- Email: [email protected]
Osmotic effects of polyethylene glycol
GASTROENTEROLOGY
Osmotic Effects of Polyethylene LAWRENCE CAROL Department
R. SCHILLER,
A. SANTA
ANA,
of Internal Medicine,
MICHAEL
1988;94:933-41
Glycol
EMMETT,
and JOHN S. FORDTRAN Baylor University
Polyethylene glycol (PEG) has been used to increase the osmotic pressure of fluids used to cleanse the gastrointestinal tract. However, little is known about its osmotic activity. To investigate this activity systematically, solutions of PEG of differing molecular weights were made and subjected to measurement of osmolality by both freezing point depression and vapor pressure osmometry. Measured osmolality was increasingly greater than predicted from average molecular weight as PEG concentration increased. Measurement of sodium activity in NaCl/PEG solutions by means of an ion-selective electrode suggested that the higher than expected osmolality could be due in part to interactions that, in effect, sequestered water froti the solution. Osmolality was consistently greater by freezing point osmometry than by vapor pressure osmometry. To determine which osniorhetry method reflected biologically relevant osmolality, normal subjects underwent steady-state total gut perfusion with an electrolyte solution containing 105 g/L of PEG 3350. This produced rectal effluent that was hypertonic by freezing point osmometry but isotonic by vapor pressure osmometry. Assuming that luminal fluid reaches osmotic equilibrium with plasma during total gut perfusioh, this result suggests that the vapor pressure osmometer accurately reflects the biologically relevant osmolality of intestinal contents. We conclude that PEG exerts more of an osmotic effect than wotild be predicted from its molecular weight. This phenomenon may reflect interactions between PEG atid water molecules that alter the physical chemistry of the solution and sequester water from the solution.
olyethylene glycol (PEG) is a substance often used as a nonabsorbable marker in studies of gastrointestinal physiology. Its availability, ability to be measured in biological fluids, lack of toxicity, and miniscule absorption [when higher molecular weight polymers are used) make it a nearly ideal marker substance (1).In recent years, PEG has also
Medical Center. Dallas. Texas
been used as a component of gastrointestinal lavage solutions (2). In such solutions it is used as a nonabsorbable, nonmetabolizable solute to increase the osmotic pressure of the lavage fluid in order to minimize water absorption. In spite of this use there has been no systematic study of the osmotic activity of PEG. To investigate the osmotic behavior of PEG, we measured the osmolality of solutions of various species of PEG over a wide range of concentrations, using both freezing point and vapor pressure osmometers. In additional in vitro experiments we analyzed the effect of PEG on the physical chemistry of PEGiNaCl solutions by measuring sodium activity with an ion-selective electrode. Finally, we investigated the in vivo osmotic properties of a highconcentration PEG solution by means of intestinal perfusion. Materials
and Methods
Construction Solutions
of Test Solutions of polyethylene
glycol,
mannitol,
and
NaCl were made on a molal basis: a certain mass of the test substance was dissolved in ZOO g of deionized water, mixed with an electromagnetic stirrer for 15 min or until the solution was clear, and then subjected to measurement of osmolality. Polyethylene glycols of various molecular glycol weight ranges were used (Table 1). Polyethylene 400 and PEG 600 were obtained from Eastman Kodak Company (Rochester, N.Y.). Polyeihylene glycol 1450 and PEG 3350 were obtained from J. T. Baker Company (Jackson, Tenn.). For purposes of calculating molality, the average molecdlar weight of the PEG was used to determine the number of moles of PEG dissolved in water. Mannitol (mol wt 182.2) and NaCl (mol wt 58.4) were obtained from Mallinckrodt, Inc. (Paris, KY.).
P
Measurement Osmolality pression
of Osmolality was determined
on an Advanced
6 1988 by the American
Instruments
by freezing Osmometer,
Gastroenterological 0016-5085/88/$3.50
point
de-
model
Association
934
SCHILLER ET AL.
GASTROENTEROLOGY Vol. 94, No. 4
3DII (Needham Heights, Mass.). The stated range of this instrument is O-2000 mosmolikg. Prefreezing was common with very dilute solutions (~15 mosmol/kg), however, and results in this range of concentrations are not considered accurate. The reproducibility of duplicate samples of more concentrated solutions was ?3%. Osmolality was also measured with a Wescor 5500 Vapor Pressure Osmometer (Logan, Utah). The stated range of this instrument was 0-200~1 mosmol/kg. However, in preliminary experiments with salt solutions the response of this instrument was not linear below 50 mosmol/kg. The reproducibility of duplicate samples of more concentrated solutions was al%.
Sodium
Concentration
Measurements
Sodium concentration was measured by means of an ion-selective electrode system (Kodak Ektachem 400 Analyzer). This system uses a potentiometric technique to compare sodium activity in the test solution with a reference solution containing electrolytes and an agent to increase viscosity to approximately that of human serum. Sodium concentration is then automatically calculated for serum based on the activity coefficient for sodium and plasma water volume as estimated from the known displacement effects of solids in plasma. This instrument produces artifactually high results when measuring sodium concentrations in pure crystalloid solutions in which there are no plasma solids. In preliminary experiments with NaCl solutions made to yield concentrations of 70, 100, 120, 140, and 175 mmol/L, the concentrations reported by the ion-selective electrode were 77, 109, 134, 157, and 194 mmol/L, an average overestimation of 11%.
All electrode-measured sodium concentrations reported in this paper have been adjusted down to compensate for this artifact. Sodium concentrations were also measured by flame photometry.
Results Comparison of Osmolality With Equimolal Amounts Glycol
Measurements of Polyethylene
Figure 1 shows a plot of PEG concentration as calculated from the moles of solute added to 200 g of water versus osmolality measured by freezing point depression. Lines of identity are also shown. Solutions of mannitol, PEG 400, and PEG 600 closely follow the lines of identity in the range of 15-120 mosmol/kg. In contrast, observed osmotic activity is higher than predicted with PEG 1450 and PEG 3350. The difference between osmolality expected from the molality of the solution and the measured osmolality is progressively greater as these solutions are more concentrated. For example, at the highest concentration of PEG 3350 studied (400 g/kg or 119.4 mmol/kg), the measured osmolality is 1230 mosmoli kg, >lO times the theoretical osmolality. Figure 2 shows a plot of PEG concentration as calculated from the mass of solute in moles added to 200 g of water versus osmolality measured by the vapor pressure method. Solutions of mannitol and the two lower molecular weight PEG species closely follow the lines of identity at concentrations >50 mmol/kg water. At concentrations ~50 mmol/kg the vapor pressure osmometer does not record osmolality accurately (see Materials and Methods] and consistently overestimates the osmolality of these solutions. Osmolality with the two higher weight PEGS is higher than predicted from molality but not to as great an extent as with the freezing point osmometer. For instance, the largest variance is six times the theoretical osmolality. Comparison With Equal
of Osmolality Measurements Masses of Polyethylene Glycol
In Vivo Studies Total gut perfusion was used as previously described (2) to measure the osmolality of fluid after it had traveled through the intestine and had the opportunity to equilibrate with plasma osmotic pressure. Normal subjects were studied after giving written informed consent on forms approved by an Institutional Review Board. A PEGelectrolyte solution containing 105 g/L of PEG 3350, 2.8 g/L of NaCl (48 mM), 0.37 g/L of KC1 (5 mM), and 1.43 g/L of NaHCO, (17 mM) was infused into the stomach by a single-lumen gastric tube at a rate of 30 mlimin. A bolus of sulfobromophthalein was injected intragastrically at the start of perfusion and steady-state conditions were assumed after sulfobromophthalein had cleared from the rectal effluent. Perfusion was continued for 2 h thereafter and rectal effluent collected during the steady state was analyzed by both freezing point osmometry and vapor pressure osmometry.
