Chemosphere, Vol.21, Nos.l—2, pp 119-133, 1990
`Printed in Great Britain
`
`0045—6535/90 $3.00 + .00
`Pergamon Press plc
`
`AN INTERLABORATORY EVALUATION OF THE STIR-FLASK
`
`METHOD FOR THE DETERMINATION OF OCTANOL-WATER
`
`PARTITION COEFFICIENTS (LOG POW)
`
`Dave Brooke*, Ilga Nielsen
`
`Building Research Establishment,
`
`Department of the Environment,
`
`Garston, Watford, Hertfordshire, WDZ 7JR, UK.
`
`Jack de Bruijn, Joop Hermens
`
`Research Institute of Toxicology,
`
`Environmental Toxicology Section,
`
`University of Utrecht,
`
`PO Box 80.176, 3508 TD Utrecht,
`The Netherlands.
`
`SUMMARY
`
`The following note gives experimental details of the stir-flask method of log Pow
`determination.
`A summary of other more established methods of estimating or measuring
`
`octanol-water partition coefficients is also included.
`
`The results given originate from an interlaboratory project involving the Building Research
`
`initiated
`(The Netherlands),
`Establishment and the Research Institute of Toxicology, Utrecht
`in order to evaluate the applicability of this method for measuring log Pow values greater
`
`than A or 5 (at which point problems arise with the current Organisation for Economic
`Cooperation and Development
`(OECD) Guideline Method) and also to assess the comparability of
`results obtained by the two laboratories using the same basic method.
`
`The results obtained by the two participating laboratories were in good agreement both with
`
`each other and with previously determined values reported in the literature.
`
`It is
`
`recommended that the method should be the subject of a proper ring test, and subsequently
`
`considered for inclusion in the OECD recommended test methods for this property.
`
`119
`
`ALCON 2243
`Apotex Inc. v. Alcon Pharmaceuticals, Ltd.
`CaselPR201300012
`
`

`

`120
`
`INTRODUCTION
`
`It has long been acknowledged that the n-octanol-water partition coefficient (Pow) plays an
`important part in the early stages of an environmental risk assessment for a chemical.
`The
`
`Pow is an indication of the extent to which the chemical will bioconcentrate in an aquatic
`
`organism.
`
`The Pow also correlates well with the soil sorption coefficient (Koc), with water
`
`solubility and with parameters predicting biological, biochemical and toxic effects; because
`
`of its many applications,
`
`the Pow is often part of the basic data for computer models
`
`predicting the distribution and eventual fate of chemicals released to the environment.
`
`The octanol-water partition coefficient is calculated as the ratio of the activities of the
`relevant chemical
`in octanol and in water (when the two solvents and the chemical are under
`
`conditions of equilibrium). At
`
`low concentrations however, activity is assumed to be
`
`equivalent to concentration and the partition coefficient can be expressed as follows:
`
`Pow = Concentration of chemical
`
`in n-octanol saturated with water
`
`Concentration of chemical
`
`in water saturated with n-octanol
`
`Pow is usually expressed as its base 10 logarithm, i.e.
`
`log Pow.
`
`Several methods have been used to measure log Pow for various chemicals’“;
`
`the OECD
`
`(Organisation for Economic Cooperation and Development) Test Guideline 107l specifies the
`shake—flask method. This method involves the thorough mixing of the chemical with the two
`
`followed by the separation of the phases and the
`phases (octanol and water) by shaking,
`determination of the chemical's concentration in each phase.
`The method does however have
`
`limitations;
`
`the experience of many laboratories has shown that the method is not suitable
`
`for chemicals having log Pow values greater than 4 or 5. Glass and surface adsorption
`effects are a problem with this method whatever the chemical being studied and they must be
`
`minimised to achieve good results. Another difficulty is the possible formation of emulsions
`
`in the aqueous layer during shaking, and the consequent contamination of the aqueous phase.
`Again, emulsion formation can occur for any chemical, not just those with higher log Pow
`
`values. However, when the concentration of a chemical
`
`in the water phase at equilibrium is
`
`low (as is the case when the compound has a higher log Pow value), even small amounts of
`
`octanol contaminating the aqueous phase can lead to large errors.
`
`The results obtained using
`
`this method can also be affected by impurities in the equilibration vessel. This method is
`
`unsuitable for surface-active compounds and organometailics, and in cases where the compound
`
`is highly hydrophobic and of low water solubility,
`
`the analytical method required to
`
`determine the equilibrium concentration in the aqueous phase may need considerable
`
`development.
`
`In order to overcome the main difficulties of the shake-flask method, a modified method,
`
`the
`
`instead the
`In this the two layers are not shaken together,
`stir—flask method was devisedz.
`system is stirred for a period of time (at least 36 hours) allowing equilibrium to be
`
`reached.
`
`The problems of sorption of chemical onto glass, and of the analysis of chemicals
`
`of low water solubility remain; however,
`
`the method removes the possibility of gross
`
`

