Salt Effects On Caffeine Solubility, Distribution,

And Self-Association
AHMAD AL-MAAIEH, DOUGLAS R. FLANAGAN
Pharmaceutics Division, College of Pharmacy, University of Iowa, Iowa City, Iowa 52242
Received 5 June 2001; revised 12 September 2001; accepted 3 October 2001
ABSTRACT: In this investigation, salt effects on monomeric solubility and distribution
are separated from self-association for caffeine. For self-associating compounds, the
Setschenow equation is inadequate because it does not separate salt effects into their
different contributions. Solubilities of caffeine, theophylline, and theobromine were
determined in water and salt solutions at 258C. Caffeine, theophylline, and theobromine
solubilities decreased with added Na
2
SO
4
or NaCl (i.e., salting-out) and increased with
added NaClO
4
or NaSCN (i.e., salting-in). Caffeine distribution coefcients (D
W/O
) also
decreased with added Na
2
SO
4
or NaCl and increased with added NaClO
4
or NaSCN.
To separate saltcaffeine effects from salt effects on caffeine self-interaction, salting
parameters (k
s
) were calculated fromD
W/O
at innite dilution instead of solubilities with
the Setschenow equation. Caffeine k
s
values were smaller than the Setschenow con-
stants (K) indicating that, for caffeine, K is not simply a salting-in/out parameter.
Distribution data were used to characterize caffeine self-association using either a
dimerization model (k
d
, dimerization constant) or an isodesmic model (k
iso
, stepwise
association constant). Caffeine self-association constants (k
d
or k
iso
) decreased with
NaClO
4
or NaSCN and increased with Na
2
SO
4
or NaCl. 2002 Wiley-Liss, Inc. and the
American Pharmaceutical Association J Pharm Sci 91:10001008, 2002
Keywords: salting-in/out; Hofmeister; kosmotrope; chaotrope; nonelectrolyte activity
coefcient; Setschenowequation; salting parameter; self-interaction/association; isodes-
mic model; free energy relationship
INTRODUCTION
Solubility and distribution phenomena of phar-
maceutical compounds are important to under-
stand drug behavior in solutions and biological
systems.
1
Moreover, many processes pertaining
to drug formulation or application take place in
the presence of varying concentrations of different
salts. Thus, the effect of salts on such phenomena
is important.
Xanthines are pharmaceutically important
compounds whose partition behavior and ability
to self-associate or to form molecular complexes
with other compounds have been extensively
investigated.
24
However, salt effects on their
solution properties have not been well studied.
In this work, the solubilities of three xanthines
(caffeine, theophylline, and theobromine) in dif-
ferent salt solutions were investigated. These
xanthines, at equilibrium with their saturated
solutions, have the same chemical potentials (m
i
)
in both phases.
5
Thus, caffeine has the same
chemical potential in the solid phase (m
i
solid
) as in
the saturated aqueous solution (m
i
aq
). This will
also be true if the saturated caffeine solution
contains added salt (i.e., m
i
solid
=m
i
salt
, where m
i
salt
is
caffeine's chemical potential in a salt-containing
saturated solution). If the salt does not change
m
i
solid
, then, at equilibrium, m
i
aq
=m
i
salt
. Further-
more, because m
i
=m
o
RTln a, then, m
i
aq
=m
o
aq

RTln a
aq
:m
i
salt
=m
o
salt
RTln a
salt
(m
o
is the stan-
dard reference chemical potential of caffeine, R is
1000 JOURNAL OF PHARMACEUTICAL SCIENCES, VOL. 91, NO. 4, APRIL 2002
Correspondence to: Douglas R. Flanagan (Telephone: 319-
335-8827; Fax: 319-335-9349;
E-mail: douglas-anagan@uiowa.edu
Journal of Pharmaceutical Sciences, Vol. 91, 10001008 (2002)
the gas constant, T is absolute temperature, a is
activity, and the superscripts salt or aq refer to
aqueous solutions with or without added salt, re-
spectively). Thus, at saturation, caffeine will have
equivalent activities in water and in salt solutions
at a given temperature if m
o
aq
=m
o
salt
.
6,7
Because
activities are products of concentrations and acti-
vity coefcients (g), salts may change the aqueous
solubility of xanthines by changing their aqueous
activity coefcients. This change may be directly
related to the effects of a salt on the aqueous envi-
ronment, which are different among various salts.
Inorganic salts were rst described by Hofme-
ister to have a systematic effect on solubility.
8
Salts were also found to have a systematic effect
on the behavior of aqueous solutions and were
divided into kosmotropes (polar water-structure
makers) or chaotropes (water-structure break-
ers).
9
This distinction was based on the degree to
which ions would interact with adjacent water
molecules. Akosmotrope like a doubly charged ion
(e.g., SO
4
2
) or an ion with a high charge density
(e.g., F