In the preceding experiments, equal molal concentrations of PEG were compared. Because of the eightfold range of molecular weights of the different PEG species, there were large differences in the weights of added PEG and large differences in the
Table
I. Range of Molecular Weights and Polymer Lengths for Polyethylene GJycoJs Used in These Studies Range of
Species
PEG 400 PEG 600 PEG1450 PEG 3350 n, average number H(OCH2CHZ),0H.
molecular
weight
n 8.2-9.1 12.5-13.9 29-36 68-84
380-420 570-630 1300-1600 3000-3700 of ethylene
glycol
units
in polyethylene
glycol.
April
OSMOTIC
1988
EFFECTS
OF POLYE:THYLENE
GLYCOL
935
FREEZING POINT OSMOMETRY
PEG 1450
1200
PEG 3350
1200
0 I 1
900
900
I I
CONCENTRATION (mmolkg water) Figure
1. Osmolality measured lines represent lines
by freezing of identity.
point osmometry versus molal concentration for mannitol Note the scale change on the ordinate between the upper
total number of ethylene glycol subunits in the different solutions. To evaluate the possible effects of these differences, we performed a second series of experiments, in which equal masses (expressed on a
and four species of PEG. The solid and lower panels.
gram basis] of the different PEG species were added to 200 g of water and osmolality was measured with both the freezing point osmometer and the vapor pressure osmometer. The relative osmolality, de-
VAPOR PRESSURE OSMOMETRY
PEG 1450
PEG 3350
,. 600 900
t
600 900
t 4’
“Ot&&
“Oki 40
80
80
120
CONCENTRATION (mmolkg water) Figure
2. Osmolality measured by vapor pressure osmometry versus molal concentration for mannitol and four species of PEG. The solid lines represent lines of identity. Note the scale change in the ordinate between the upper and lower panels. The response of the vapor pressure osmometer is not linear below a concentration of 50 mmol/kg of HZO.
936
SCHILLER
GASTROENTEROLOGY
ET AL.
Vol. 94. No. 4
MANNITOL
NaCl
EXPECTED J(=* g::*::::::,+
L
Wb) (mmolkg)
I
I
3.7 7.3 62.5 125
*snu’g)
,,......,...#I
EXPECTED
:z*
I
14.6 250
I 29.2
22.8 45.6
500
125 250
91.1 500
182.2 1000
CONCENTRATION Figure
3. Relative osmolality (measured osmolality divided containing NaCl (left) and mannitol (right). Open symbols by vapor pressure osmometry.
by mass in grams symbols represent
fined as the measured osmolality divided by the concentration of solute [in grams per kilogram of water) was used to compensate for the fact that lower molecular weight PEGS should have more osmoles per gram. For a substance that behaves ideally, the increase in osmolality should be exactly proportional to the increase in concentration and the result of a plot of relative osmolality versus concentration should be a straight line parallel to the abscissa (slope = 0). The y-intercept of this line (the ratio at infinite dilution) is the inverse of the molecular weight. As shown in Figure 3, results with NaCl and mannitol solutions are similar for both freezing point and vapor pressure measurements and fall on a straight line parallel to the abscissa. The results for NaCl solutions fall below the line expected from the molecular weight of NaCl because NaCl is not fully dissociated at these concentrations. Mannitol solutions (right panel) fall almost precisely on the expected values. Results with the PEG solutions (Figure 4) fit linear regressions well (r > 0.98 in all cases). In contrast to NaCl or mannitol solutions, these regression lines have slopes >O, indicating that measured osmolality increases with the square of the concentration [slope = dyldx = (osmoles/C)/(C) = osmoles/C’, where C is the concentration] for each of the different PEG species. As shown in Table 2, the y-intercepts for these lines with either osmometry method are similar to the theoretical value, supporting the stated average molecular weights of these polymers. As
of solute added) versus concentration for solutions measurements by freezing point osmometry; closed
might be expected from the data presented in Figures 1 and 2 in which freezing point osmometry yielded higher osmolalities than vapor pressure osmometry, the slopes of the lines with freezing point osmometry are greater than the slopes with vapor pressure osmometry. For PEG 400, PEG 1450, and PEG 3350 the slopes are similar when using one method of measuring osmolality. This suggests that the increasing osmolality observed with the higher molecular weight PEG molecules is not unique to the large size of these molecules but is at least in part related to the mass of PEG or the number of ethylene glycol subunits present at a given molality, or both. The slopes obtained with PEG 600 were quantitatively different than those for the other PEG species. The reason for this difference is unknown. Measurement Polyethylene
of Sodium Concentrations Glycol 3350 Solutions
in
One possible explanation for a higher than predicted osmolality is that PEG forms complexes with the water in solution and makes the water less available for other interactions (i.e., decreases water activity). This would increase the activity of solutes dissolved in the remaining “free” water. To examine this possibility, solutions of PEG 3350 and NaCl were made and sodium concentration was measured with an ion-selective electrode as an estimate of sodium activity. Sodium activity is related to the amount of NaCl present, the extent of dissociation of NaCl, and the volume of water with which the
April 1988
OSMOTIC EFFECTS OF POLYETHYLENE
GLYCOL
937
PEG 600
PEG 400
. ..’
. . *A
,*** ,,**
,a*’
,,** *.
**.......‘
*’
_.a* *_. . . . ...1.‘” : .=.
I
50 63
rt EXPECTED (1.67mOsm’gPEG)
1 400 667
I
I
200 333
100 167
, . .A
CONCENTRATION
PEG 3350
PEG 1450
5-
.’
,o’
I’
,..*
..**’
a
p::.w* ;.’ hlnco) (mmohd
II
50 34
100 69
I
I
400 276
200 136
;o 15
l&l 30
*.**
EXPECTED rr (0.3mDm’g PEG)
2do 60
460 119
CONCENTRATKIN Figure 4. Relative osmolality (measured osmolality divided by mass in grams of solute added) versus concentration for various PEG species. Open symbols represent measurements by freezing point osmometry; closed symbols by vapor pressure osmometry. The dotted lines are regression lines calculated by the method of least squares.
sodium ions can interact. As the amount of NaCl is known and NaCl is substantially dissociated at these concentrations, the main determinant of sodium activity will be the volume of distribution of sodium. The composition and analysis of solutions of NaCl and PEG in water are shown in Table 3. Sodium concentrations in the PEG-free solution (solution A) as measured by flame photometry or electrode are similar to the ideal concentration expected from the amount of NaCl added (140 mmol/L). Sodium con-
centrations measured by flame photometry decrease slightly with the highest concentrations of PEG, perhaps because of viscosity changes. Addition of a small amount of PEG, which has only a negligible effect on vapor pressure osmolality (solution B), does not increase electrode-measured sodium concentration substantially, indicating that sodium activity is relatively unchanged. However, addition of larger amounts of PEG (solutions C and D) increases electrode-measured sodium concentrations by 25%-60%
938
Table
GASTROENTEROLOGY Vol. 94. No. 4
SCHILLER ET AL.
