Computer Simulation of the Effect of Temperature on pH
JAMES
E. KIPPX AND DAVIDF. SCHUCK
Received March 8, 1995, from the William Graham Science Center, Baxter Healthcare Corporation, IV Systems Division,
Round Lake, Illinois 60073. Accepted for publication August 1, 1995@.
Abstract 0 The effect of temperature on solution pH was simulated by
computer (program PHTEMP). We have determined that the change in
pH due to shifts in acid-base equilibria (ApH = pH(6O"C) - pH(25"C))
can be substantial for compounds such as aliphatic amines that have
high enthalpies for acid dissociation. This is of particular significance
during elevated temperature experiments in which changes in the p&
values of formulation components, and hence the solution pH, can
accelerate decomposition as compared to those formulations where
sensitive functionality is absent. PHTEMP afforded the following results
at initial pH = 7 (25 "C): (a) 0.1 M triethylamine (AHS = 10.4 kcal/mol)
ApH x -0.8; (b) 0.1 M acetic acid (AH" = -0.1 kcal/mol) ApH x 0;
(c) 0.1 M sulfuric acid (AH", = -12 kcal/mol; AW2 = -5.4 kcal/mol)
ApH x -0.4. Solutions of general pharmaceutical interest were also
studied and included a 12-component amino acid mixture, 0.1 M glycine,
and 0.1 M triethylamine in either 0.02 M citric acid or 0.05 M TRlS buffer.
In each case the pH change with temperature was dependent on the
concentrations of components, the enthalpies for each acid dissociation,
and the starting pH. At lower pH (<4), PHTEMP predicts that ApH is
typically smaller than at higher pH (>9). These results are interpreted
as the effect of a relative change in hydronium ion activity, AH+/H+(initial),
due to temperature-induced shifts in equilibria (acid dissociation, water
autoprotolysis). This relative change must become larger as H+ decreases
(pH increases). The output of PHTEMP was experimentally verified with
0.1 M glycine and with a multiple component amino acid solution. In
both cases, agreement with prediction was excellent. The results of this
investigation underscore the need to critically review formulation choices
for both thermodynamic and traditional kinetic effects on the resulting
product stability.
Introduction
The measurement and control of solution pH is critical in
pharmaceutical development. Accelerated screening of can-
didate formulations is often conducted at temperatures far
higher than that of the anticipated product storage condition.
High-temperature data may be fitted to a temperature-
dependent kinetic model (e.g., Arrhenius or Eyring) in order
to predict stability a t the product storage temperature, or the
results may be used qualitatively to ascertain the effects of
solution parameters (e.g., pH, ionic strength, buffer strength)
on stability. It may be erroneously assumed that the effects
of temperature on pH can be neglected. For example, test
solutions could be adjusted to a given pH at room temperature
and decomposition carried out a t elevated temperature.
Similarly, pH may be improperly neglected because of the
experimental difficulty associated with pH measurements at
elevated temperatures.
We present the results of simulations using a computer
program (PHTEMP) that is used to predict the change in pH
due to shifts in equilibria and solute activities with temper-
ature in single and multiple component formulations. We
have found such simulations useful in evaluation of buffer
@Abstractpublished in Advance ACS Abstracts, September 15, 1995.
0 1995, American Chemical Sociev and 0022-3549/95/3184-1347$09.00/0
American Pharmaceutical Association
systems for temperature sensitivity in situations where pH
measurement a t very high temperature is impractical.
The effect of temperature on measured pH depends on the
temperature dependence of the electrode response (Le., equi-
libria shifts within the electrode and changes in junction
potentials) and on shifts in solution equilibria within the
sample. Introductory physical chemistry texts typically dis-
cuss the Nernst slope' to describe the temperature dependence
of the electrode, while changes in solution equilibria with
temperature are extensively described in classroom chemistry
texts and in classic compendia on pH measuremenL2 Because
of the mathematical complexity of predicting the temperature