) was proposed to interact with adjacent

water molecules more strongly than would bulk
water. On the other hand, a chaotrope like a
large ion with a single charge (e.g., ClO
4

or SCN

)
was proposed to interact with adjacent water
molecules less strongly than would bulk water.
9
Arepresentative order of the Hofmeister series for
anions fromkosmotrope to chaotrope (left to right)
based on elution froma Sephadex G-10 column is:
9
SO
2
4
~ HPO
2
4
> F

> Cl

> Br

> I

(~ ClO

4
) > SCN

Salt effects on caffeine distribution between

aqueous and organic phases were also character-
ized. Distribution data were used in these studies
to separate salting parameters from caffeine self-
interaction parameters. Caffeine self-association
in different aqueous media was analyzed using
dimerization or isodesmic models. An isodesmic
model assumes stepwise association with equal
equilibrium constants and simplies the mathe-
matical analysis. Even though isodesmic modeling
has been extensively used in other disciplines,
10
it has found only a few applications in the phar-
maceutical literature.
11
The salts used in this study (sodium salts of
sulfate, chloride, bromide, thiocyanate, and per-
chlorate) were chosen to show that different salts
may have various, and in some cases, opposite
effects. Thus, this work shows the value of using
different salts to characterize solution properties
of nonelectrolytes rather than relying on only one
(e.g., NaCl, KCl, etc.). Also, this work provides a
resolution of salt effects into contributions due to
the monomeric solute activity and self-association
of caffeine. These effects are often erroneously
referred to as simple Setschenow constants when,
in fact, they involve more than one solution
phenomenon.
MATERIALS AND METHODS
Materials
Caffeine, theophylline, and theobromine were
purchased from Sigma (St. Louis, MO). Isooctane,
chloroform, NaCl, NaBr, NaSCN, NaClO
4
, and
Na
2
SO
4
were all of ACS grade.
Solubility
Excess xanthine was added to ~7 mL of either
distilled water (DW) or salt solution (NaCl, NaBr,
NaSCN, NaClO
4
, or Na
2
SO
4
at 0.11.0 M) in
screw-capped glass tubes. The tubes were rotated
in a water bath at 258C for 2448 h to achieve
equilibrium. The suspension was centrifuged and
the supernatant liquid ltered through a 0.45 mm
Gelman lter (Nylon Acrodisc
1
). Saturated solu-
tions were then volumetrically diluted in DW and
their ultraviolet (UV) absorbance measured at
273 nm(HP8450A photodiode array spectrophoto-
meter). The solubility was then calculated from
the standard curve of a particular xanthine.
Partitioning
A 20.0-mL aqueous aliquot containing varying
caffeine concentrations was mixed with 20.0 mL
of the organic phase (isooctane:chloroform, 9:1, v/v)
at 258C. After equilibration, a sample of the orga-
nic phase was obtained and caffeine concentration
was determined by UV analysis (273 nm). Caffe-
ine concentration in the aqueous phase was then
calculated from mass balance.
RESULTS AND DISCUSSION
Solubility
Caffeine solubility in different salt solutions is
shown in Figure 1. Different salts have different
effects on solubility: added NaClO
4
or NaSCN
increased caffeine solubility, whereas added
Na
2
SO
4
or NaCl decreased it, and added NaBr
SALT EFFECTS ON CAFFEINE 1001
JOURNAL OF PHARMACEUTICAL SCIENCES, VOL. 91, NO. 4, APRIL 2002
did not show signicant effects. The effects of
these salts on the aqueous solubility of theophyl-
line and theobromine are similar (i.e., solubility
increased with added NaClO
4
or NaSCN, decre-
ased with added Na
2
SO
4
or NaCl, and changed
little with added NaBr).
It can be seen in our work, that chaotropes
(e.g., ClO
4