`

`121
`
`contamination of one phase by the other and the danger of emulsion formation. This method
`
`also reduces labour intensity and minimises loss of the sample through evaporation. Close
`
`is however necessary to avoid emulsion formation caused by temperature
`temperature control
`cycling through the equilibration period. Results obtained using the stir—flask method were
`
`reported2 32 and shown to be very reproducible and also in good agreement with previously
`
`measured log Pow values where available.
`
`The further development and use of this method is
`
`the subject of this report.
`
`In addition to the direct measurement methods, a number of methods have been developed for
`
`indirect measurement or estimation of log Pow.
`
`Foremost amongst these other methods are:
`
`(1)
`
`estimation from High Performance Liquid Chromatography (HPLC) retention times,
`
`(2) measurement by the generator-column method and (3) calculation methods by structure
`fragmentation.
`
`(1) The use of reverse phase HPLC as a rapid method for estimating log Pow has been reported
`
`by several authorss'1‘. This method depends on establishing a correlation between HPLC
`
`capacity factors (derived from HPLC retention times) and the octanol-water partition
`coefficients for chemicals with previously measured log Pow values;
`the Pow for a new
`
`chemical can then be obtained from its retention time.
`
`The HPLC method has been reported as
`
`The method is rapid, accurate and reproducible and
`reliable over the log Pow range 0-6.
`although impurities may make interpretation of the data and assignment of peaks more
`
`difficult,
`
`the method is generally less sensitive to impurities than the shake-flask method.
`
`Only neutral compounds and compounds in non—ionic form can be measured well by this method,
`
`and since reference compounds are needed for the calibration of the method, difficulties
`arise if suitable standards are unavailable.
`
`One of the main criticisms of the HPLC method for estimating partition coefficients lies in
`
`the fact that it is a secondary method (i.e.
`
`the coefficient is not measured directly but is
`
`found by calibration of the capacity factor against existing measured log Pow values).
`reference compounds used should preferably be structurally similar to the compound being
`
`The
`
`studied. Extrapolation is necessary for compounds whose coefficients lie outside the range
`of references used,
`this will reduce the reliability of the results.
`
`The HPLC method is in general reliable; occasionally, however, results obtained differ from
`
`shake-flask determinations by more than can be explained by experimental error alone.
`
`Several studies have reported such 'outliers'5 9 1° 1‘ 1’ and have attempted to explain why
`they occur.
`
`The HPLC method has recently undergone an OECD interlaboratory comparison test,
`
`the overall
`
`in 1988‘“. This report recommended that both
`results of which were published by Klein et al
`the shake-flask and HPLC methods should be included in future OECD Guidelines, for the
`
`determination of partition coefficients over different application ranges.
`
`