dependence of multiple component systems, most of these
treatises are limited t o relatively simple systems.
A computer program (PHTEMP) was written to predict the
temperature dependence of acid-base equilibria, wherein
multiple solution components with multiple contributing
equilibria are present. PHTEMP is based on an earlier
program3 (PHCALC)which calculates ionic equilibria at fured
temperature (25 "C) of any solution comprised of up to 54 acids
and bases, including polyprotic acids. In PHTEMP, the
equilibrium constants at any user-selected temperature are
determined from the van 't Hoff equation using published
thermodynamic values (K, a t 25 "C and the standard acid-
dissociation enthalpy, AH").4 As in PHCALC, PHTEMP sets
up the system of mass balance, charge balance, and equilib-
rium equations in a matrix, and solves the matrix equation
to obtain estimates of the concentrations of all solution species.
The calculated concentrations are used to determine ionic
strength (I,in mom) and hence activity coefficients, yi, at any
given temperature by use of an extended form of the Debye
Hiickel e q ~ a t i o n : ~
hi2&
log yi = -
+
1 Bai&
Parameters A and B are defined as
B = e(2N,,leoer,AkT11'2 (3)
Constants e, No, k, e,, and er,A, T , and zi are the elementary
charge, Avogadro's constant, Boltzmann's constant, free charge
permittivity in vacuo, solvent dielectric constant, temperature
(degrees Kelvin), and charge of ion i, respectively. The ionic
is an empirical parameter that is entered by the user
size (ai)
as an average value. The ionic strength, I , is defined as
where ci and zi are the concentration (molarity) and charge
of each ionic species present in solution.
The programming strategy of PHTEMP is otherwise identi-
cal to PHCALC.
We have used PHTEMP to estimate the change in pH for a
temperature increase from 25 to 60 "C as a function of the
initial pH. The following solutions were studied: (a) single
Journal of Pharmaceutical Sciences / 1347
Vol. 84, No. 11, November 1995
component solutions of triethylamine (0.001 and 0.1 M), Table 1-Standard Enthalpies of Acid Dissociation for Different
sulhric acid (0.1 M), and acetic acid (0.1M), (b) triethylamine Functional Group Classes
(0.1 M) in the presence of buffer systems [citrate and tris- Standard
(hydroxymethy1)aminomethane (TRIS)], and (c) a multiple Acidic Enthalpy
component amino acid formulation. Functional p&at Change
These examples were chosen in order to understand the Group Examples 25 'Ca (kcallmol) RefC
temperature effects in common with pharmaceutical prepara- Sulfate, Sulfonate Hydrogen sulfate 1.99 -5.4 4, 79
tions. (sdfuric, pK2)
In order to test the accuracy of PHTEMP, we determined Carbonate Carbonic acid 6.35 (PKi) 2.0 4, 37
the pH change in solutions of glycine and in a prepared 10.33 (pKz) 3.5 4,37
mixture of amino acids (see Table 2 for composition). Carboxylic Formic acid 3.74 -0.04 5,284
Acetic acid 4.76 4.1 5,284
Lactic acid 3.86 -0.08 5,290
Experimental Section Glycine (R-COOH) 2.35 1.o 1,1
Preparation of Solution and pH Measurement-Glycine Phosphate Phosphoric acid 2.15 (PKi) -1.9 4,56
(Baker), proline (Aldrich), and phenylalanine (CalBiochem) were of 7.20 (PKz) 0.8 4,56
analytical reagent grade quality. All other amino acids (leucine, 12.4 ( P W 3.5 4,56
alanine, methionine, isoleucine, valine, histidine, threonine, tryp- Glycerol-2-phosphate 1.34 (pK) -2.9 3,172
tophan, and tyrosine) were purchased from Ajinomoto Co., Inc. (Tokyo, 6.65 ( p k ) -0.4 3,172
Japan), and were USP grade. All reagents were used as is without a-o-(+)-glucose-1-(dihydrogen 1.46 (pKi)b NAd 3,173
further purification. Prior to pH measurement, test solutions were phosphate) 6.50 (P&) -0.4 3,173
equilibrated in a thermostated constant temperature bath. The
temperature of each solution sample was monitored with a n Omega Phenols Phenol 9.98 5.5 5,445
871 digital thermometer. Solution pH was measured with an Orion Salicylic acid (Ar-OH) 13.74 8.56 5,336
720A digital pH meter with a combination glass electrode (Corning). Catechol 9.23 (pKi)b 6.0b 3,200
The pH meter was calibrated at the experimental temperature by 13.0 ( ~ K z ) ~ 56 3,200
bracketing with two commercially available standards of either Thiols Ethanethiol 10.61 6.4 3,335