) increase xanthine solubility whereas

kosmotropes (e.g., SO
4
2
) decrease it. Because all
the investigated salts in this study were sodium
salts, the differences in salt effects were attri-
buted to their anions.
The solubility data were treated using the
empirical Setschenow equation:
log
S
o
S

= KC
s
(1)
where S
o
is the nonelectrolyte (i.e., xanthine)
solubility in water, S is its solubility in a salt
solution, K is the Setschenow constant, and C
s
is
salt concentration.
12
Figure 2 is a Setschenow
plot for caffeine in which Kvalues are the slopes of
the lines. The K values for all xanthines (Table 1)
are positive for salts that decrease solubility
(i.e., salting-out) and negative for salts that incre-
ase it (i.e., salting-in).
A plot of K values versus intrinsic solubility
(i.e., S
o
; Figure 3) shows that K constants are
larger for xanthines that are more water-soluble.
This result means that the more soluble caffeine
(S
o
&21 mg/mL) is more sensitive to added salt
(either salting-in or -out) than the least soluble
theobromine (S
o
&0.44 mg/mL). This relation
between K and S
o
for xanthines is interesting
because less polar solutes are generally more
affected by salts in aqueous solutions.
13,14
How-
ever, the intrinsic aqueous solubilities of the
investigated xanthines are not solely based on
their polarities and interactions with water, but
are also affected by the crystal lattice energies
and the melting points of their respective solid
phases. In caffeine, the three nitrogen atoms at
positions 1, 3, and 7 are methylated, whereas only
two of these positions are methylated in theobro-
mine and theophylline. Thus, theobromine and
theophylline have higher melting points (3578C
Figure 1. Caffeine solubility in different salt solu-
tions at 258C. Key: (*) NaClO
4
; (*) NaSCN; (!) NaBr;
(!) NaCl; (&) Na
2
SO
4
.
Table 1. Setschenow K Constants (M
1
) for Xanthines in Different Salt Solutions at
258C
Salt Caffeine Theophylline Theobromine
Na
2
SO
4
0.760.01
a
0.47 0.02 0.420.003
NaCl 0.17 0.007 0.11 0.01 0.056 0.005
NaBr 0.027 0.006 0.0015 0.005 0.052 0.004
NaSCN 0.47 0.02 0.33 0.02 0.35 0.01
NaClO
4
0.55 0.03 0.38 0.01 0.37 0.02
a
Standard deviation.
Figure 2. Setschenow plots for caffeine in different
salt solutions at 258C. Key: (*) NaClO
4
; (*) NaSCN;
(!) NaBr; (!) NaCl; (&) Na
2
SO
4
.
1002 AL-MAAIEH AND FLANAGAN
JOURNAL OF PHARMACEUTICAL SCIENCES, VOL. 91, NO. 4, APRIL 2002
for theobromine and 2702748C for theophylline)
and lower intrinsic aqueous solubilities than caf-
feine (mp, 2388C) because the extra methyl group
in caffeine decreases the extent of intermolecular
interaction in the solid phase, resulting in a lower
crystal energy (also, lower melting point) and a
higher aqueous solubility. Thus, the less polar
caffeine is more affected by salts in aqueous solu-
tions even though its intrinsic aqueous solubility
is higher than that of the other two compounds.
The relationship between K and S
o
is sometimes
seen for drug salts for which the Setschenow eq-
uation has been inappropriately applied. Solubi-
lity product relationships and the DebyeHuckel
equation should be used to describe common and
uncommon ion equilibria in these cases.
15
Setschenow K constants are salting constants,
if S
o
is low or in the absence of nonelectrolyte
self-interactions. Otherwise, the applicable ex-
pression is:
13,16
log
g
g
o