2. Characteristics
of Relative
Osmolality
Versus Concentration
Mannitol Intercepts (mosmollg) Theoretical
PEG 400
5.49 5.28 5.51
Vapor pressure osmometry Freezing point osmometry Slopes (mosmollg per g)
Theoretical
0 -0.0009 -0.00001
Vapor pressure osmometry Freezing point osmometry
of Aqueous
0.30 0.26 0.22
0 0.0034 0.0068
0 0.0015 0.0037
0 0.0043 0.0069
0 0.0040 0.0071
Solution B
C
D
140 0 997 259
140 5 993 262
140 100 913 390
140 200 829 678
140 141 140 993 4
141 144 141 972 21
134 177 153 791 122
128 230 169 609 220
PEG 3350
0.69 0.62 0.74
NaCUPEG 3350
A
PEG 1450
1.67 1.52 1.50
Solutions
Composition NaCl added (mmol) PEG 3350 added (g) HZ0 added to make 1 L (ml] Vapor pressure osmolality (mosmollkg) Sodium concentration (mmol/L) By flame photometry By ion-selective electrode” Expected for amount of water addedb Volume of distribution for sodium (ml)” Volume of water “bound” (mild
PEG 600
2.50 2.48 2.60
(Table 3), indicating that sodium activity has increased. As shown in Table 3, assuming that sodium dissolves in the total volume of water added, the sodium concentration expected for the amount of water added would rise as more PEG (and less water) is added (the displacement effect of PEG). However, only one-third of the observed increase in sodium concentration as measured by ion-selective electrode could be due to this effect. The fact that the increase in electrode-measured sodium concentration is greater than this is consistent with the notion that PEG increases sodium activity by forming complexes with water and preventing a fraction of the water from interacting with sodium. The size of this hypothetical fraction can be estimated by dividing the moles of sodium added by the electrode-measured sodium concentration. This gives volumes of distribution of 791 and 609 ml (per liter of solution) for solutions C and D, respectively, and suggests that substantial amounts of water (122 and 220 ml/L, respectively) are “bound” by the PEG and are unavailable for interaction with sodium in the solution.
Table 3. Properties
Curves
’ Reflecting sodium activity and adjusted for corrections make for measurement of sodium concentration in plasma (see Materials and Methods). b Calculated by dividing 140 mmol by the volume of HZ0 added to make 1 L. ‘Calculated by dividing 140 mmol by sodium concentration measured by ion-selective electrode. d Calculated by subtracting volume of distribution from Hz0 added to make I L.
If the concept of a bound fraction of water is correct, one should be able to predict the electrodemeasured sodium concentrations in solutions containing known masses of PEG and NaCl. To test this, two additional solutions containing 200 g/L of PEG 3350 and 70 or 108 mmol of NaCl were made. The sodium concentration that would be predicted to be measured by the ion-selective electrode was calculated by dividing the moles of NaCl added by 609 ml, the volume of distribution calculated for solution D (which also contained 200 g/L of PEG 3350). The predicted results are 115 and 177 mmol/L, respectively. Electrode-measured concentrations were 114 and 176 mmol/L, respectively, both in excellent agreement with the predicted results.
It is conceivable that the increase in electrodemeasured sodium concentration noted with PEG 3350 is an artifact caused by increasing osmolality per se. To examine this possibility an additional set of experiments analogous to those shown in Table 3 was performed. In these studies increasing amounts of mannitol were added to saline to produce a range of osmolalities similar to those measured for the PEG/saline solutions in Table 3. As shown in Table 4, electrode-measured sodium concentrations rose only slightly with increasing concentrations of mannitol, and this rise was not out of proportion to the decreasing amounts of water added. This suggests that the electrode measurements of sodium concentration are not affected by increasing osmolality per se and that, in contrast to PEG, mannitol sequesters very little water from the solution. In Vivo Total Gut Pe$usion As freezing point osmolality and vapor pressure osmolality differ in vitro, it is critically important to evaluate which (if either) represents the “true” value in vivo. To do this, we perfused the intestines of normal volunteers with an electrolyte solution containing a high concentration of PEG 3350. As the intestinal to water and electrolytes in osmotic equilibrium
mucosa is highly permeable (31, rectal effluent should be with plasma after traversing
April 1988
OSMOTIC EFFECTS OF POLYETHYLENE GLYCOL
the entire small intestine and colon and should have an effective osmolality of 290 mosmol/kg (the osmolality of plasma). The osmolalities of the perfusion solution as calculated on a theoretical basis and as measured by the vapor pressure osmometer and by the freezing point osmometer are shown in Table 5. On a theoretical basis (i.e., if the osmolality were due strictly to the number of ions and PEG molecules in solution), the infusate is markedly hypotonic to plasma. The infusate is also hypotonic as measured by vapor pressure osmometry but isotonic by freezing point osmometry. Results of this study are shown in Table 5. Freezing point osmolality of rectal effluent rises to a level that would be hypertonic to plasma. As there are no known mechanisms for this to occur (3) and as we are assuming that luminal fluid would be in osmotic equilibrium with plasma after traversing the entire intestine, it seems likely that the freezing point osmometer is giving an artifactually high measurement of osmolality. In contrast, the theoretical calculated osmolality of rectal effluent is improbably low. The calculated value of 180 ? 2 mosmol/kg for rectal effluent could not occur if equilibrium with plasma had been reached. However, vapor pressure osmometry yields a value that is both physiologically realistic and very close to that of plasma. It seems likely, therefore, that vapor pressure osmometry is giving an accurate measure of biologically effective osmolality .
Discussion Our studies indicate that PEG produces a greater osmotic effect than can be accounted for by the number of PEG molecules in solution. This is true for all species of PEG studied and the discrepancy seems to correlate with the mass of PEG (and
Table
4. Properties
of Aqueous
NaWMannitol
Solutions Solution
Composition NaCl added (mmol) Mannitol added (mmol) HZ0 added to make 1 L (ml) Vapor pressure osmolality (mosmollkg) Sodium concentration (mmollL) By flame photometry By ion-selectivr electrodea Expected for amount of water addedb
E
F
G
H
140 0 998 250
140 125 984 375
140 250 968 511
140 538 934 826
143 142 142 141 143 145 145 153 140 142 145 150
a Reflecting sodium activity and adjusted for corrections made for measurement of sodium concentration in plasma (see Methods]. b Calculated by dividing 140 mmol by the volume of Hz0 added to make 1 L.
939
Table 5. Theoretical and Measured Osmolalities (in milliosmoles per kilogram) of Infusate Effluent in 14 Total Gut Perfusion ExperimenW
Theoretical calculatedb Measured by freezing point Measured by vapor pressure
and
Infusate
Effluent
168 * 2 288 k 1 245 rt2
180 2 2 346 f 4 292 2 2
a Measured plasma osmolality [by freezing point osmometry) was 290 * 1 mosmol/kg in these subjects. b Calculated as 2 ([Na’] + [K+]) + [PEG in grams per liter]/3350. Electrolyte concentration increased slightly from infusate to effluent from 137 k 2 to 148 k 3 mmol/L (p < 0.005). The concentration of polyethylene glycol [PEG] was not significantly altered (105vs.109 k 2 g/L,p > 0.1).
the number of ethylene glycol subunits) added to make the solution. In every instance, measured osmolality is greater with the freezing point osmometer than with the vapor pressure osmometer. Our in vivo perfusion experiment extends these observations and indicates that the unexpectedly high osmolality of PEG solutions is not just a laboratory artifact but has real biological significance. When PEG-containing fluid is perfused through the intestine and permitted to equilibrate with plasma (osmolality = 290 mosmol/kg), the resulting concentrations of electrolytes and PEG combine to yield a theoretical calculated osmolality that is substantially hypotonic compared to plasma [Table 5). As the actual osmolality should be 290 mosmol/kg, the PEG solution must exert a higher osmotic pressure than this calculated value. On the other hand, the freezing point osmometer indicates that rectal effluent is hypertonic, an impossible situation (3). As the vapor pressure osmometer indicates that rectal effluent is isotonic as expected, the in vivo perfusion experiment suggests that the vapor pressure osmometer is indicating the “true” osmolality and is to be preferred for measurement of osmolality in solutions containing high concentrations of PEG. Several explanations for the greater than expected osmolality of PEG could be possible. The basis for the measurement of osmolality by the vapor pressure and freezing point osmometers is Raolt’s law: the equivalence between the fractional lowering of vapor pressure and the mole fraction of solute. Both osmometers use the change in phase transition temperature (dew point temperature or freezing point temperature) produced by the alteration in vapor pressure to measure osmolality. However, Raolt’s law is only an approximation. Both positive and negative deviations have been recognized (4-6). These deviations are usually only a few percent at most and occur at much higher molal fractions than used in our experiments with PEG. It is unlikely,
940
SCHILLER ET AL.