potassium tetraoxalate (pH 1.68 at 25 "C, Radiometer), potassium Propane-2-thiol 10.86 5.4 3,335
hydrogen phthalate (pH 4.0 a t 25 "C, Baxter Diagnostics), disodium 2-Methylpropane-2-thiol 11.22 5.3 3,335
hydrogen phosphate (pH 6.84 a t 37 "C, Radiometer), or sodium borate
(pH 9.18 a t 25 "C, Radiometer). Standard pH values for the buffer Ammonium Methylamine 10.64 13.2 2, 1
solutions a t the measured experimental temperatures were taken Diethylamine 10.93 12.8 2,72
from Bates (ref 2c, p 76). Amino acid solutions were prepared a t the Triethylamine 10.72 10.4 2,112
concentrations listed in Tables 2 and 3, using filtered, distilled, and Glycine (RNH3+) 9.78 10.6 f,1
deionized water which had been thoroughly degassed by sparging with Tris(hydroxymethy1)- 8.075 11.36 2, 20
nitrogen. The solutions were pH adjusted with either 5 M sodium aminomethane (TRIS)
hydroxide (Ricca) or 5 M HCl (Ricca), a s appropriate.
C o m p u t e r Simulations-The program PHTEMP was written in Water 13.997 13.34 4, 1
Pascal (Turbo Pascal version 7.0, by Borland). All data processing a A t zero ionic strength, unless otherwise indicated. Ionic strength = 0.1, T
was performed on a 80486 microcomputer with floating-point copro- = 25 "C. All references are from Critical Stability Constants; Volume and page
cessor using extended-precision reals (10 bytes). A copy of the
number (see ref 4). Not applicable.
executable file with instructions is available upon request.
PHTEMP was used to simulate titration of the acid in
Results and Discussion question at constant volume with either base (NaOH) or acid
(HC1) at 25 and 60 "C. The data were graphed as plots of
General Features-The extent to which the acid-base ApH versus pH(Tl), where ApH = pH(T2)- pH(Tl), TI= 25
equilibrium will change upon heat input is dependent on the "C, and T2= 60 "C. Constant ionic strength was not employed
standard enthalpy of acid dissociation, AH'? during the simulations. However, the salt effect on pH shift
with temperature must be slight because calculations at 25
and 60 "C which were performed with and without activity
corrections predicted shifts that were very similar, even
though the differences between uncorrected ( I = 0) and
Enthalpies of many common organic and inorganic acids corrected pH estimates a t each temperature were significant
and bases are available from the literature and are reproduced for certain initial pH values. In other words, although ionic
in Table 1for reference. Typically, enthalpies of dissociation strength has a slight impact on absolute pH, this effect is
for weak organic acids (e.g., carboxylic acids such a s acetic nearly canceled in the difference (ApH).
acid) are close t o zero. Therefore, the pH-temperature The standard enthalpies of acid dissociation and the dis-
dependence for many pharmaceutical formulations that con- sociation constants used to generate pH-temperature curves
tain carboxylic acid groups is expected to be insignificant. are shown Table 2.
Weak acids with high pK, (>9), such as conjugate acids of PHTEMP does not account for changes in the enthalpies of
aliphatic amines, generally have much higher positive en- dissociation with changing temperature. Enthalpy corrections
thalpies. Deprotonation of triethylammonium ion, for ex- are typically made by employing quadratic or other math-
ample, is quite endothermic (ca. 10 kcal/mol). On the other ematical corrections t o the standard enthalpy value. Since
hand, the deprotonation of some strong acids is significantly such expressions are not readily available for each substance
exothermic. For example, the acid dissociation of bisulfate input into PHTEMP, no attempt was made to incorporate heat
ion to form sulfate has a standard enthalpy of approximately capacity corrections into PHTEMP. The degree of error
-5 kcdmol. We chose, therefore, to test PHTEMP with acetic associated with the constant enthalpy assumption in the
acid, triethylamine, and sulfuric acid as representative single temperature range from 25 to 60 "C is predicted in most cases
component formulations for which the behavior could be to be minor. This is because carboxylic acid compounds
readily understood before simulating more complex formula- typically have small enthalpies over a wide variety of tem-
tions. peratures.6 On the other hand, other compounds (amines)