= log
S
o
S

= k
s
C
s
k
i
S S
o
( ) (2)
where g is the nonelectrolyte activity coefcient
in salt solution, g
o
is nonelectrolyte activity
coefcient in water, and S
o
, S, and C
s
are as
already dened, and k
s
and k
i
are salting
and nonelectrolyte self-interaction parameters,
respectively. Thus, only for low S
o
or in the ab-
sence of nonelectrolyte self-interaction will the
Setschenow constant (K) be equal to the salting
parameter (k
s
). It should be noted that the
Setschenow equation often holds even if these
conditions are not met, but in this case, the em-
pirical Setschenow constant (K) would not equal
the salting parameter (k
s
).
13,16
Equation 2 indicates that in order for solubility
to decrease (i.e., salting-out), salts increase aq-
ueousactivitycoefcients (e.g., withSO
4
2
, akosmo-
trope). This increase may be explained by such
ions binding more water, which decreases the free
water available for solvating the nonelectrolyte.
Alternately, such salts may increase solvent struc-
ture, making the aqueous environment enthalpi-
cally or entropically less favorable for solute
insertion. Salting-in, on the other hand, involves
decreasing xanthine activity coefcients (e.g.,
with ClO
4

, a chaotrope). This behavior may be

related to such salts reducing the aqueous struc-
ture, making it entropically and/or enthalpically
more favorable for the nonelectrolyte insertion.
1
Distribution
The effect of different salt solutions on caffeine
distribution between aqueous and organic phases
is shown in Figure 4. Salts that increased caffeine
aqueous solubility (e.g., NaClO
4
and NaSCN) de-
creased caffeine partitioning. On the other hand,
Na
2
SO
4
and NaCl had the opposite effect on
caffeine solubility and partitioning. This result
may be explained by chaotropes (e.g., ClO
4

and
SCN

) decreasing aqueous caffeine activity coef-

cients, leading to lower caffeine activity, and
thus less drug partitioning into the organic phase
at a given concentration (Figure 4). On the other
hand, kosmotropes (e.g., SO
4
2
) increased activity
coefcients, leading to more partitioning. This
explanation is consistent with solubility data,
Figure 3. Relationship between Setschenow con-
stants (K) and xanthine intrinsic aqueous solubility.
Key: (*) NaClO
4
; (&) Na
2
SO
4
.
Figure 4. Caffeine distribution between aqueous
and organic phases at different caffeine concentrations.
Organic phase is isooctane:chloroform (9:1 by volume).
Aqueous phase: (*) 1.0 M NaClO
4
; (&) 0.50 M NaClO
4
;
(!) 0.25 MNaClO
4
; (^) water; (!) 0.25 M Na
2
SO
4
; (&)
0.50 M Na
2
SO
4
; (*) 1.0 M Na
2
SO
4
.
SALT EFFECTS ON CAFFEINE 1003
JOURNAL OF PHARMACEUTICAL SCIENCES, VOL. 91, NO. 4, APRIL 2002
which indicated similar trends for aqueous caf-
feine activity coefcients with salts (i.e., g incre-
asing with kosmotropes and decreasing with
chaotropes).
Distribution studies in water and salt solutions
were investigated to obtain k
s
values. Typically,
aqueous phase nonelectrolyte concentrations with
water or salt solutions (C

i
and C
i
, respectively)
are experimentally determined at equilibrium
with a given organic phase concentration (C
i
R
).
13
Because the nonelectrolyte chemical potential
(m
i
) in all phases are equal at equilibrium, and,
assuming the salts do not partition to the organic
phase and do not change the nonelectrolyte stan-
dard chemical potential (m
o
) in the aqueous phase,
the nonelectrolyte would have the same activity
in water and in a salt solution at C

i
and C
i
.
7
The
value of C
i
R
is usually chosen to be low enough
that nonelectrolyte self-interaction or self-asso-
ciation is negligible.
13
Values of k
s
would then be
determined from
log
g
g
o

= log
C
o
i
C
i

= k
s
C
s
(3)
and k
i
could be obtained using
k
i
=
log
S
o
S

k
s
C
s

S S
o
( )
(4)
For example, at C
i
R
=0.4 mM caffeine, k
s
values
with Na
2
SO
4
and NaClO
4
(from eq. 3) are 0.646
and 0.440, respectively. The absolute k
s
values
are smaller than those of the Setschenow K con-
stants (Table 1), illustrating that the Setschenow
K values for caffeine cannot be simply taken as
salting constants as has been reported.
17
Instead of using a nite concentration in the
organic phase (C
i
R
), distribution behavior for non-
electrolytes at innite dilution can be utilized.
For caffeine, distribution coefcients (D
W/O
) were
calculated and plots of D
W/O
versus organic phase
caffeine concentration in water or Na
2
SO
4
/NaClO
4
solutions were constructed (Figure 5). The inter-
cepts in Figure 5 (i.e., D
W/O
at innite dilution)
are listed in Table 2 and were used instead of C