therefore, that this can explain the multifold deviation noted with these comparatively dilute PEG solutions. Another explanation for our results is an error in the stated molecular weight of PEG used to calculate osmolality. There is good agreement between the stated average molecular weights and the inverse of the y-intercepts of the relative osmolality versus concentration curves (Figure 4 and Table 2), suggesting that the stated average molecular weights are accurate. However, like most polymers, the molecular weight of commercially available PEG falls within a range. For instance, PEG 3350 includes molecules with molecular weights from 3000 to 3 700 (Table 1).As the solutions are constructed with a given weight of PEG and molality is calculated by dividing through by molecular weight, the error in calculated osmolality using a single midpoint value for molecular weight may be as much as 10%. Nonetheless, this error cannot explain the magnitude of the unexpectedly high osmolalities that we measured. Nor can it explain the increasing divergence between calculated and measured osmolality with increasing PEG concentration (Figures 1 and 2). Another possibility is contamination of PEG by some low molecular weight, osmotically active substance. If this occurred, one would expect a linear displacement from the line of identity because a fixed proportion of contaminant would be added per gram of PEG. As this is not observed (Figure 2), it is unlikely that PEG contamination explains the deviations observed. A more likely explanation is that PEG has noncolligative properties that produce the unexpected osmotic effects of PEG in aqueous solution. It is conceivable that PEG increases osmolality by altering the activity of water in solution. This could reduce the availability of water to interact with PEG and other solutes present in the solution (4-6)and would have the effect of dissolving the solute in less water, thereby increasing the chemical and osmotic activity of the solute. To examine whether this might occur with PEG, we measured sodium activity in a series of aqueous NaCUPEG 3350 solutions (Table 3). These results show that sodium activity as reflected by an ion-selective electrode rises disproportionately as PEG concentration rises. This result is consistent with the concept that high concentrations of PEG sequester water from solution and can affect the chemical and osmotic activity of other substances in solution. The extent to which this putative water sequestration effect accounts for the unexpectedly high measured osmolality of pure PEG solutions is not clear from our data. Given the volumes of distribution for sodium calculated in Table 3, the maximum increase
GASTROENTEROLOGY
Vol. 94,
No.4
in osmolality from this effect that would be predicted for a pure 200-g/L PEG solution would be somewhat less than twofold. In reality, vapor pressure osmolality is approximately fourfold greater than expected at this concentration of PEG 3350 (Figure 2). It seems likely, therefore, that different solutes (or combinations of solutes) have different water sequestration effects. For instance, when PEG is the sole solute present, the volume of distribution for PEG might be even smaller and the volume of water “bound” even greater than those noted when sodium was also present in the solution. One can calculate that a volume of distribution of just under 300 ml could account for the increase in measured osmolality observed with a pure 200-g/L PEG 3350 solution. These interactions are probably due to the physical structure of PEG molecules. Polyethylene glycol is a linear polymer made up of repeating two-carbonlength hydrocarbons joined by oxygen in an ether linkage (7). This allows for hydrogen bonding of the oxygens with surrounding water molecules. As the PEG molecules of higher molecular weight are long, averaging 29-36 ethylene glycol units in length for PEG 1450 and 68-84 units for PEG 3350 (7), they would be capable of ordering a relatively large region of water by hydrogen binding. To some extent the smaller PEG molecules appear to do the same when present in high enough concentration that the number of ethylene glycol subunits is similar to the longer PEG species (Table 2). This ordering would reduce vapor pressure by making it more difficult for water molecules to escape from the solution into the air above the solution. The reduction in vapor pressure would be detected as a change in dew point temperature in the vapor pressure osmometer and would be reported as an increase in osmolality as the equipment converts a reduction in dew point temperature to an increase in osmolality (4-6). A change in vapor pressure would also affect the freezing point temperature and would thereby alter the measurement of osmolality in the freezing point osmometer (4-6). It should be stressed that, as established by the in vivo perfusion experiment, this represents a true change in osmolality and not just a measurement artifact. This noncolligative osmotic behavior allows much lower concentrations of PEG 3350 to be used to increase osmotic pressure in lavage solutions than would otherwise be the case. As has already been mentioned, the freezing point osmometer gives consistently higher results than the vapor pressure osmometer both with our in vitro and in vivo experiments. The reason for this is uncertain. One possibility is that high concentrations of PEG inhibit ice crystallization, producing a further lowering of freezing point temperature and a higher
April 1988
measured osmolality. When ice crystalizes, it does so by adding a layer of water molecules to the plane Certain fish in face of the growing ice crystal (8-10). the Antarctic and Arctic regions have developed a series of glycopeptide molecules that inhibit ice crystallization, probably by binding onto the crystallization plane of the ice crystal (8-10). This chemical adaptation is important in allowing these fish to live in sea water that is often colder than the freezing point of ordinary body fluids. The molecular structure of these extended peptide chains resembles in some ways that of PEG. It is conceivable, therefore, that PEG molecules work in an analogous way to inhibit ice crystallization. This would lower the freezing point temperature and would be recorded as a higher osmolality in a freezing point osmometer but would not be a factor in the vapor pressure osmometer.
References 1. Fordtran
JS. Marker perfusion techniques for measuring intestinal absorption in man. Gastroenterology 1986;51:108993. 2 Davis GR, Santa Ana CA, Morawski SG, Fordtran JS. Development of a lavage solution associated with minimal water and electrolyte absorption or secretion. Gastroenterology 1980;78:991-5.
OSMOTIC EFFECTS OF POLYETHYLENE GLYCOL
941
3. Schultz SG. Salt and water absorption by mammalian small intestine. In: Johnson LR, ed. Physiology of the gastrointestinal tract. New York: Raven, 1981:983-9. 4. Briggs DR. Osmotic pressure measurements. In: Uber FM, ed. Biophysical research methods. New York: Interscience Publishers, 1950:39-66. 5. Kupke DW. Osmotic pressure. Adv Protein Chem 1960;15:57130. 6. Marshall AG. Chemical potential. In: Biophysical chemistry: principles, techniques and applications. New York: John Wiley & Sons, 1978:17-50. 7. The Merck index. Rahway, N.J.: Merck & Co., 1983:7439. 8. Duman JG, Patterson JL, Kozak JJ, DeVries AL. Isopiestic determination of water binding by fish antifreeze glycoproteins. Biochim Biophys Acta 1980;626:332-6. 9. DeVries AL. Biological antifreeze agents in cold water fishes. Comp Biochem Physiol 1982;73A:627-40, 10. Eastman JT, DeVries AL. Antarctic fishes. Sci Am 1986; 255:106-14.
Received July 29, 1987. Accepted November 9, 1987. Address requests for reprints to: Lawrence R. Schiller, M.D., Department of Internal Medicine, Baylor University Medical Center, 3500 Gaston Avenue, Dallas, Texas 75246. This work was supported by U.S. Public Health Service grant l-ROl-AM37172-02 from the National Institute of Arthritis, Metabolism, and Digestive Diseases; and the Southwestern Medical Foundation’s Abbie K. Dreyfuss Fund, Dallas, Texas. The authors thank Marcia Horvitz for help in preparing this manuscript and Peter A. Dysert II, M.D., for assistance with the chemical analysis.