`

`122
`
`(2) A second method for the indirect measurement of octanol-water coefficients based on
`
`In this method, water is
`chromatographic techniques is the generator-column method‘5,15.
`pumped through columns containing a solid support coated with an organic stationary phase; an
`aqueous solution is generated, which is in equilibrium with the stationary phase. When the
`
`organic stationary phase is pure solute,
`
`the concentration of the solute in the eluted
`
`aqueous phase is the aqueous solubility of the substance. When the organic stationary phase
`
`is octanol containing a solute of concentration Co and the concentration of the solute in the
`
`Thus the generator-column method can measure
`then the Pow = Co/cw'
`aqueous phase is Cw,
`both aqueous solubilities and octanol-water partition coefficients.
`The generator—column is
`
`intended as a supplement
`
`to the shake-flask method, not as a replacement and can be used if
`
`the log Pow value is so high that analysis of the water phase is difficult if the
`
`shake-flask method is used.
`
`In the generator-column method,
`
`the water passes through the
`
`generator-column onto an extractor column, which is flushed out after a suitable period of
`
`time onto an HPLC column for analysis.
`
`The method has several advantages over the shake-flask method.
`
`The water flow through the
`
`generator-column can be made slow enough to avoid formation of emulsions in the aqueous
`phase, and the large interfacial area between organic and aqueous phases allows rapid
`
`equilibration to occur. When the column is made part of a continuous and closed flow system,
`
`the walls become equilibrated with the aqueous solution and errors due to adsorption are
`
`there is no exposure of the solution to the atmosphere,
`Since in a closed system,
`avoided.
`errors due to loss of volatile components are also minimised.
`
`Despite the fact that the generator-column method overcomes many of the intrinsic errors of
`the shake-flask method,
`there are doubts as to whether some of the log Pow values obtained
`
`represent the true thermodynamic equilibrium of the material between octanol and water (as
`the stir-flask and shake-flask methods do). Questions raised about
`the method include
`
`whether sufficient time is allowed for equilibrium to be attained, especially if the compound
`is of low water solubility, and whether chemical or other interactions between the
`
`generator-column and the test substance could lead to errors, i.e.
`
`if the compound
`
`hydrolyses in the water on the way through the column.
`
`It is also possible that if compounds
`
`with high water solubility pass through the extractor column,
`
`there may be a danger of
`
`overloading the column, or that for especially adsorbent compounds the contents of the
`
`extractor column may not be completely flushed out onto the HPLC column for analysis.
`
`Recovery and repeat experiments may therefore be necessary to ensure that these occurrences
`
`do not affect results.
`
`The generator-column is also much more elaborate and time—consuming
`
`in terms of the apparatus used when compared to the shake-flask method and to a greater
`
`degree when compared to the HPLC method.
`
`(3) The third method calculates log Pow values from the structures of the individual
`
`The procedure involves the division of the molecule under scrutiny into
`chemicals’ “.
`suitable substructures for which reliable Pow increments are known.
`The respective fragment
`
`

`

`123
`
`values are summed and correction terms (accounting for intramolecular attractions,
`
`for
`
`instance) are added in to produce the log Pow estimate. Although the method is of course
`
`limited by the size and scope of the existing fragment data base,
`
`the estimated values are in
`
`good agreement with a wide range of measured log Pow values. This method does however tend
`to overestimate values for those compounds with higher octanol-water partition coefficients,
`
`probably because as yet not all relevant interactions can be accounted for in what are
`
`generally complex molecules. Despite these problems and being limited by the availability of
`fragment values,
`this calculation method is being continually re-evaluated, developed and
`
`updated.
`
`The method is also available commercially as a computer programmeal.
`
`This calculation method can be used to make rough estimates of log Pow values so that the
`
`appropriate conditions and experimental method can be chosen for determination of the
`
`coefficient, and in some cases, due to the intrinsic nature of the compound, calculation
`
`methods are the only way of estimating the octanol-water partition coefficient (e.g. for
`
`compounds with very high coefficients).
`
`Calculation methods in general require a certain amount of expertise when dealing with
`
`complex molecules and are less accurate than the shake—flask method, however they have been
`
`The HPLC ring-test report‘“ recommended
`widely used to calculate log Pow values3 “.
`calculation methods for inclusion in the OECD Guideline primarily for screening compounds to
`
`ascertain the most suitable determination method to use (either shake-flask or HPLC), but
`
`also as a 'last resort' determination if the log Pow of a chemical could not be measured
`
`experimentally due to its nature.
`
`As part of the development of the stir-flask method, a collaborative project was initiated
`
`(BEE) and the Research Institute of Toxicology
`between the Building Research Establishment
`(RITOX)
`(Netherlands).
`The aim of the project was to assess the possible applications of
`
`the point at which the
`this method for compounds with log Pow values greater than H or 5,
`current OECD Guideline method tends to become unreliable;
`the project also sought to assess
`
`the comparability of the results obtained by two different laboratories using the same basic
`stir-flask method but with slightly different experimental set-ups.
`
`The test compounds for the project were chosen from the list of chemicals used for the OECD
`
`ring-test of the HPLC method‘“, selecting compounds with HPLC-derived log Pow values greater
`than N.
`The test compounds were as follows:
`
`n-Butylbenzene
`
`Diphenylether
`Phenanthrene
`
`Fluoranthene
`
`1-Hydroxyanthraquinone
`
`Triphenylamine
`pp'-DDT
`
`Di-2-ethylhexylphthalate
`
`2-Ethylanthraquinone
`
`Dinoseb (2,u—Dinitro-6-sec butylphenol)
`
`