i
and C
i
to calculate k
s
using eq. 3. For example,
such a plot for Na
2
SO
4
, in which the slope is k
s
is
shown in Figure 6. The k
s
values calculated
by this method are listed in Table 2 and these
values are slightly smaller and probably more
accurate than those calculated using a nite
caffeine concentration.
The self-interaction contribution can be neglec-
ted when the nonelectrolyte is not very soluble
(i.e., S
o
<10 mM).
16
Thus, for theobromine
(S
o
=2.4 mM), the Setschenow constants (K) can
be used as estimates for the salting parameter
(k
s
). Excellent linearity is seen when caffeine k
s
values were plotted versus those of theobromine
(Figure 7). This result suggests that a salting
order exists for these compounds because the
salting parameters for one xanthine correlate well
with those of another. This salting order clearly
indicates that even though different salts have
various effects on the solubility of a given com-
pound (with smaller or larger k
s
values that are
either positive or negative depending onsalt type),
these effects can be predicted for a structurally
related compound. Moreover, such a plot is a
linear free energy relationship because k
s
is rela-
ted to the free energy of transfer (DG
tr
) of a mole
of a nonelectrolyte from pure water to a salt solu-
tion of concentration C
s
by DG
tr
=RT ln(g/g8) =
2.303RTk
s
C
s
.
13
This relationship also allows the
calculation of DG
tr
from salting parameters.
The self-interaction parameter (k
i
) describes
how a nonelectrolyte affects its own activity coef-
cient in solution and is related to activity coef-
cient by log g =k
s
C
s
k
i
C
i
, where C
i
is the
nonelectrolyte concentration and the other terms
have the same meaning as before.
13
The k
i
para-
meters are generally negative and, thus, a non-
electrolyte generally decreases its escaping
tendency and increases its own solubility by self-
interaction, as expected. Moreover, a rapid drop
in g has been observed for compounds that asso-
ciate in solution and is accompanied by large
Figure 5. Caffeine distribution coefcients (D
W/O
) at
different organic phase caffeine concentrations. Aqu-
eous phase: (*) 1.0 M NaClO
4
; (&) 0.50 M NaClO
4
; (!)
0.25 M NaClO
4
; (^) water; (!) 0.25 M Na
2
SO
4
; (&)
0.50 M Na
2
SO
4
; (*) 1.0 M Na
2
SO
4
.
1004 AL-MAAIEH AND FLANAGAN
JOURNAL OF PHARMACEUTICAL SCIENCES, VOL. 91, NO. 4, APRIL 2002
negative values of k
i
(e.g., purine
18
). Caffeine [k
i
[
values calculated using eq. 4 decrease with NaClO
4
concentration. For example, k
i
=1.51, 1.02, or
0.50 M
1
at 0.25, 0.50, or 1.0 M NaClO
4
, respec-
tively. This result indicates less caffeine self-
interaction at higher NaClO
4
concentrations. On
the other hand, k
i
is more negative in Na
2
SO
4
solutions (e.g., k
i
=1.71 or 2.06 M
1
at 0.30 or
1.0 M Na
2
SO
4
, respectively).
Self-association
It can be seen (Figure 5) that D
W/O
is not constant
but rather a function of caffeine concentration,
because caffeine self-associates in aqueous solu-
tions but is monomeric in organic solutions.
2
Thus, at higher caffeine concentrations, a lower
caffeine fraction is monomeric, which is assumed
Figure 6. Aplot of log[(D
W/O
)
aqu
/(D
W/O
)
salt
] at innite
dilution versus Na
2
SO
4
concentration.
Table 2. Salt Effects on Caffeine Solubility, Distribution, and Self-Association at
258C
a
Salt
Salt
Conc. (M) D
W/O
b
k
s
(M
1
)
c
k
d
(M
1
)
d
k
iso
(M
1
)
e
Water 23.0 0.2 16.50.2 12.2 0.1
Na
2
SO
4
0.590 0.01
0.25 15.7 0.1 19.90.2 15.8 0.2
0.50 11.4 0.1 24.10.3 19.7 0.2
1.0 5.85 0.02 30.50.4 26.5 0.3
NaClO
4
0.417 0.05
0.25 34.0 0.4 11.80.4 8.63 0.2
0.50 41.9 0.3 10.20.1 7.610.07
1.0 61.8 0.4 7.57 0.06 5.760.05
NaSCN 0.363 0.005
1.0 52.9 0.3 8.930.1 6.54 0.1
NaCl 0.128 0.005
1.0 17.1 0.1 16.80.2 13.9 0.1
NaBr 0.005800.005
1.0 23.6 0.2 13.20.2 10.50.05
a
Values are reported standard deviations. For D
W/O
, k
d
, and k
iso
, n=916.
b
Distribution coefcient at innite dilution (y-intercepts in Figure 5).
c
Saltcaffeine interaction (salting) parameter. Linear regression was used to obtain k
s
for
Na
2
SO
4
and NaClO
4
. For NaSCN, NaCl, and NaBr, k
s
was calculated using data from 1.0 M salt
concentrations only.
d
Dimerization constant.
e
Isodesmic association constant.
Figure 7. Linear free energy relationship between
caffeine and theobromine salting parameters (k
s
). For
theobromine, Setschenow constants (K) are used as
estimates for k
s
.
SALT EFFECTS ON CAFFEINE 1005
JOURNAL OF PHARMACEUTICAL SCIENCES, VOL. 91, NO. 4, APRIL 2002
to be the form in equilibrium with the organic
phase, leading to increasing D
W/O
values at higher
caffeine concentrations.
Different approaches can be used to character-
ize caffeine self-association in aqueous solutions:
one involves different equilibrium constants for
each step of the association,
2
whereas an isodes-
mic model assumes a stepwise association with an
identical free energy and equilibrium constant for
each step of the aggregation.
10
The partitioning
data were analyzed by these two approaches to
calculate k
d
(dimerization constant using the rst
approach) and k
iso
(stepwise association constant
using the second approach). We do not intend to
indicate that dimerization alone describes caf-
feine association, because previous investigations
have shown that to be inaccurate,
19
but rather
values of k
d
are used to show that the trends
for salt effects on caffeine association are model
independent.
Caffeine association may be described by the
following equilibrium (n _ 2):
A
n1
A=
k
n
A
n
(5)
where A represents one caffeine molecule, A
n
is
a caffeine n-mer (e.g., dimer, trimer, etc.), and
k
n
is the association equilibrium constant. The
distribution coefcient is represented by the total
caffeine concentration in aqueous and organic
phases:
D
W=O
=
(A)
aqu
2(A
2
)
aqu
3(A
3
)
aqu
n(A
n
)
aqu
(A)
org
=
C
T
(A)
org
(6)
where C
T
represents the total caffeine concen-
tration in the aqueous phase at equilibrium. The
monomeric caffeine fraction in the aqueous phase
(a) is the ratio of aqueous monomeric caffeine con-
centration, (A)
aqu
, to C
T
:
a =
(A)
aqu
C
T
(7)
Thus, the distribution coefcient is related to the
monomer concentration ratio by:
D
W=O
=
1
a

