
43
(No Transcript)
44
Comparison of Commercial Product to Particle Size
Fractions During Voiding
45
Spatial distribution of Ibuprofen, lactose and
microcrystalline cellulose
Single channel image 1620 nm
Data treatment Savitzky-Golay smoothing and SNV