`

`124
`
`Initially, five other compounds with lower, well-defined partition coefficient values
`
`(toluene, chlorobenzene, nitrobenzene, 1,u-dichlorobenzene and aniline) were studied to allow
`familiarisation with the technique and to identify any early problems.
`
`The following sections of this report give information on the methods, materials and
`
`apparatus used by both BRE and RITOX in their study of the stir-flask method for partition
`coefficient determination. Also included are the results obtained by both laboratories.
`
`Previously determined log Pow results obtained by other laboratories using other methods are
`included for comparison.
`
`EXPERIMENTAL
`
`Materials
`
`The solvents and test compounds used in this project were obtained from various sources
`gap:
`at high grades of purity.
`Two of the compounds, 2-ethylanthraquinone and
`
`1-hydroxyanthraquinone were left over from the OECD interlaboratory comparison test of the
`
`HPLC method for octanol-water partition coefficient determination‘“,
`
`in which BRE took part.
`
`The water needed for the experiments was distilled in an all-glass apparatus.
`
`For the second
`
`set of dinoseb runs, distilled water buffered with orthophosphoric acid to pH 2.0 was used.
`
`
`RITOX:
`
`The test chemicals were purchased from several different sources at the highest grade
`
`of purity available.
`
`The water that was used in all the experiments was distilled in an
`
`For the second set of dinoseb determinations, distilled water buffered
`all-glass apparatus.
`to pH 2.0 with 0.1N citric acid was used.
`
`Apparatus
`BRE: The apparatus used consisted of a 1 litre glass aspirator bottle with a ground glass
`
`The two phases were contained in the aspirator with a magnetic stirring bar
`stopper and tap.
`and the whole assembly was kept at constant temperature in an incubator oven.
`
`
`RITOX:
`
`The reaction vessel consisted of a straight double-walled Pyrex glass flask with a
`
`capacity of approximately 1 litre.
`
`The top of the flask tapered off to a ground glass
`
`stopper. Water samples were taken from a Teflon tap connected to the interior of the vessel
`approximately 2 cm from the bottom. Water at a constant temperature was pumped through the
`
`wall cavity around the vessel.
`
`The vessel was covered with insulating material to keep
`
`temperature fluctuations to a minimum and to prevent any influences from daylight.
`
`By
`
`connecting the wall cavities of several vessels with plastic tubing, multiple determinations
`could be carried out simultaneously.
`
`Procedure
`
`BRE:
`
`A known amount of the test compound was dissolved in octanol prior to equilibration.
`
`