(A)
aqu
(A)
org
(8)
The monomer concentration ratio (aqueous-
to-organic) is the true partition coefcient for
caffeine monomers, which will be the partition
coefcient at very low caffeine concentrations
where self-association is negligible. The inter-
cepts from D
W/O
plots versus caffeine concentra-
tion (Figure 5) are accurate estimates for the true
caffeine partition coefcients, leading to the fol-
lowing equation:
a
conc =j
=
lim
conc0
D
W=O
D
W=O
(9)
This equation enables us to calculate a at any
concentration ( j) in a model-independent fashion
from the distribution coefcients. (A)
aqu
can then
be calculated at different C
T
values using eq. 7.
In the isodesmic model, the equilibrium con-
stants are assumed to be equal; thus, k
2
=
k
3
= =k
n
=k
iso
, and k
iso
can be obtained by
nonlinear regression using eq. 10:
10
C
T
=
(A)
aqu
1 k
iso
(A)
aqu
h i
2
(10)
On the other hand, if only dimerization is con-
sidered, then
C
T
= (A)
aqu
2(A
2
)
aqu
= (A)
aqu
2k
2
(A)
2
aqu
(11)
and the dimerization constant k
d
(=k
2
) can be
obtained using:
C
T
=
1 a
2a
2

1
k
d

(12)
The k
d
and k
iso
values for caffeine in different
salt solutions are listed in Table 2. The values of
k
d
and k
iso
in water agree well with literature
values.
2,10,19,20
These equilibrium constants de-
crease with added NaClO
4
or NaSCN and in-
crease with added Na
2
SO
4
or NaCl. This result
is consistent with earlier reports that showed
increased stability of caffeine aggregates in KCl
solutions.
21
The isodesmic model can be applied to the
solubility data to calculate monomeric caffeine
fraction (a) at any caffeine concentration (C
T
)
using eq. 13:
10
a =
2k
iso
C
T
1