Osmotic Effects of Polyethylene LAWRENCE CAROL Department
R. SCHILLER,
A. SANTA
ANA,
of Internal Medicine,
MICHAEL
1988;94:933-41
Glycol
EMMETT,
and JOHN S. FORDTRAN Baylor University
Polyethylene glycol (PEG) has been used to increase the osmotic pressure of fluids used to cleanse the gastrointestinal tract. However, little is known about its osmotic activity. To investigate this activity systematically, solutions of PEG of differing molecular weights were made and subjected to measurement of osmolality by both freezing point depression and vapor pressure osmometry. Measured osmolality was increasingly greater than predicted from average molecular weight as PEG concentration increased. Measurement of sodium activity in NaCl/PEG solutions by means of an ion-selective electrode suggested that the higher than expected osmolality could be due in part to interactions that, in effect, sequestered water froti the solution. Osmolality was consistently greater by freezing point osmometry than by vapor pressure osmometry. To determine which osniorhetry method reflected biologically relevant osmolality, normal subjects underwent steady-state total gut perfusion with an electrolyte solution containing 105 g/L of PEG 3350. This produced rectal effluent that was hypertonic by freezing point osmometry but isotonic by vapor pressure osmometry. Assuming that luminal fluid reaches osmotic equilibrium with plasma during total gut perfusioh, this result suggests that the vapor pressure osmometer accurately reflects the biologically relevant osmolality of intestinal contents. We conclude that PEG exerts more of an osmotic effect than wotild be predicted from its molecular weight. This phenomenon may reflect interactions between PEG atid water molecules that alter the physical chemistry of the solution and sequester water from the solution.
olyethylene glycol (PEG) is a substance often used as a nonabsorbable marker in studies of gastrointestinal physiology. Its availability, ability to be measured in biological fluids, lack of toxicity, and miniscule absorption [when higher molecular weight polymers are used) make it a nearly ideal marker substance (1).In recent years, PEG has also
Medical Center. Dallas. Texas
been used as a component of gastrointestinal lavage solutions (2). In such solutions it is used as a nonabsorbable, nonmetabolizable solute to increase the osmotic pressure of the lavage fluid in order to minimize water absorption. In spite of this use there has been no systematic study of the osmotic activity of PEG. To investigate the osmotic behavior of PEG, we measured the osmolality of solutions of various species of PEG over a wide range of concentrations, using both freezing point and vapor pressure osmometers. In additional in vitro experiments we analyzed the effect of PEG on the physical chemistry of PEGiNaCl solutions by measuring sodium activity with an ion-selective electrode. Finally, we investigated the in vivo osmotic properties of a highconcentration PEG solution by means of intestinal perfusion. Materials
and Methods
Construction Solutions
of Test Solutions of polyethylene
glycol,
mannitol,
and
NaCl were made on a molal basis: a certain mass of the test substance was dissolved in ZOO g of deionized water, mixed with an electromagnetic stirrer for 15 min or until the solution was clear, and then subjected to measurement of osmolality. Polyethylene glycols of various molecular glycol weight ranges were used (Table 1). Polyethylene 400 and PEG 600 were obtained from Eastman Kodak Company (Rochester, N.Y.). Polyeihylene glycol 1450 and PEG 3350 were obtained from J. T. Baker Company (Jackson, Tenn.). For purposes of calculating molality, the average molecdlar weight of the PEG was used to determine the number of moles of PEG dissolved in water. Mannitol (mol wt 182.2) and NaCl (mol wt 58.4) were obtained from Mallinckrodt, Inc. (Paris, KY.).
P
Measurement Osmolality pression
of Osmolality was determined
on an Advanced
6 1988 by the American
Instruments
by freezing Osmometer,
Gastroenterological 0016-5085/88/$3.50
point
de-
model
Association
934
SCHILLER ET AL.
GASTROENTEROLOGY Vol. 94, No. 4
3DII (Needham Heights, Mass.). The stated range of this instrument is O-2000 mosmolikg. Prefreezing was common with very dilute solutions (~15 mosmol/kg), however, and results in this range of concentrations are not considered accurate. The reproducibility of duplicate samples of more concentrated solutions was ?3%. Osmolality was also measured with a Wescor 5500 Vapor Pressure Osmometer (Logan, Utah). The stated range of this instrument was 0-200~1 mosmol/kg. However, in preliminary experiments with salt solutions the response of this instrument was not linear below 50 mosmol/kg. The reproducibility of duplicate samples of more concentrated solutions was al%.
Sodium
Concentration
Measurements
Sodium concentration was measured by means of an ion-selective electrode system (Kodak Ektachem 400 Analyzer). This system uses a potentiometric technique to compare sodium activity in the test solution with a reference solution containing electrolytes and an agent to increase viscosity to approximately that of human serum. Sodium concentration is then automatically calculated for serum based on the activity coefficient for sodium and plasma water volume as estimated from the known displacement effects of solids in plasma. This instrument produces artifactually high results when measuring sodium concentrations in pure crystalloid solutions in which there are no plasma solids. In preliminary experiments with NaCl solutions made to yield concentrations of 70, 100, 120, 140, and 175 mmol/L, the concentrations reported by the ion-selective electrode were 77, 109, 134, 157, and 194 mmol/L, an average overestimation of 11%.
All electrode-measured sodium concentrations reported in this paper have been adjusted down to compensate for this artifact. Sodium concentrations were also measured by flame photometry.
Results Comparison of Osmolality With Equimolal Amounts Glycol
Measurements of Polyethylene
Figure 1 shows a plot of PEG concentration as calculated from the moles of solute added to 200 g of water versus osmolality measured by freezing point depression. Lines of identity are also shown. Solutions of mannitol, PEG 400, and PEG 600 closely follow the lines of identity in the range of 15-120 mosmol/kg. In contrast, observed osmotic activity is higher than predicted with PEG 1450 and PEG 3350. The difference between osmolality expected from the molality of the solution and the measured osmolality is progressively greater as these solutions are more concentrated. For example, at the highest concentration of PEG 3350 studied (400 g/kg or 119.4 mmol/kg), the measured osmolality is 1230 mosmoli kg, >lO times the theoretical osmolality. Figure 2 shows a plot of PEG concentration as calculated from the mass of solute in moles added to 200 g of water versus osmolality measured by the vapor pressure method. Solutions of mannitol and the two lower molecular weight PEG species closely follow the lines of identity at concentrations >50 mmol/kg water. At concentrations ~50 mmol/kg the vapor pressure osmometer does not record osmolality accurately (see Materials and Methods] and consistently overestimates the osmolality of these solutions. Osmolality with the two higher weight PEGS is higher than predicted from molality but not to as great an extent as with the freezing point osmometer. For instance, the largest variance is six times the theoretical osmolality. Comparison With Equal
of Osmolality Measurements Masses of Polyethylene Glycol
In Vivo Studies Total gut perfusion was used as previously described (2) to measure the osmolality of fluid after it had traveled through the intestine and had the opportunity to equilibrate with plasma osmotic pressure. Normal subjects were studied after giving written informed consent on forms approved by an Institutional Review Board. A PEGelectrolyte solution containing 105 g/L of PEG 3350, 2.8 g/L of NaCl (48 mM), 0.37 g/L of KC1 (5 mM), and 1.43 g/L of NaHCO, (17 mM) was infused into the stomach by a single-lumen gastric tube at a rate of 30 mlimin. A bolus of sulfobromophthalein was injected intragastrically at the start of perfusion and steady-state conditions were assumed after sulfobromophthalein had cleared from the rectal effluent. Perfusion was continued for 2 h thereafter and rectal effluent collected during the steady state was analyzed by both freezing point osmometry and vapor pressure osmometry.