`

`125
`
`The concentration of the solution ranged from 0.01M to 0.1M;
`dependent on the limits of the analytical techniques used.
`
`the concentration chosen being
`980 ml of water and 20 ml of the
`
`solution of the chemical
`
`in octanol were placed in the aspirator and left for at least 36
`
`hours in the incubator oven,
`
`the contents of the aspirator being stirred constantly at
`
`Problems with the incubator oven meant that the temperature could
`approximately 100 rpm.
`only be kept
`to 25 : 3°C.
`
`After the period of equilibration, samples of the octanol layer were removed by pipette and
`water samples were drawn off from the tap and analysed (this arrangement also avoided
`contamination of the water samples).
`
`Analysis of the samples was by one of three techniques,
`are given below:
`
`the instrumentation details of which
`
`1. High resolution gas chromatography (GC-FID).
`
`A Carlo Erba Fractovap A160 series gas
`
`chromatograph was used, equipped with a dimethyl silicone BP1 capillary column (25 m long,
`internal diameter 0.22 mm, phase thickness 0.25 pm).
`The GC was fitted with a Carlo Erba
`
`injection sizes and techniques,
`flame ionisation detector (FID A0). Appropriate gas flows,
`and temperatures were chosen for each compound under analysis.
`Integration of the GC traces
`
`was carried out using a Trivector Trio chromatography computing integrator.
`
`A Milton Roy liquid chromatograph
`2. High performance liquid chromatography (HPLC).
`equipped with a Constametric Model III pump and a Spectro Monitor III 120HA UV detector was
`
`used.
`
`The HPLC was fitted with a Du Pont Zorbax 250 mm x ”.6 mm, 216N6 theoretical plate,
`
`reverse phase Cla column.
`
`Sample introduction was via a Rheodyne 7125 injection valve fitted
`
`with a 20 ul
`
`loop. Appropriate detection wavelengths and mobile phase compositions were
`
`The mobile phases were degassed before use by
`chosen for each compound under analysis.
`refluxing for one hour.
`The same integrator was used as for the GC analysis.
`
`A Shimadzu doublebeam UV 190 Spectrophotometer was used,
`3. UV/visible absorption (UV/vis).
`set at the appropriate maximum absorbance wavelength for each compound.
`
`For the compounds with
`The water samples were extracted, where necessary, with hexane.
`higher log Pow values,
`the hexane extracts from 2 or 3 water samples were combined and
`concentrated.
`In some cases, where hexane interfered with the detection of the chemical
`
`being analysed for,
`different solvent.
`
`the hexane was removed completely under nitrogen flow and replaced by a
`In the case of HPLC analysis, methanol was used as the new solvent.
`
`Extraction efficiencies were calculated using spiked samples and the relevant extraction
`
`method, and where necessary the concentrations recorded for the water samples were adjusted
`
`The water samples were analysed directly in the cases where
`according to their efficiencies.
`extraction and concentration were not necessary (i.e. for the UV analysis and some of the
`HPLC analyses).
`In all cases,
`the extraction efficiencies were over 80% for the relevant
`chemicals.
`
`

`

`126
`
`Subsamples of the original octanol samples taken were diluted to appropriate concentration
`levels before analysis.
`Samples for CC analysis were diluted with hexane, and samples for
`both HPLC and UV analysis were diluted with methanol.
`
`Multiple samples were taken where possible to improve the precision of the results.
`
`
`RITOX’Z: At
`
`the beginning of an experiment approximately 9N0 ml of distilled water was
`
`placed in the reaction vessel together with a Teflon coated magnetic stirring bar. When the
`
`aqueous phase had reached the desired temperature (the temperature of the outcoming water
`
`from the insulating jacket was constantly monitored and remained at 25 t 0.1°C), 20 to 50 m1
`
`in octanol (concentration from 0.5 to 100 mg/ml) was
`of a solution of the test chemical
`brought into contact with the water phase without allowing mixing to occur. When the vessels
`
`were completely filled with the test substances the stirring rate was approximately 200 rpm.
`
`Both octanol and water samples were taken on several successive days.
`
`For compounds with log
`
`Pow values <5, samples were taken on days 2 and 3.
`
`For compounds with higher log Pow values
`
`An aliquot of the octanol phase was taken by
`H and 532.
`sampling took place on days 3,
`pipette, and diluted with 1:1 acetone/hexane (by volume) or methanol. Water samples were
`
`taken in triplicate.
`
`The volume of the water sample taken depended on the type of the test
`
`chemical,
`
`the expected log Pow value and the concentration of the octanol solution and varied
`
`from 3 to 100 ml.
`
`The treatment of the water sample depended on the method of analysis used.
`
`Samples analysed by HPLC were injected onto the column directly after sampling, whereas
`
`samples analysed by GC were first extracted with hexane, and samples were stored in a
`
`If necessary,
`refrigerator until needed.
`times under a stream of dry nitrogen gas.
`and 100%.
`
`the hexane extracts were concentrated 2 to 200
`The extraction efficiencies were always between 80
`
`The GC analyses were carried out on a Carlo Erba 5360 gas chromatograph equipped with an
`electron-capture detector (ECD) and a flame ionisation detector (FID), on a Pye Unicam HSSOGC
`with ECD and FID or on a Tracor 550 CC with ECD.
`The Carlo Erba and Pye Unicam GCs were
`
`fitted with CP 811 8C8 capillary columns (0.32 mm ID,
`
`length 25 m).
`
`The Tracor was fitted
`
`with a wide bore 8P1 CB capillary column (0.53 mm ID,
`
`length 25 m).
`
`Injections of the hexane
`
`extracts (2 ul) were made on either a split injection or an on-column injection system.
`
`Injector and column temperatures were chosen as relevant for the compounds under analysis.
`
`The HPLC equipment used consisted of a Pye Unicam u01o double piston pump that operated at
`
`0.” ml/min and a Pye Unicam u020 UV detector. All solvents used were filtered over a o.u5 um
`
`filter, and degassed prior to use. Determinations were carried out at room temperature on a
`reverse-phase C18 column (Chrompack Chromspher C19, particle size 5 pm,
`length 10 cm,
`ID 3
`
`mm). Water samples were injected directly onto the HPLC column, octanol samples were first
`diluted with methanol.
`
`UV analysis of aniline was carried out using a Pye Unicam PU8600 Spectrophotometer.
`
`