4k
iso
C
T
1

2(k
iso
C
T
)
2
(13)
The a values at saturation (i.e., at C
T
=S
o
or S)
are listed in Table 3. Monomeric caffeine concen-
trations at saturation (i.e., monomer solubilities)
were calculated by multiplying a values (Table 3)
by the solubility (S
o
or S). It can be seen that 23
55% of caffeine is monomeric and varies with the
added salt.
1006 AL-MAAIEH AND FLANAGAN
JOURNAL OF PHARMACEUTICAL SCIENCES, VOL. 91, NO. 4, APRIL 2002
Xanthines self-associate by stacking as a result
of being squeezed out by the aqueous solution.
14
Addition of Na
2
SO
4
or NaCl changes solution
properties such that addition of monomeric caf-
feine into solution is less favorable, resulting in
lower monomeric caffeine solubility (Table 3) and
providing a higher driving force towards stacking
or self-association. Thus, Na
2
SO
4
or NaCl incre-
ase association constants (k
d
and k
iso
, Table 2).
Chaotropes (NaClO
4
or NaSCN), on the other
hand, make the aqueous medium more favorable
for monomeric caffeine, decreasing the driving
force towards association resulting in lower asso-
ciation constants.
Even though added Na
2
SO
4
and NaCl increase
association constants (Table 2), they result in
higher monomeric caffeine fraction at saturation
(Table 3). On the other hand, added NaClO
4
and
NaSCN decrease association constants and result
in lower monomeric fraction at saturation. In
solubility studies, enough caffeine will dissolve
until the chemical potential of monomeric caffeine
equals that of the solid phase. Thus, at saturation,
as discussed before, monomeric caffeine activity
in pure water will be the same as that in salt
solutions. Salts will then either increase or decr-
ease monomeric caffeine solubility by changing
its activity coefcient. Added Na
2
SO
4
or NaCl
increase monomer activity coefcient, resulting in
lower monomeric solubility, whereas added NaClO
4
or NaSCN have the opposite effect. The extent
of caffeine self-association depends on both mono-
mer concentration and association constants.
With added Na
2
SO
4
or NaCl, the net effect of
larger association constants and lower monome-
ric solubility is a lower extent of association at
saturation, as indicated by a larger monomeric
fraction (Table 3). The net effect of added NaClO
4
or NaSCN, on the other hand, is a decrease in
monomeric fraction.
CONCLUSIONS
For solutions of self-associating compounds, add-
ed salts affect self-association as well as mono-
mer solubility and distribution. The empirical
Setschenow equation offers a general treatment
that does not separate the overall salt effects into
these different contributions. This is the rst
study that uses solution behavior at innite dilu-
tion to separate salt effects on monomers from
effects on self-association. Also, this study points
out how different inorganic salts have diverse
effects on nonelectrolyte properties. Kosmotropes
(e.g., Na
2
SO
4
or NaCl) destabilize monomeric
caffeine (i.e., increase its activity coefcient),
leading to salting-out and larger self-association
constants. Chaotropes, on the other hand, stabi-
lize caffeine monomers, leading to salting-in and
resulting in smaller self-association constants.
REFERENCES
1. Grant DJW, Higuchi T. 1990. Solubility behavior of
organic compounds. New York: Wiley-Interscience,
Chapter 8, pp. 355398.
2. Guttman D, Higuchi T. 1957. Reversible associa-
tion of caffeine and of some caffeine homologs in