1348 / Journal of Pharmaceutical Sciences

Vol. 84, No. 11, November 1995
Table 2-Computer Simulation Parameters Used in PHTEMPa-cse 0.2 -

Concn, AWi, - --
Components M p6 kcalhol pK2, plG
Ap2,Ap3>
kcallmol Refd
0
pK - 40.72

~
(D -0.2 -
Triethylamine
Single-Component Solutions
0.001, 0.1 10.72 10.4 2,112
m
2
0
-0.4 - \
Acetic Acid 0.1 4.76 -0.1 5,284 r
Sulfuric Acid 0.1 4 -12 0.97 -5.4 4,79 O. -0.6 -
Glycine 0.1 2.350 1 9.778 10.6 1,1 i \ \ 1
Mixture 1
Triethylamine 0.1 10.72 10.4 2,112 : : ! I : : :
TRlS 0.05 8.075 11.36 2, 20 0 2 4 6 8 1 0 1 2 1 4
Mixture 2 pH at 25 Degrees C
Triethylamine 0.1 10.72 10.4 2,112 Figure 2-ApH versus pH(25 "C) for 0.001 M triethylamine, a = 3 x cm
Citric Acid 0.02 3.182 1.0 4.761,6.396 0.58, -0.80 3, 161 (O),output from PHTEMP. The solid line is calculated using eqs 6a and 8a.
Amino Acid Mixture
Alanine 0.200 2.348 0.7 9.867 10.8 1,4
Glycine 0.230 2.350 1 9.778 10.6 1,1
Leucine 0.040 2.329 0.4 9.747 10.8 1,ll
Phenylalanine 0.032 2.20 1 9.31 10.5 1,18
Methionine 0.033 2.20 0 9.05 10 1,50
lsoleucine 0.031 2.319 0.3 9.754 10.8 1, 12
Valine 0.033 2.286 0.1 9.718 10.8 1,g
Histidine 0.024 1.7 0.7 6.02, 9.08 7, 10.5 1,61
Proline 0.031 1.952 0.3 10.640 10.6 1,69
Threonine 0.030 2.088 1.3 9.100 10 1,37
Tryptophan 0.0074 2.35 0 9.33 10.7 1, 63 : : : : : : : : : : : : :
Tyrosine 0.0019 2.17 0 9.19, 10.47 10.1,5.8 1, 31 0 2 4 6 8 1 0 1 2 j 4
pH at 25 Degrees C
a Adjusted with NaOH or HCI. bAll simulations run from 25 "C (TI) to 60 "C
( 5 )using an effective ion size of 3 A (see the text) and critical constants for Figure 3-ApH versus pH(25 "C) for 0.1 M acetic acid using a = 3 x cm
water dissociation (see Table 1). Some equilibrium constants not extrapolated (0)and a = 7 x cm (0),output from PHTEMP. The solid line is calculated
to zero ionic strength. See reference for exact experimental conditions. dAll using eqs 6a and 8a.
references are from Critical Stabikty Constants (see ref 4); volume, page number.
"KW = 13.997 (A," = 13.34 kcallmol) was used in all simulations (see Table
incorporation into eq 1. PHTEMP was run twice for 0.1 M
1).
triethylamine and 0.1 M acetic acid using effective average
ion sizes of 3 and 7 A. Ion size had no discernible influence
on the predicted pH change. An average effective diameter
-- I
of 3 A was therefore chosen in other calculations.s
0
Y. pK - 10.72 Figures 1-3 illustrate a general aspect of ApH vs pH(T1)
plots: The pH-temperature dependence increases as pH(T1)
increases. A linear plateau a t low pH is anticipated based
on the large hydronium ion activity, and the fact that pH is a
logarithmic function of this activity. Any shift in acid-base
equilibrium is overshadowed by the large hydronium ion
activity a t lower pH. The pH shift should therefore approach
zero as pH is reduced, assuming activity and concentration
0 2 4 6 8 1 0 1 2 1 4 are equal. When pH is sufficiently high, the hydronium ion
pH at 25 Degrees C contribution due to a shift in weak acid-base equilibrium is
Figure I-ApH versus pH(25 "C) for 0.1 M triethylamine, a = 3 x cm ( 0 ) significant compared to the initial hydronium ion concentra-
and a = 7 x cm (0),output from PHTEMP. The solid line is calculated tion. Because dissociation of triethylamine is endothermic
using eqs 6a and 8a. ( A F > O), an increase in temperature must lead to a shift in
equilibrium such that more hydronium ion is produced (by
may show a more significant dependence with temperature. Le Chatelier's principle). As pH(T1) is increased, a point is
The resultant error in the ApH estimate is, in most cases, reached where the magnitude of this shift in hydronium ion
likely to be minor. This is supported by the comparable concentration is equal to or greater than the initial concentra-
results of PHTEMP with the experimental data presented tion of hydronium ion. This results in a significant decrease
below for the amino acid mixture. Olofsson, et aL7 have in pH in the midregion of the pH(T])-ApH profile (pH 5-7),
reported heat capacity data from 25 to 75 "C for a variety of leading to an inflection point in the curve (Figures 1 and 2).
different m i n e s and have also determined the corresponding In contrast, acetic acid (Figure 31, with a nearly zero enthalpy
enthalpies of acid dissociation. Heat capacities were typically of dissociation, shows very little change in acid-base equi-
less than 25 kcaYmol and yielded only minor changes in librium with temperature, and the inflection point is accord-
enthalpy over the temperature range (25 to 60 "C) covered in ingly shifted to above pH 8.
PHTEMP. Figures 1-3 illustrate the effects of a high enthalpy of
Solution pH-Temperature Effects for Single Com- dissociation from the autoprotolysis of water (AHH"= 13.3 kcaY
ponent Formulations (A?fo> 0, @ zz 0, and AHo < mol). The inflection in the acetic acid curve (Figure 3) is
0)-(a) Triethylamine (H > 0) versus Acetic Acid (AHQ x almost entirely due to a shift in the autoprotolysis equilibrium,
0)-Figures 1, 2, and 3 show the difference between the pH while in the triethylamine curves (Figures 1and 2), the shift
at 25 and 60 "C versus pH a t 25 "C for 0.1 M triethylamine, in the autoprotolysis equilibrium is clearly observable only
0.001 M triethylamine, and 0.1 M acetic acid, respectively. at very high pH (> 10) and results in a second inflection point
PHTEMP requires input of the average ion size (ail for in the curve.