`

`127
`
`Both HPLC and GC signals were integrated with a Shimadzu CR1-A or CR3-A integrator.
`
`RESULTS
`
`1.B_RE_
`The overall results from the BRE determinations are presented in Table 1.
`
`The values quoted
`
`As can be seen,
`are the averages of several log Pow determinations for each compound.
`standard deviations are in almost all cases (0.1 log units, even for the compounds with
`
`the
`
`higher log Pow values. Results for dinoseb are presented for both neutral pH and pH 2.0
`determinations.
`
`Previous experiments2 had shown that 72 hours was a sufficient time period for a compound as
`
`lipophilic as hexachlorobenzene to reach equilibrium. However, it was thought prudent to
`
`allow a longer time period for the more lipophilic compounds in this study, and so these were
`
`left for N-S days. Results from RITOX32 (see below)
`unnecessary .
`
`show that this was probably
`
`2.
`
`
`RITOX
`
`The overall results from the determinations by RITOX are presented in Table 1. Results for
`
`dinoseb are presented for both neutral pH and pH 2.0 determinations.
`
`It can be seen that the
`
`reproducibility between runs for each compound is very good, with the standard deviation
`
`never exceeding 1 0.06 log units.
`
`In these runs the approach to equilibrium was monitored by
`
`taking samples from both phases on consecutive days until a steady value was obtained. This
`was found to occur within 2-3 days for all the chemicals studied; however,
`it is recommended
`
`that this procedure be followed when beginning tests on any new substance.
`
`DISCUSSION
`
`Comparison of results from two laboratories
`One of the primary objectives of this study was to compare the results obtained using the
`stir-flask technique at two different laboratories.
`The two sets of results are presented in
`Table 1.
`
`An immediately noticeable feature is the difference in the variability of the results from
`the two laboratories, as shown by the standard deviations.
`It is likely that the increased
`
`variability in the results obtained at BRE was due to the less accurate temperature control
`
`provided by the incubator oven used.
`
`As described above, this allowed the temperature to
`
`vary by 2-3°C during the course of some experiments, whereas the water-jacket system used by
`RITOX was able to maintain a temperature of 25 1 0.1°C.
`It has been suggested2 that
`
`temperature cycling can lead to the formation of emulsions, which would result in a higher
`concentration being measured in the aqueous phase, and hence a lower log Pow value being
`
`obtained.
`
`A comparison of the results from the two laboratories shows that the BBB results
`
`are in general
`
`lower than those from RITOX, supporting this interpretation.
`
`