In the preceding experiments, equal molal concentrations of PEG were compared. Because of the eightfold range of molecular weights of the different PEG species, there were large differences in the weights of added PEG and large differences in the
Table
I. Range of Molecular Weights and Polymer Lengths for Polyethylene GJycoJs Used in These Studies Range of
Species
PEG 400 PEG 600 PEG1450 PEG 3350 n, average number H(OCH2CHZ),0H.
molecular
weight
n 8.2-9.1 12.5-13.9 29-36 68-84
380-420 570-630 1300-1600 3000-3700 of ethylene
glycol
units
in polyethylene
glycol.
April
OSMOTIC
1988
EFFECTS
OF POLYE:THYLENE
GLYCOL
935
FREEZING POINT OSMOMETRY
PEG 1450
1200
PEG 3350
1200
0 I 1
900
900
I I
CONCENTRATION (mmolkg water) Figure
1. Osmolality measured lines represent lines
by freezing of identity.
point osmometry versus molal concentration for mannitol Note the scale change on the ordinate between the upper
total number of ethylene glycol subunits in the different solutions. To evaluate the possible effects of these differences, we performed a second series of experiments, in which equal masses (expressed on a
and four species of PEG. The solid and lower panels.
gram basis] of the different PEG species were added to 200 g of water and osmolality was measured with both the freezing point osmometer and the vapor pressure osmometer. The relative osmolality, de-
VAPOR PRESSURE OSMOMETRY
PEG 1450
PEG 3350
,. 600 900
t
600 900
t 4’
“Ot&&
“Oki 40
80
80
120
CONCENTRATION (mmolkg water) Figure
2. Osmolality measured by vapor pressure osmometry versus molal concentration for mannitol and four species of PEG. The solid lines represent lines of identity. Note the scale change in the ordinate between the upper and lower panels. The response of the vapor pressure osmometer is not linear below a concentration of 50 mmol/kg of HZO.
936
SCHILLER
GASTROENTEROLOGY
ET AL.
Vol. 94. No. 4
MANNITOL
NaCl
EXPECTED J(=* g::*::::::,+
L
Wb) (mmolkg)
I
I
3.7 7.3 62.5 125
*snu’g)
,,......,...#I
EXPECTED
:z*
I
14.6 250
I 29.2
22.8 45.6
500
125 250
91.1 500
182.2 1000
CONCENTRATION Figure
3. Relative osmolality (measured osmolality divided containing NaCl (left) and mannitol (right). Open symbols by vapor pressure osmometry.
by mass in grams symbols represent
fined as the measured osmolality divided by the concentration of solute [in grams per kilogram of water) was used to compensate for the fact that lower molecular weight PEGS should have more osmoles per gram. For a substance that behaves ideally, the increase in osmolality should be exactly proportional to the increase in concentration and the result of a plot of relative osmolality versus concentration should be a straight line parallel to the abscissa (slope = 0). The y-intercept of this line (the ratio at infinite dilution) is the inverse of the molecular weight. As shown in Figure 3, results with NaCl and mannitol solutions are similar for both freezing point and vapor pressure measurements and fall on a straight line parallel to the abscissa. The results for NaCl solutions fall below the line expected from the molecular weight of NaCl because NaCl is not fully dissociated at these concentrations. Mannitol solutions (right panel) fall almost precisely on the expected values. Results with the PEG solutions (Figure 4) fit linear regressions well (r > 0.98 in all cases). In contrast to NaCl or mannitol solutions, these regression lines have slopes >O, indicating that measured osmolality increases with the square of the concentration [slope = dyldx = (osmoles/C)/(C) = osmoles/C’, where C is the concentration] for each of the different PEG species. As shown in Table 2, the y-intercepts for these lines with either osmometry method are similar to the theoretical value, supporting the stated average molecular weights of these polymers. As
of solute added) versus concentration for solutions measurements by freezing point osmometry; closed
might be expected from the data presented in Figures 1 and 2 in which freezing point osmometry yielded higher osmolalities than vapor pressure osmometry, the slopes of the lines with freezing point osmometry are greater than the slopes with vapor pressure osmometry. For PEG 400, PEG 1450, and PEG 3350 the slopes are similar when using one method of measuring osmolality. This suggests that the increasing osmolality observed with the higher molecular weight PEG molecules is not unique to the large size of these molecules but is at least in part related to the mass of PEG or the number of ethylene glycol subunits present at a given molality, or both. The slopes obtained with PEG 600 were quantitatively different than those for the other PEG species. The reason for this difference is unknown. Measurement Polyethylene
of Sodium Concentrations Glycol 3350 Solutions
in
One possible explanation for a higher than predicted osmolality is that PEG forms complexes with the water in solution and makes the water less available for other interactions (i.e., decreases water activity). This would increase the activity of solutes dissolved in the remaining “free” water. To examine this possibility, solutions of PEG 3350 and NaCl were made and sodium concentration was measured with an ion-selective electrode as an estimate of sodium activity. Sodium activity is related to the amount of NaCl present, the extent of dissociation of NaCl, and the volume of water with which the
April 1988
OSMOTIC EFFECTS OF POLYETHYLENE
GLYCOL
937
PEG 600
PEG 400
. ..’
. . *A
,*** ,,**
,a*’
,,** *.
**.......‘
*’
_.a* *_. . . . ...1.‘” : .=.
I
50 63
rt EXPECTED (1.67mOsm’gPEG)
1 400 667
I
I
200 333
100 167
, . .A
CONCENTRATION
PEG 3350
PEG 1450
5-
.’
,o’
I’
,..*
..**’
a
p::.w* ;.’ hlnco) (mmohd
II
50 34
100 69
I
I
400 276
200 136
;o 15
l&l 30
*.**
EXPECTED rr (0.3mDm’g PEG)
2do 60
460 119
CONCENTRATKIN Figure 4. Relative osmolality (measured osmolality divided by mass in grams of solute added) versus concentration for various PEG species. Open symbols represent measurements by freezing point osmometry; closed symbols by vapor pressure osmometry. The dotted lines are regression lines calculated by the method of least squares.
sodium ions can interact. As the amount of NaCl is known and NaCl is substantially dissociated at these concentrations, the main determinant of sodium activity will be the volume of distribution of sodium. The composition and analysis of solutions of NaCl and PEG in water are shown in Table 3. Sodium concentrations in the PEG-free solution (solution A) as measured by flame photometry or electrode are similar to the ideal concentration expected from the amount of NaCl added (140 mmol/L). Sodium con-
centrations measured by flame photometry decrease slightly with the highest concentrations of PEG, perhaps because of viscosity changes. Addition of a small amount of PEG, which has only a negligible effect on vapor pressure osmolality (solution B), does not increase electrode-measured sodium concentration substantially, indicating that sodium activity is relatively unchanged. However, addition of larger amounts of PEG (solutions C and D) increases electrode-measured sodium concentrations by 25%-60%
938
Table
GASTROENTEROLOGY Vol. 94. No. 4
SCHILLER ET AL.