Journal of Pharmaceutical Sciences / 1349

Vol. 84, No. 11, November 1995
 :::i
0
- -+. pK 1.W
'B
p-0.2.
g -0.4 :' 1
5-0.6 \
-o.8
-1 ~
0 2 4 6 8 1
pH at 25 Degrees C
Figure 4-ApH versus pH(25 "C) for 0.1 M sulfuric acid, a = 3 x 10V cm,
output from PHTEMP.
(b) Sulfuric Acid (Al?< 0)-Figure 4 depicts a constant-
volume titration of 0.1 M sulfuric acid, illustrating the effect
on the solution pH of raising the temperature from 25 to 60
"C. Acid dissociation of bisulfate ion is exothermic (AP =
-5.4 kcal/mol, from Table 1) and is expected to cause an
increase in pH with temperature. Thus, a linear plateau (ApH
> 0 a t pH 2-6) in the pH(T+ApH profile is observed when
the decrease in hydronium ion concentration upon equilibrium
shift is significant compared to the initial hydronium ion
present. At alkaline pH (>6) the shift in K , leads to an
inflection point in the curve and finally a second plateau (pH
> 8) as the bisulfate-sulfate equilibrium no longer contributes the opposing signs of W 1 ( 1 . 0 kcal/mol) and AH"z (0.58 kcal/
to the change in pH.
(c) Mathematical Confirmation of PHTEMP Output-A
simple equation, eq 6a, that calculates the derivative D p ~ solution drug products (3-7), PHTEMP illustrates that the
(dpWdT) for monoprotic acids is derived in the Appendix. This
equation can be extended in eq 7a to apply to solutions
containing more than one monoprotic acid. From D p ~one , simulation shows that the addition of a n amine-type buffer
can approximate ApH for a finite temperature differential (eq
8a). We have used eqs 6a and 8a to validate PHTEMP for
single-component systems. Figures 1-3 compare the results
from this simple model with the output of PHTEMP. The
graphs show good agreement between the predicted equation
and PHTEMP. There is a noticeable difference, however,
between the two results as they approach the corresponding
inflection points. The discrepancy in this region was antici-
pated, since equation 8a is a finite difference equation that
approximates an integral expression. The error in eq 8a
should increase as the derivative in the ApH-pH(T1) plot
increases. PHTEMP solves a system of nonlinear equations
and thereby more accurately determines the "true" (inte-
grated) result.
Solution pH-Temperature Effects for Triethylamine
in the Presence of TRIS and Citric Acid Buffers-
Pharmaceutical formulations often use buffers to enhance
stability over a product's shelf life. Considerations for the
proper choice of formulation buffer normally would include
formulation compatibility, buffer capacity, and pK,. Even
though testing a t 25 "C and elevated temperatures may be
typically planned, the pH-temperature sensitivity of buffers
and other excipients is not a routine consideration. PHTEMP
may be used to simulate the pH-temperature sensitivity of
a buffered formulation. Because many commercial drug
products contain drug substances with a n amine functionality
(e.g., procaine, lidocaine), triethylamine (Al?= 10.4 kcal/mol) temperature. We were interested in examining potential drug
was chosen to represent a "prototype" drug for use in the
simulation experiment with PHTEMP. As representative
buffers, we chose citric acid and tris(hydroxymethy1)ami-
nomethane (TRIS) because of their widespread formulation
usage.
Figure 5 shows the results of using PHTEMP to model the
pH-temperature dependence of 0.1 M triethylamine between
