Quantitative Component Analysis
of Solid Mixtures by Analyzing
Time-Domain 1H and 19F T1 Saturation
Recovery Curves (QSRC)
Dirk Stueber1 and Stefan Jehle2
1Materials
& Molecular Characterization,
Merck & Co., Inc., Kenilworth, NJ, USA
2NMR Application, Bruker Biospin, Fällanden, Switzerland
Background
Active pharmaceutical ingredients (APIs) often exhibit extensive polymorphism and the tendency to form solvates and
hydrates. In addition, the interaction of the desired API lead form with excipients in formulations during processing or during
long-term storage may lead to form change and/or amorphization. Consequently, API and formulated materials studied in early
drug development often contain complex mixtures composed of the desired API lead form in the presence of other polymorphs,
solvates, amorphous material, and excipients. The ability to characterize and quantify relevant API forms in these complex
mixtures in the presence of each other and excipients is crucial in the early development process because polymorphs often
exhibit distinct physical properties that may alter the dissolution and bioperformance, processability, and/or chemical stability of
formulated drug product.1
Typical analytical tools to analyze API and formulated pharmaceutical materials include X-ray powder diffraction, optical and
vibrational spectroscopy, and thermometric methods like differential scanning calorimetry (DSC) and thermogravimetry (TG).2
In recent years, high-field and high-resolution solid-state NMR (ssNMR) has emerged as an invaluable tool for analyzing API
and formulated pharmaceutical materials in the solid state.3 Several ssNMR-based methods to quantify components in
mixtures have been proposed and successfully applied. These methodologies include a number of chemometrics approaches,
signal deconvolution, corrected signal integration, and relaxation-based methods. Among the chemometrics NMR tools, the
direct exponential curve resolution algorithm (DECRA) has been applied most frequently on a variety of materials, including
pharmaceuticals, polymers, and human brain MRI.4
The method proposed here, QSRC, represents a new and very efficient
approach for quantifying the components in solid mixtures. It utilizes 1H and
19F T saturation recovery curves (SRCs) measured on a Bruker Minispec
1
mq20 benchtop TD-NMR instrument.5 For the analysis of a given mixture,
the SRCs for the relevant pure components, as well as for the mixture itself,
are measured. The relative amounts of the mixture components are
obtained from a fit of the mixture SRC with a linear combination of weighted
pure component SRCs.
Polymorphs
Solvates
Cocrystals
Salts
Amorphous
1H
Ibuprofen/indomethacin
5%-50%m Ibu
1H
Ibuprofen/itraconazole
5%-50%m Ibu
1H
2-trifluoromethyl cinnamic acid/
6-trifluoromethyl uracil
5%-90%m
2TFMCA
19F
2-trifluoromethyl cinnamic acid/
fluoxetine HCl
10%-70%m
2TFMCA
19F
T1 [s]
M [g/mol]
x
0.64
3.40
0.64
0.69
2.08
4.22
206.29
357.79
206.29
705.64
216.16
180.09
18
16
18
38
3
3
2.08
1.85
216.16
345.79
3
3
and 19F QSRC analysis for model systems (binary blends)
A. Selected
experimental pure
component and
blend SRCs, and
corresponding
QSRC fits
A
B. Point-by-point
deviations for
QSRC fit for
50.2% Ibu/
Indo blend
B
Int
Int
 Ibu, raw
 50.2% Ibu blend, raw
 Indo, raw
 Ibu, raw
 49.9% Ibu blend, raw
 Itra, raw
4 scans/inc
50 points
 50.2%
Ibu blend,
scaled
 Ibu, scaled
 QSRC Fit  50.6%
rms = 0.0082
 Indo, scaled
 2TFMCA, raw
 51.4% 6TFMU blend, raw
 6TFMU, raw
128 scans/inc
50 points
 49.9%
Ibu blend,
scaled
 Ibu, scaled
32 scans/inc
50 points
 QSRC Fit  49.2%
rms = 0.0020
 QSRC Fit  51.3%
rms = 0.0020
 51.1%
6TFMU blend,
scaled
 Itra, scaled
 2TFMCA, scaled  6TFMU, scaled
/ms
/ms
Correlation plots for QSRC 1H and 19F fitting on model systems
Ibu/Indo
4 scans/inc, 50 points, 2ms-20s
2
r =0.9995, m=1.01, b=-0.30
0.8
0.5
X
• Model compounds are small
pharmaceutical molecules
Nucleus
Ibu=ibuprofen; Indo=indomethecin; Itra=itraconazole; 2TFMCA=2-trifluoromethyl cinnamic acid; 6TFMCA=6-trifluoromethyl cinnamic acid; FXT=fluoxetine
FID
t [s]
0.015 0.020
• Model compounds have different
and similar 1H/19F T1s
Blends
Bruker Minispec mq20
Int
1.0 Spectrum
0.010
Model System
/ms
TD-NMR on a Bruker Minispec mq20
Int
1.0
• Binary blends of model compounds
with known compositions were
analyzed with QSRC
Electronics
box
0.6
FFT
Ibu/Indo
4 scans/inc, 30 points, 2ms-20s
r2=0.9994, m=1.05, b=-0.24
Ibu/Indo
4 scans/inc, 10 points, 2ms-20s
r2=0.9906, m=1.48, b=0.97
Ibu/Indo
32 scans/inc, 50 points, 2ms-20s
r2=0.9985, m=1.00, b=0.20
Ibu/Indo
32 scans/inc, 30 points, 2ms-20s
r2=0.9980, m=1.00, b=0.48
Ibu/Indo
32 scans/inc, 10 points, 2ms-20s
r2=0.9978, m=0.99, b=0.42
Magnet
Chiller
0.4
Ibu
0.2
-.05
50
Int
1.000
Avg int
of first
points
0.995
150
ppm
200
Int
1.0
0.8
0.6
0.985
0.4
Indo
• No FFT, no spectrum
Spectrum
0.990
0.980
100
• Record only first points of
FID and average → build
relaxation curves
Sample:
10-mm glass tube
100–200 mg
Ibu/Itra
4 scans/inc, 50 points, 2ms-20s
r2=0.8836, m=0.77, b=24.8
• Very simple and fast sample prep
10
15
tau [s]
20
Itra
• Linear coefficients, ci, are relative concentrations
SRCmix =
i=1
ci SRCi +b
SRC
norm
i
Int
 = 5.0 s
1.0
SRCi = Ii,1 ,Ii,2 ,Ii,3 ,…,Ii,n
SRC i x nuclei/molecule