2. Characteristics
of Relative
Osmolality
Versus Concentration
Mannitol Intercepts (mosmollg) Theoretical
PEG 400
5.49 5.28 5.51
Vapor pressure osmometry Freezing point osmometry Slopes (mosmollg per g)
Theoretical
0 -0.0009 -0.00001
Vapor pressure osmometry Freezing point osmometry
of Aqueous
0.30 0.26 0.22
0 0.0034 0.0068
0 0.0015 0.0037
0 0.0043 0.0069
0 0.0040 0.0071
Solution B
C
D
140 0 997 259
140 5 993 262
140 100 913 390
140 200 829 678
140 141 140 993 4
141 144 141 972 21
134 177 153 791 122
128 230 169 609 220
PEG 3350
0.69 0.62 0.74
NaCUPEG 3350
A
PEG 1450
1.67 1.52 1.50
Solutions
Composition NaCl added (mmol) PEG 3350 added (g) HZ0 added to make 1 L (ml] Vapor pressure osmolality (mosmollkg) Sodium concentration (mmol/L) By flame photometry By ion-selective electrode” Expected for amount of water addedb Volume of distribution for sodium (ml)” Volume of water “bound” (mild
PEG 600
2.50 2.48 2.60
(Table 3), indicating that sodium activity has increased. As shown in Table 3, assuming that sodium dissolves in the total volume of water added, the sodium concentration expected for the amount of water added would rise as more PEG (and less water) is added (the displacement effect of PEG). However, only one-third of the observed increase in sodium concentration as measured by ion-selective electrode could be due to this effect. The fact that the increase in electrode-measured sodium concentration is greater than this is consistent with the notion that PEG increases sodium activity by forming complexes with water and preventing a fraction of the water from interacting with sodium. The size of this hypothetical fraction can be estimated by dividing the moles of sodium added by the electrode-measured sodium concentration. This gives volumes of distribution of 791 and 609 ml (per liter of solution) for solutions C and D, respectively, and suggests that substantial amounts of water (122 and 220 ml/L, respectively) are “bound” by the PEG and are unavailable for interaction with sodium in the solution.
Table 3. Properties
Curves
’ Reflecting sodium activity and adjusted for corrections make for measurement of sodium concentration in plasma (see Materials and Methods). b Calculated by dividing 140 mmol by the volume of HZ0 added to make 1 L. ‘Calculated by dividing 140 mmol by sodium concentration measured by ion-selective electrode. d Calculated by subtracting volume of distribution from Hz0 added to make I L.
If the concept of a bound fraction of water is correct, one should be able to predict the electrodemeasured sodium concentrations in solutions containing known masses of PEG and NaCl. To test this, two additional solutions containing 200 g/L of PEG 3350 and 70 or 108 mmol of NaCl were made. The sodium concentration that would be predicted to be measured by the ion-selective electrode was calculated by dividing the moles of NaCl added by 609 ml, the volume of distribution calculated for solution D (which also contained 200 g/L of PEG 3350). The predicted results are 115 and 177 mmol/L, respectively. Electrode-measured concentrations were 114 and 176 mmol/L, respectively, both in excellent agreement with the predicted results.
It is conceivable that the increase in electrodemeasured sodium concentration noted with PEG 3350 is an artifact caused by increasing osmolality per se. To examine this possibility an additional set of experiments analogous to those shown in Table 3 was performed. In these studies increasing amounts of mannitol were added to saline to produce a range of osmolalities similar to those measured for the PEG/saline solutions in Table 3. As shown in Table 4, electrode-measured sodium concentrations rose only slightly with increasing concentrations of mannitol, and this rise was not out of proportion to the decreasing amounts of water added. This suggests that the electrode measurements of sodium concentration are not affected by increasing osmolality per se and that, in contrast to PEG, mannitol sequesters very little water from the solution. In Vivo Total Gut Pe$usion As freezing point osmolality and vapor pressure osmolality differ in vitro, it is critically important to evaluate which (if either) represents the “true” value in vivo. To do this, we perfused the intestines of normal volunteers with an electrolyte solution containing a high concentration of PEG 3350. As the intestinal to water and electrolytes in osmotic equilibrium
mucosa is highly permeable (31, rectal effluent should be with plasma after traversing
April 1988
OSMOTIC EFFECTS OF POLYETHYLENE GLYCOL
the entire small intestine and colon and should have an effective osmolality of 290 mosmol/kg (the osmolality of plasma). The osmolalities of the perfusion solution as calculated on a theoretical basis and as measured by the vapor pressure osmometer and by the freezing point osmometer are shown in Table 5. On a theoretical basis (i.e., if the osmolality were due strictly to the number of ions and PEG molecules in solution), the infusate is markedly hypotonic to plasma. The infusate is also hypotonic as measured by vapor pressure osmometry but isotonic by freezing point osmometry. Results of this study are shown in Table 5. Freezing point osmolality of rectal effluent rises to a level that would be hypertonic to plasma. As there are no known mechanisms for this to occur (3) and as we are assuming that luminal fluid would be in osmotic equilibrium with plasma after traversing the entire intestine, it seems likely that the freezing point osmometer is giving an artifactually high measurement of osmolality. In contrast, the theoretical calculated osmolality of rectal effluent is improbably low. The calculated value of 180 ? 2 mosmol/kg for rectal effluent could not occur if equilibrium with plasma had been reached. However, vapor pressure osmometry yields a value that is both physiologically realistic and very close to that of plasma. It seems likely, therefore, that vapor pressure osmometry is giving an accurate measure of biologically effective osmolality .
Discussion Our studies indicate that PEG produces a greater osmotic effect than can be accounted for by the number of PEG molecules in solution. This is true for all species of PEG studied and the discrepancy seems to correlate with the mass of PEG (and
Table
4. Properties
of Aqueous
NaWMannitol
Solutions Solution
Composition NaCl added (mmol) Mannitol added (mmol) HZ0 added to make 1 L (ml) Vapor pressure osmolality (mosmollkg) Sodium concentration (mmollL) By flame photometry By ion-selectivr electrodea Expected for amount of water addedb
E
F
G
H
140 0 998 250
140 125 984 375
140 250 968 511
140 538 934 826
143 142 142 141 143 145 145 153 140 142 145 150
a Reflecting sodium activity and adjusted for corrections made for measurement of sodium concentration in plasma (see Methods]. b Calculated by dividing 140 mmol by the volume of Hz0 added to make 1 L.
939
Table 5. Theoretical and Measured Osmolalities (in milliosmoles per kilogram) of Infusate Effluent in 14 Total Gut Perfusion ExperimenW
Theoretical calculatedb Measured by freezing point Measured by vapor pressure
and
Infusate
Effluent
168 * 2 288 k 1 245 rt2
180 2 2 346 f 4 292 2 2
a Measured plasma osmolality [by freezing point osmometry) was 290 * 1 mosmol/kg in these subjects. b Calculated as 2 ([Na’] + [K+]) + [PEG in grams per liter]/3350. Electrolyte concentration increased slightly from infusate to effluent from 137 k 2 to 148 k 3 mmol/L (p < 0.005). The concentration of polyethylene glycol [PEG] was not significantly altered (105vs.109 k 2 g/L,p > 0.1).
the number of ethylene glycol subunits) added to make the solution. In every instance, measured osmolality is greater with the freezing point osmometer than with the vapor pressure osmometer. Our in vivo perfusion experiment extends these observations and indicates that the unexpectedly high osmolality of PEG solutions is not just a laboratory artifact but has real biological significance. When PEG-containing fluid is perfused through the intestine and permitted to equilibrate with plasma (osmolality = 290 mosmol/kg), the resulting concentrations of electrolytes and PEG combine to yield a theoretical calculated osmolality that is substantially hypotonic compared to plasma [Table 5). As the actual osmolality should be 290 mosmol/kg, the PEG solution must exert a higher osmotic pressure than this calculated value. On the other hand, the freezing point osmometer indicates that rectal effluent is hypertonic, an impossible situation (3). As the vapor pressure osmometer indicates that rectal effluent is isotonic as expected, the in vivo perfusion experiment suggests that the vapor pressure osmometer is indicating the “true” osmolality and is to be preferred for measurement of osmolality in solutions containing high concentrations of PEG. Several explanations for the greater than expected osmolality of PEG could be possible. The basis for the measurement of osmolality by the vapor pressure and freezing point osmometers is Raolt’s law: the equivalence between the fractional lowering of vapor pressure and the mole fraction of solute. Both osmometers use the change in phase transition temperature (dew point temperature or freezing point temperature) produced by the alteration in vapor pressure to measure osmolality. However, Raolt’s law is only an approximation. Both positive and negative deviations have been recognized (4-6). These deviations are usually only a few percent at most and occur at much higher molal fractions than used in our experiments with PEG. It is unlikely,
940
SCHILLER ET AL.