25 and 60 "C and 0.1 M triethylamine in the presence of 0.05
M TRIS and 0.02 M citric acid. (See Table 2 for simulation
parameters.)
1350 / Journal of Pharmaceutical Sciences
Vol. 84, No. 11, November 1995
O':
0 1 2 1 4
o 2 4 E a i o i z w
pH at 25 Degrees C
Figure 5-ApH versus pH(25 "C) for (a) 0.1 M triethylarnine (O), (b) 0.1 M
triethylamine in 0.05 M TRIS (0), and (c) 0.1 M triethylamine in 0.02 M citric acid
(+). a = 3 x cm. Output from PHTEMP.
PHTEMP predicts the typical initial plateau at lower pH
and a second plateau at higher pH common to each of the
three curves. In general, the pH shifts in all three solutions
show the expected downward trends over the pH range
studied. An exception, however, occurs in the triethylamine/
citric acid solution, which shows a slight increase in pH shift
between pH 3 and 6. Around pH 6.0, the pH shift is very
slightly positive. This behavior is likely due t o the triple acid
dissociation of citric acid (pK, = 3.182, 4.761, and 6.396) and
mol) versus A P 3 (-0.80 kcal/mol).
Within a pH region commonly used for the formulation of
pH-temperature sensitivity of a 0.1 M triethylamine solution
is greatly dependent on the type of buffer present. The
(TRIS, 0.05 M) actually enhances pH-temperature sensitivity
over a large pH range, whereas addition of a carboxylic acid
buffer (citric acid, 0.02 M) dramatically decreases sensitivity.
This can be largely explained by comparing the acid dissocia-
tion enthalpies of both buffers. Because the pK, of TRIS
(8.075) is lower than triethylamine (10.72), addition of TRIS
to the triethylamine solution shifts the inflection point of the
curve to lower pH in accordance with eq 6a. Furthermore,
the large enthalpy of acid dissociation for TRIS (11.36 kcal/
mol) increases the pH-temperature sensitivity of the solution
from pH 4 to 11. In constrast, the low enthalpies of acid
dissociation for citric acid result in reduced pH-temperature
sensitivity from pH 5.5 to 9. The much higher pH-temper-
ature sensitivity in triethylamine with added TRIS versus
citric acid buffer at pH 8 is interesting because the buffer
capacity of a 0.05 M TRIS solution a t pH 8 is predicted to be
approximately 20-fold higher than that of a 0.02 M citric acid
solution at the same pH.
The opposing results of triethylamine in the presence of two
different buffer systems illustrate that pH-temperature
relationships may be difficult to predict in multiple component
solutions unless a computer simulation is applied.
Solution pH-Temperature Effects in Amino Acid
Formulations: Glycine and a Multicomponent Amino
Acid Mixture-PHTEMP is particularly useful for prediction
of pH changes in multicomponent formulations a t elevated
stability issues associated with equilibrium shifts in complex
multicomponent amino acid solutions. Amino acid formula-
tions are extensively used in the pharmaceutical industry as
nutritional supplements; therefore, a thorough understanding
of pH changes that result due to changes in temperature is
essential and of wide applicability,
Figures 6 and 7 show the results of using PHTEMP to
model the pH-temperature dependence of a single component
and a 12-component amino acid formulation, respectively.
Because glycine represents the simplest of amino acids, a 0.1
 Table 3-Comparison of Simulation Results with Experiment
0 PHTEMP PredictionaIb ExperimentalC
-0.2
0) 5,"C Tz,OC pH(&) pH(&) ApH pH(&) ApH