aqueoussolution. JAmPharmAssocSci Ed46:410.
3. Thakkar AL, Tensmeyer LG, Wilham WL. 1971.
NMR evidence for self-association of theophylline
in aqueous solution. J Pharm Sci 60:12671269.
4. Higuchi T, Lach JL. 1954. Investigation of some
complexes formed in solution by caffeine. V. Inter-
actions between caffeine and p-aminobenzoic acid,
m-hydroxybenzoic acid, picric acid, o-phthalic acid,
Table 3. Salt Effects on Fraction and Concentration of Monomeric Caffeine at
Saturation (258C)
Salt Salt Conc. (M) Monomer Fraction Monomer Solubility (M)
Water 0.327 0.0350
Na
2
SO
4
0.25 0.373 0.0247
0.50 0.429 0.0175
1.0 0.546 0.0099
NaClO
4
0.25 0.306 0.0518
0.50 0.271 0.0630
1.0 0.233 0.0898
NaSCN 1.0 0.243 0.0776
NaCl 1.0 0.383 0.0273
NaBr 1.0 0.348 0.0390
SALT EFFECTS ON CAFFEINE 1007
JOURNAL OF PHARMACEUTICAL SCIENCES, VOL. 91, NO. 4, APRIL 2002
suberic acid, and valeric acid. J Am Pharm Assoc
Sci Ed 43:524530.
5. Klotz IR, Rosenberg RM. 1986. Chemical thermo-
dynamics, basic theory and methods, 4th edition.
Menlo Park: The Benjamin/Cummings Publishing
Company, Chapter 12, pp. 257270.
6. Denbigh K. 1981. The principles of chemical
equilibrium, 4th edition. Cambridge: Cambridge
University Press, Chapter 9, pp. 270291.
7. James KC. 1986. Solubility and related properties.
NewYork: Marcel Dekker, Inc., Chapter 1, pp. 159.
8. Hofmeister F. 1888. Zur Lehre von der Wirkung der
Salze. II. Arch Exp Pathol Pharmakol 24:247260.
9. Collins KD, Washabaugh MW. 1985. The Hofmeis-
ter effect and the behaviour of water at interfaces.
Quart Rev Biophys 18:323422.
10. Martin RB. 1996. Comparisons of indenite self-
association models. Chem Rev 96:30433064.
11. Charman WN, Lai CS, Finnin BC, Reed BL. 1991.
Self-association of nicotinamide in aqueous solu-
tion: Mass transport, freezing-point depression,
and partition coefcient studies. Pharm Res 8:
11441150.
12. Setschenow J. 1889. Uber die konstitution der
salzlosungen auf grund ihres verhaltens zu koh-
lensaure. Z Physik Chem 4:117125.
13. Long FQ, McDevit WF. 1952. Activity coefcients
of nonelectrolyte solutes in aqueous salt solutions.
Chem Rev 51:119169.
14. Yalkowsky SH. 1999. Solubility and solubilization
in aqueous media. New York: American Chemical
Society and Oxford University Press, Chapter 8, pp.
321396.
15. Bogardus JB. 1982. Common ion equilibria of
hydrochloride salts and the Setschenow equation.
J Pharm Sci 71:588590.
16. Gordon JE. 1975. The organic chemistry of electro-
lyte solutions. New York: John Wiley and Sons,
Chapter 1, pp. 1151.
17. Perez-Tejeda P, Maestre A, Balon M, Hidalgo J,
Munoz M, Sanchez M. 1987. Setschenow coef-
cients for caffeine, theophylline, and theobromine
in aqueous electrolyte solutions. J Chem Soc,
Faraday Trans I 83:10291039.
18. Ts'o POP, Melvin IS, Olson AC. 1963. Interac-
tion and association of bases and nucleosides in
aqueous solutions. J Am Chem Soc 85:1289
1296.
19. Lilley TH, Linsdell H. 1992. Association of caffeine
in water and in aqueous solutions of sucrose.
J Chem Soc, Faraday Trans 88:28652870.
20. Zahalka J, Donbrow M, Yanuka Y. 1993. Study
of self-association of 7-alkylxanthines by nuclear
magnetic resonance spectroscopy. J Chem Res (S):
351.
21. Cesaro A, Russo E, Tessarotto D. 1980. Thermo-
dynamics of caffeine in aqueous denaturant solu-
tions. J Solution Chem 9:221235.
1008 AL-MAAIEH AND FLANAGAN
JOURNAL OF PHARMACEUTICAL SCIENCES, VOL. 91, NO. 4, APRIL 2002