`

`128
`
`A criterion of i 0.3 log units (as used by OECD) was used to assess the 2 sets of data
`
`presented here.
`
`It can be seen that only 2-ethylanthraquinone, 1-hydroxyanthraquinone,
`
`pp'—DDT and di-2-ethylhexylphthalate differed by more than i 0.3 log units.
`
`For most of the
`
`compounds studied,
`
`the agreement was better than 1 0.1 log units. These results suggest that
`
`the stir-flask technique is at least as reproducible from one laboratory to another as the
`
`conventional methods of measuring log Pow.
`
`Comparison of results with literature values
`In addition to an internal comparison of the data, it is also useful to compare the results
`
`from this study with values obtained from the literature.
`
`A collection of such values for
`
`the chemicals in this study is presented in Table 1, determined by the shake-flask, HPLC,
`generator—column and calculation methods.
`
`For some of the chemicals a range of published values was found, while for others only the
`
`result of one indirect determination was located. Consequently, it is possible to make a
`number of different comparisons.
`For those compounds for which at least one shake-flask
`
`measurement was found,
`
`the agreement with the two sets of results presented here is, on the
`
`whole, good.
`
`For the chemicals with low log Pow values ((U),
`
`the agreement is better than
`
`1 5%,
`
`thus the stir-flask technique is capable of duplicating shake-flask values in the area
`
`where this method is known to be reliable.
`
`And with the exception of
`
`di-2-ethylhexylphthalate and pp'-DDT,
`
`there is also good agreement with the shake-flask
`
`values for higher log Pow compounds where these have been measured. Here the stir-flask
`
`values are much higher than the reported shake-flask data;
`
`from the comments in the
`
`introduction on the problems encountered with the shake-flask method a low value might be
`expected.
`It is worth noting that the values measured in this work are in much better
`
`agreement with the HPLC—derived values as measured in the OECD ring—test of the HPLC
`
`triphenylamine and 2-ethylanthraquinone both show good agreement
`method‘“. Similarly,
`between the stir-flask values and those derived from HPLC. There are one or two compounds
`for which the agreement is not so good.
`No obvious explanation can be suggested for the
`
`discrepancy for 1-hydroxyanthraquinone, although possible instability of the compound in the
`
`The discrepancy with the dinoseb results determined under
`aqueous phase can not be excluded.
`neutral conditions is due to ionisation,
`in that the solutions used in the HPLC work were
`
`buffered whereas for this run they were not. This is borne out by the repeated
`
`determinations for dinoseb carried out at lower pHs by both laboratories which are closer in
`
`value to the HPLC result, and also to the calculated value.
`
`In general, stir—flask values
`
`agree well with calculated values, with the exception of di-2-ethylhexylphthalate.
`
`Nevertheless,
`
`the agreement between the HPLC data and the directly measured stir-flask data
`
`for hydrophobic chemicals suggests that we can have more confidence in these methods than in
`
`values that are obtained using the shake-flask procedure. Also,
`
`the stir‘flask method can be
`
`used to supply directly measured reference values for use in further HPLC work.
`
`It is worth
`
`