'
g-0.4

I,
-0.6

-0.8
23.5
23.1
55.2
60.1
4.34
3.95
0.1 M Glycine
4.28
3.87
-0.06
-0.08
4.35
3.94
+0.01
-0.01
23.3 60.2 4.18 4.10 -0.08 4.13 -0.05
-1 23.2 60.1 3.92 3.85 -0.07 3.89 -0.03
23.5 55.2 8.67 7.92 -0.75 7.91 -0.76
0 2 4 6 8 1 0 1 2 1 4 23.3 60.0 8.96 8.10 -0.86 8.13 -0.83
pH at 25 Degrees C 23.2 60.1 8.97 8.10 4.87 8.13 -0.84
Figure 6-Plot of ApH versus pH(25 "C)for 0.1 M glycine, a = 3 x cm 23.2 60.1 9.01 8.14 -0.87 8.15 -0.86
(+) and a = 7 x cm (O), output from PHTEMP. 0.69 M Amino Acid Mixture (Table 2)
23.4 60.0 4.26 4.1 8 -0.08 4.16 -0.10
23.2 59.9 6.25 5.67 -0.58 5.63 -0.62
23.5 59.9 8.82 7.98 -0.84 7.98 -0.84
0

0 -0.2
en
2 4.4
0
I
-0.6
-0.8
-1
0 2 4 6 6 1 0 1 2
pH at 25 Degrees C
Figure 7-Plot of ApH versus pH(25 "C) for mixture of 12 amino acids (from
Table Z),a = 3 x cm, output from PHTEMP.
M solution, the behavior of which could be easily understood,
was chosen for comparison with the multicomponent amino
acid mixture. The identity and concentration of amino acids
typify some multicomponent solutions that are commercially
available as parenteral nutrition products. A temperature
differential of 35 "C (25-60 "C) was again chosen in the
simulation. Table 2 summarizes the parameters used.
The most obvious differences between Figures 6 and 7 are
the inflection points and the slopes of the curves at the
corresponding inflection points. The gradual inflection in
Figure 7 may be attributed to a smoothing effect of the various
pKa values for the different amino acids. The 12 amino acids
possess pK2 values ranging from 6.02 (histidine) to 10.640
(proline). All possess substantial enthalpies of acid dissocia-
tion. The pK1 values for the amino acids (carboxylic acid
dissociation) show a considerably narrower range with rela-
tively smaller enthalpies of dissociation (0-1 kcaumol). The
total concentration of the amino acid mixture (0.69 M) is
greater than in the glycine formulation (0.1 M); thus, the
inflection point of the curve is predicted to shift to lower pH
(eq 6a), while increasing the negative slope at this point. The
inflection point of the amino acid mixture is observed to shift
t o lower pH compared with glycine alone; however, an
increased negative slope is not observed. This result suggests
that the range of pKa values of components as well as the total
acid-base concentration is important in determining the pH-
temperature sensitivity profile of a multicomponent solution.
The maximum pH change remains the same for either the
mixture or the single component amino acid, as was antici-
pated, since the enthalpies of the second acid dissociation for
all the amino acids except histidine are nearly equivalent
(approximately 10.5 kcavmol).
Laboratory Confirmation of Simulation Results-In
order to test the accuracy of PHTEMP, we experimentally
determined the pH change in solutions of glycine and a
prepared mixture of amino acids (see Table 2 for composition).
Table 3 shows the results of experiments conducted at various
pH values and the results from PHTEMP. In all cases the
agreement of PHTEMP with experimental work was excellent.
a All simulations run using an effective ion size of 3 8, (see the text). pH
adjusted with NaOH or HCI to match experimental pH at the lower temperature.
Calibration using standard buffers at the temperatures shown.
The minor differences in ApH between PHTEMP predictions
and experimental work were likely due to either systematic
error associated with temperature variability of the thermo-
1 4
stated temperature bath (k2 "C), error in pH measurement,
inaccuracies in the parameters (e.g., heats of dissociation and
pKa values) used in the simulation, or other factors that were
not considered in the model (e.g., solute association, effective
sizes of individual ionic species).
Conclusion
PHTEMP clearly shows that the effects of temperature on
equilibria can be significant. Therefore, during the formula-
tion of pharmaceutical solutions, care should be taken to
identify components suspected to possess high enthalpies of
dissociation. From this study we have identified amine
functional groups t o be of particular concern, and those
solutions containing components with this functionality should
be evaluated for pH-temperature sensitivity prior to under-
taking elevated-temperature experiments. In contrast, those
formulations containing carboxylic acids are expected to show
i n s i d c a n t effects of temperature in the common formulation
range (3-7) because of the typically small enthalpies of
dissociation. Factors that contribute to higMow enthalpies
of dissociation have been reviewed in various texts and will
not be repeated here.g
As expected, PHTEMP predicts that the sensitivity of pH
to temperature is dependent on the formulation pH. In
general, as solution pH is increased, a critical point is reached
(region about an inflection point) in which a further pH
increase results in a larger IApWATI. At low pH (pH < 4 in
Figure 2, for example), the solution is saturated with hydro-
nium ions and is unaffected by minor shifts in dissociation
equilibria that produce hydronium ions. At high pH the
opposite is true, and large pH differences may be observed
even with relatively small changes in temperature (e.g., 0.001
M triethylamine, AT = 35 "C, ApH = 0.9; Figure 2). The
magnitude of ApH above the critical point will be dependent
on the enthalpy change associated with the equilibrium
causing the pH shift, and a number of other factors that are
less intuitive (e.g., concentration, number of components,pKa,
ionic strength), each of which are taken into account by
PHTEMP.
Differences in pH are only one of several considerations
when interpreting accelerated stability data. When analyzing
the effects of temperature on formulation stability, the usual
caveats must always apply. The experimenter must exclude,