I i, τ 5T1 M i
xnuclei/molecule= moles of
observed nuclei per
moles of molecules.
0.6
• POC for using QSRC method for 1H and 19F SRCs has been shown for several model systems
• Separation of components with close T1s requires more scans/inc
C1: T1 = 1.0 s
• Significant total time savings with respect to conventional ssNMR techniques
• Advantages of QSRC – Bruker Minispec mq20:
– Robust, accurate, and fast
– Trivial sample preparation (glass tube sample holder), T-control, and automation possible
– No requirements on sample texture or homogeneity (tablets, gels, polymers, …)
– Amenable to industrial high-throughput settings, production sits (Pharma Industry, …)
SCR for mix:
C1:C2 = 1:1
• Patent for QSRC – Bruker Minispec mq20 filed (QSRC module in Dynamics Center)
0.4
References
• Optimal coefficients are determined in minimization
0.2
• Optimization routine implemented in mathematica code
HCl
• SRCs are efficiently collected on a Bruker Minispec mq20 benchtop NMR instrument
M = molecular mass.

 N norm
norm
norm 
MinimizeSRC mix    ci SRC i  b 
 i 1


FXT
• Proposed QSRC method uses 1H and 19F T1 SRCs as fingerprints for expected components in solid mixtures
– SRCmix = weighted linear comb SRCscomp
C2: T1 = 4.0 s
0.8
• Component SRCs are normalized
2TFMCA
Summary/Conclusions
Illustration of method:
Hypothetical twocomponent system
• SRCmix is a linear combination of component SRCs (SRCi)
N
FXT/2TFMCA
128 scans/inc, 50 points, 2ms-20s
r2=0.9999, m=0.99, b=0.18
• Automation and T-control available
QSRC approach:
Intensities Ii at n
recovery time points
6TFMU/2TFMCA
32 scans/inc, 50 points, 2ms-40s
r2=0.9999, m=0.99, b=0.28
6TFMU
QSRC: Form Quantification Using TD-NMR SRC Data
N mixture components
• Excellent agreement
between prep and fitted
compositions
• Notably more scans/
inc necessary for
components with close T1
• Well-characterized relaxation curves
0.2
5
Ibu/Itra
128 scans/inc, 50 points, 2ms-20s
r2=0.9968, m=1.01, b=0.07
• No resolution necessary → obs. 1H
Saturation recovery
 T1
0.975
Ibu/Itra
32 scans/inc, 50 points, 2ms-20s
r2=0.9868, m=0.93, b=1.90
5
10
15
20
𝐼𝑚𝑖𝑥,𝜏=5.0𝑠 =1 2 𝐼1,𝜏=5.0𝑠 + 1 2 𝐼2,𝜏=5.0𝑠
tau [s]
1. Chemburkar SR. Org Process Res Dev. 2000;4:413-417.
2. Brittain HG. Physical Characterization of Pharmaceutical Solids.
New York, NY: Marcel Dekker, Inc.; 1995.
3. Pham TN. Mol Pharm. 2010;7:667-1691.
4. Alam TM. Ann Rep NMR Spect. 2004;54:41-80.
5. Schwartz LJ. J Chem Ed. 1988;65:959-963.
6. van Duynhoven J. Ann Rep NMR Spect. 2010;69:145-197.
7. Dalitz F. Prog NMR Spect. 2012;60:52-70.
Contact
Corresponding author: [email protected]
Copyright © 2016 Merck Sharp & Dohme Corp., a subsidiary of Merck & Co., Inc. All rights reserved.
stueberd_189874-0001_ENC_poster • 4/6/2016 2:02 PM • ENC • 44” x 44” @200%