`

`129
`
`
`
`
`
`$31.38
`
`caioonLoumecwo
`
`
`
`mung;mmPEmMHHJ
`
`
`
`.wa.N.zmTN
`
`{NEW.0.—m.m
`
`:mo.m(«mod
`
`meEonmzm
`
`cowumgmu
`
`coUmgmu
`
`moo.oumafim
`
`modumod3538‘
`
`ULaucmum“:8woq$8.2:
`
`:8m33:;
`
`
`
`unnucwun«9222200
`
`
`
`
`
`
`
`pumcdmuno~96ngmwsdm>3:283:3:comicEco“nozumexmmdufium93
`Engx09;canmumnumEELmumu33?:0;0Hm«.53..
`
`"—vans.
`
`
`
`30505.550
`
`
`
`.
`
`
`
`2mm;.:5.N.:$.N:am.N[Sim:85«mom.m$0.0n=3.~2523225
`
`3.mm.m..zN.m
`
`
`
`2mm.m.Lmé
`
`
`
`:5;.3».—
`
`~.mm.m.:2:4”
`3mm.m.nL.m.m
`
`emod«.134$06ammm.m
`
`50.0«wmmé30.0nwmmimcuucoaonuuz
`mcmucoDOLoHcoaoI=._
`
`:ow.m
`
`
`
`
`
`.ché.23.:.29....film...
`
`L9:.1.0.:.39..
`
`
`
`
`_—Nm.:.»«©3.2000.0NNwmua:MO.O.fl:5M.3E:3...
`
`
`
`
`
`:00.».:wo._{no.0
`
`9.5.0{and.
`
`$3.0...006
`
`.::85...mm.o
`
`:2;
`
`
`
`000.0M030.00—.0.oHsz‘oOCAAHC<
`
`
`
`
`
`.335!08.0«Em...306nmow...235235-:
`
`
`
`.2E....23...80.0H:5...Rodnm=m.mLufiwicmfig
`
`
`
`
`
`
`
`
`
`:34..9;(«3.9.
`
`
`
`.236.255$3.0._:36...om.momoé«:3696.0“Ema.8933
`
`
`
`
`
`0—0.0«m:F.mmmo.onmwm.~AxaHmLuzw:umvnwnocuo
`
`
`
`Fmo.oHmhm.m_:o.o“Nom.mAo.m:aanyamnocdo
`
`
`:36!owedN=mm.m370HEndufisaicwzfié
`
`13.:«no.0“zoma$0.0.4EAwcocfisumgcucmioézxi
`
`
`
`
`
`
`
`10w...hmodu2%.:3.0.0¢SuimcocfizcmLzucmszumuw
`
`
`
`
`
`
`
`
`
`
`
`
`
`
`
`
`
`
`

`

`130
`
`noting that two of the authors have measured a log Pow value of 8.27 for decachlorobiphenyl
`
`by this method”.
`
`CONCLUSIONS
`
`The results presented here Show that the stir-flask method is able to produce comparable
`
`results when used in different laboratories, and that these values can be of high precision
`
`if accurate temperature control can be maintained.
`
`The results agree with those obtained by
`
`the shake-flask method over the range of log Pow values where that method is considered
`
`accurate;
`
`they also agree with measurements on more lipophilic compounds by surrogate methods
`
`such as HPLC.
`
`Thus it appears that this is a direct method for the measurement of the
`
`including high
`partition coefficient which is applicable over a wide range of log Pow values,
`values.
`The method overcomes the problems associated with emulsion formation and volatile
`
`chemicals which can occur with the shake—flask procedure but without
`
`the complexity of
`
`equipment and procedure of chromatographic methods.
`
`As such, it would appear to merit a
`
`proper ring test, and subsequently be considered for inclusion in the OECD recommended test
`
`methods for this property.
`
`ACKNOWLEDGEMENTS
`
`Acknowledgements are made to Norman Williams and Frans Busser for their helpful advice on
`
`analytical techniques.
`
`REFERENCES
`
`(1)
`
`(2)
`
`(3)
`
`(A)
`
`(5)
`
`(6)
`
`OECD Guideline for Testing of Chemicals, 107, Partitional Coefficient (n-octanol/water)
`- Flask Shaking Method, OECD, Paris, 1981.
`
`D N Brooke, A J Dobbs, N Williams. Octanol:Water Partition Coefficients (P):
`Measurement, Estimation and Interpretation; Particularly for Chemicals with P>105.
`Ecctox. and Env. Safety, 1986, 11, 251-260.
`
`Pharmaco-Chemistry Library, Vol 1, The Hydrophobic Fragmental Constant.
`R F Rekker.
`Elsevier, Amsterdam, 1977.
`
`C Hansch, A J Leo. Substituent Constants for Correlation Analysis in Chemistry and
`Biology, 1979, John Wiley and Sons, New York.
`
`A Rapid Method for Estimating Log P for Organic
`G D Veith, N M Austin, R T Morris.
`Chemicals. Water Research, 1979, 1;, u3-u7.
`
`J E Garst, W C Wilson. Accurate, Wide-Range, Automated, High-Performance Liquid
`Chromatographic Method for the Estimation of Octanol/Water Partition Coefficients. I:
`Effect of Chromatographic Conditions and Procedure Variables on Accuracy and
`Reproducibility of the Method.
`J. Pharm. Sci., 1984, 1;, 1616-1623.
`
`

`

`131
`
`(7)
`
`J E Garst. Accurate, Wide-Range, Automated, High-