Journal of Pharmaceutical Sciences / 1351

Vol. 84, No. 11, November 1995
for example, potential changes in mechanism, phase changes, or
and changes in acid or base ionization of the drug and
excipients due to shifts in equilibria. The latter is important
in determining possible buffer effects in a general acid- or D,=- AWK
base-catalyzed reaction and in examining the effect of the ionic R P
distribution of the drug on the observed rate constant.
PHTEMP provides ion distribution estimates as well as the Likewise,
predicted pH as a function of temperature. This would enable,
for example, correlation between the observed rate constant
m & W
and ionic speciation of the drug at each temperature. Dw=-
PHTEMP cannot account for all possible experimental R P
conditions. Ion-pairing, hydration, and heat capacity changes
are not considered. Experimental confirmation of ApH is Thus
therefore always warranted when precise data are required.

Appendix D~H=-[ 1 ){ +
KCAHO } *K
a l n ( l 0 ) R P (K + [H+1I2 [H+12
(6a)

Let S equal the concentration of the counterion of the strong

acid or base added to adjust pH. For example, S = "a+] for Equation 6a can be extended to 7a to treat cases where n
NaOH. Species [X-] is the conjugate base of the drug at monoprotic acids are present:
formal concentration C , with an acid dissociation constant K.
From charge balance and equilibrium expressions, eq l a can
be written. Note that S may be positive or negative.

where

Upon differentiation of eq l a , S is eliminated because the

counterion concentration is invariant with temperature. There-
fore,

KC(DK+DH)
D, = -
( K + [H+U2
+ K +CDK
[H'I
K$H
-- +-Dw
[H+12 [H'I
(2a) For small AT (assuming enthalpies remain constant over An,

where ApH M D,,AT (84

In the approximate equation 8a, D p should

~ be calculated
using the average temperature in the interval AT, and the
corresponding Ki values a t the average temperature. Ki may
Collecting terms gives be estimated at any temperature using the van 't Hoff
equation,

References and Notes

1. For example: Levine, I. Physical Chemistry; McGraw-Hill: New
where York, 1978; BD 388-390.
2. (a) Perrin D.-D.; Dempsey, B. Buffers for p H and Metal Ion
KC K
W
+ [H'])' +-[H+I2
Control; Chapman and Hall: New York, 1974; p 18. (b) Perrin,
a=l+ D. D. Aust. J. Chem. 1964,17, 484-488. (c) Bates, R. Determi-
(K nation of pH, Theory and Practice; Wiley: New York, 1964; pp
112-117.
, = -log [H'].
Assuming [H+] = a ~ +pH Therefore, 3. Kipp, J. E. J . Chem. Ed. 1994, 71, 119-121.
4. A significant compliation of enthalpy and dissociation constant
data may be found: Smith, R. M.; Martel, A. E. Critical Stability
Constants; Plenum Press: New York, 1974-1977, 1982, Vol.
1-5.
5. (a) Klotz, I. M.; Rosenberg, R. M. Chemical Thermodynamics-
Basic Theory and Methods; BenjamidCummings: Menlo Park,
CA; p 442; (b) For a discussion of the limitations of the Debye-
Huckel equation, see ref 1, p 246.
6. Olofsson, G. J . Chem. Thermodyn. 1984,16, 39-44.
7. Olofsson, G.; Bergstrom, S. J . S o h . Chem. 1978,7 , 497-513.
By the van 't Hoff law 8. For a discussion on ion size, see ref 5, p 441.
9. Maskil, H. The Physical Basis of Organic Chemistry; Oxford
University Press: New York, 1985; Chapter 5.
JS9500924

1352 / Journal of Pharmaceutical Sciences

Vol. 84, No. 11, November 1995