therefore, that this can explain the multifold deviation noted with these comparatively dilute PEG solutions. Another explanation for our results is an error in the stated molecular weight of PEG used to calculate osmolality. There is good agreement between the stated average molecular weights and the inverse of the y-intercepts of the relative osmolality versus concentration curves (Figure 4 and Table 2), suggesting that the stated average molecular weights are accurate. However, like most polymers, the molecular weight of commercially available PEG falls within a range. For instance, PEG 3350 includes molecules with molecular weights from 3000 to 3 700 (Table 1).As the solutions are constructed with a given weight of PEG and molality is calculated by dividing through by molecular weight, the error in calculated osmolality using a single midpoint value for molecular weight may be as much as 10%. Nonetheless, this error cannot explain the magnitude of the unexpectedly high osmolalities that we measured. Nor can it explain the increasing divergence between calculated and measured osmolality with increasing PEG concentration (Figures 1 and 2). Another possibility is contamination of PEG by some low molecular weight, osmotically active substance. If this occurred, one would expect a linear displacement from the line of identity because a fixed proportion of contaminant would be added per gram of PEG. As this is not observed (Figure 2), it is unlikely that PEG contamination explains the deviations observed. A more likely explanation is that PEG has noncolligative properties that produce the unexpected osmotic effects of PEG in aqueous solution. It is conceivable that PEG increases osmolality by altering the activity of water in solution. This could reduce the availability of water to interact with PEG and other solutes present in the solution (4-6)and would have the effect of dissolving the solute in less water, thereby increasing the chemical and osmotic activity of the solute. To examine whether this might occur with PEG, we measured sodium activity in a series of aqueous NaCUPEG 3350 solutions (Table 3). These results show that sodium activity as reflected by an ion-selective electrode rises disproportionately as PEG concentration rises. This result is consistent with the concept that high concentrations of PEG sequester water from solution and can affect the chemical and osmotic activity of other substances in solution. The extent to which this putative water sequestration effect accounts for the unexpectedly high measured osmolality of pure PEG solutions is not clear from our data. Given the volumes of distribution for sodium calculated in Table 3, the maximum increase
GASTROENTEROLOGY
Vol. 94,
No.4
in osmolality from this effect that would be predicted for a pure 200-g/L PEG solution would be somewhat less than twofold. In reality, vapor pressure osmolality is approximately fourfold greater than expected at this concentration of PEG 3350 (Figure 2). It seems likely, therefore, that different solutes (or combinations of solutes) have different water sequestration effects. For instance, when PEG is the sole solute present, the volume of distribution for PEG might be even smaller and the volume of water “bound” even greater than those noted when sodium was also present in the solution. One can calculate that a volume of distribution of just under 300 ml could account for the increase in measured osmolality observed with a pure 200-g/L PEG 3350 solution. These interactions are probably due to the physical structure of PEG molecules. Polyethylene glycol is a linear polymer made up of repeating two-carbonlength hydrocarbons joined by oxygen in an ether linkage (7). This allows for hydrogen bonding of the oxygens with surrounding water molecules. As the PEG molecules of higher molecular weight are long, averaging 29-36 ethylene glycol units in length for PEG 1450 and 68-84 units for PEG 3350 (7), they would be capable of ordering a relatively large region of water by hydrogen binding. To some extent the smaller PEG molecules appear to do the same when present in high enough concentration that the number of ethylene glycol subunits is similar to the longer PEG species (Table 2). This ordering would reduce vapor pressure by making it more difficult for water molecules to escape from the solution into the air above the solution. The reduction in vapor pressure would be detected as a change in dew point temperature in the vapor pressure osmometer and would be reported as an increase in osmolality as the equipment converts a reduction in dew point temperature to an increase in osmolality (4-6). A change in vapor pressure would also affect the freezing point temperature and would thereby alter the measurement of osmolality in the freezing point osmometer (4-6). It should be stressed that, as established by the in vivo perfusion experiment, this represents a true change in osmolality and not just a measurement artifact. This noncolligative osmotic behavior allows much lower concentrations of PEG 3350 to be used to increase osmotic pressure in lavage solutions than would otherwise be the case. As has already been mentioned, the freezing point osmometer gives consistently higher results than the vapor pressure osmometer both with our in vitro and in vivo experiments. The reason for this is uncertain. One possibility is that high concentrations of PEG inhibit ice crystallization, producing a further lowering of freezing point temperature and a higher
April 1988
measured osmolality. When ice crystalizes, it does so by adding a layer of water molecules to the plane Certain fish in face of the growing ice crystal (8-10). the Antarctic and Arctic regions have developed a series of glycopeptide molecules that inhibit ice crystallization, probably by binding onto the crystallization plane of the ice crystal (8-10). This chemical adaptation is important in allowing these fish to live in sea water that is often colder than the freezing point of ordinary body fluids. The molecular structure of these extended peptide chains resembles in some ways that of PEG. It is conceivable, therefore, that PEG molecules work in an analogous way to inhibit ice crystallization. This would lower the freezing point temperature and would be recorded as a higher osmolality in a freezing point osmometer but would not be a factor in the vapor pressure osmometer.
References 1. Fordtran
JS. Marker perfusion techniques for measuring intestinal absorption in man. Gastroenterology 1986;51:108993. 2 Davis GR, Santa Ana CA, Morawski SG, Fordtran JS. Development of a lavage solution associated with minimal water and electrolyte absorption or secretion. Gastroenterology 1980;78:991-5.
OSMOTIC EFFECTS OF POLYETHYLENE GLYCOL
941
3. Schultz SG. Salt and water absorption by mammalian small intestine. In: Johnson LR, ed. Physiology of the gastrointestinal tract. New York: Raven, 1981:983-9. 4. Briggs DR. Osmotic pressure measurements. In: Uber FM, ed. Biophysical research methods. New York: Interscience Publishers, 1950:39-66. 5. Kupke DW. Osmotic pressure. Adv Protein Chem 1960;15:57130. 6. Marshall AG. Chemical potential. In: Biophysical chemistry: principles, techniques and applications. New York: John Wiley & Sons, 1978:17-50. 7. The Merck index. Rahway, N.J.: Merck & Co., 1983:7439. 8. Duman JG, Patterson JL, Kozak JJ, DeVries AL. Isopiestic determination of water binding by fish antifreeze glycoproteins. Biochim Biophys Acta 1980;626:332-6. 9. DeVries AL. Biological antifreeze agents in cold water fishes. Comp Biochem Physiol 1982;73A:627-40, 10. Eastman JT, DeVries AL. Antarctic fishes. Sci Am 1986; 255:106-14.
Received July 29, 1987. Accepted November 9, 1987. Address requests for reprints to: Lawrence R. Schiller, M.D., Department of Internal Medicine, Baylor University Medical Center, 3500 Gaston Avenue, Dallas, Texas 75246. This work was supported by U.S. Public Health Service grant l-ROl-AM37172-02 from the National Institute of Arthritis, Metabolism, and Digestive Diseases; and the Southwestern Medical Foundation’s Abbie K. Dreyfuss Fund, Dallas, Texas. The authors thank Marcia Horvitz for help in preparing this manuscript and Peter A. Dysert II, M.D., for assistance with the chemical analysis.