Fulvestrant: From the laboratory to commercial-scale manufacture

Article abstract of DOI:10.1021/op900315j

The development of a commercial manufacturing process for fulvestrant (the active ingredient in 'Faslodex') is described. Key steps in the synthesis are stereoselective 1,6-addition of an organocuprate to a steroidal dienone followed by copper-mediated aromatisation of the A-ring. The strategy for dealing with noncrystalline intermediates is outlined. The production of drug substance of acceptable quality is critically dependent on limiting the formation of key impurities. The origin of these impurities is discussed, and measures to prevent or control their formation are described.

Full text of DOI:10.1021/op900315j

Organic Process Research & Development 2010, 14, 544–552
Fulvestrant: From the Laboratory to Commercial-Scale Manufacture
Eve J. Brazier, Philip J. Hogan, Chiu W. Leung, Anne O’Kearney-McMullan, Alison K. Norton, Lyn Powell,*
Graham E. Robinson, and Emyr G. Williams
Process R&D, AstraZeneca, Silk Road Business Park, Charter Way, Macclesfield, Cheshire SK10 2NA, U.K.
Abstract:
mediates, and (c) introduction of the C7 substituent by a low-
temperature addition to a dienone which afforded a 1.9:1 ratio
of isomers with the required 7R-isomer predominating. Crys-
tallisation of the final product provided the only opportunity
for nonchromatographic purification and fortunately removed
the unwanted 7ꢀ-isomer.
The development of a commercial manufacturing process for
fulvestrant (the active ingredient in ‘Faslodex’) is described. Key
steps in the synthesis are stereoselective 1,6-addition of an orga-
nocuprate to a steroidal dienone followed by copper-mediated
aromatisation of the A-ring. The strategy for dealing with
noncrystalline intermediates is outlined. The production of drug
substance of acceptable quality is critically dependent on limiting
the formation of key impurities. The origin of these impurities is
discussed, and measures to prevent or control their formation are
described.
Eventually, a supply of pentafluoropentanol 10 was estab-
lished, and its contribution to the cost of manufacture diminished
so this ceased to be a dominant factor in the choice of synthetic
route. The steroidal starting material became by far the most
expensive ingredient in the synthesis so it was appropriate to
focus on ways of utilizing this material most efficiently. It
seemed likely that this could be achieved by preassembly of
the sulfide-bearing side chain, thus reducing the number of
stages from the steroidal dienone to fulvestrant. This article
describes the investigation and development of a second
synthetic route in which the sulfide-bearing side chain is
assembled before addition to the steroid (Scheme 2). The
alternative approach of using the sulfoxide-bearing side chain
in the organocopper-mediated coupling was discounted because
of incompatibility of the alkyl sulfoxide with the reaction
conditions. The second route involves fewer noncrystalline
steroid intermediates than the first route, was found to give a
rather more favorable ratio of 7R- and 7ꢀ-isomers in the
organocuprate addition step (2.5:1 versus 1.9:1), and made
significantly more efficient use of the expensive steroidal starting
material. It became the route of choice for large-scale manu-
facture.2
Introduction
Fulvestrant 6 is the active ingredient in ‘Faslodex’, a novel
oestrogen-receptor downregulator that has proved to be a safe
and effective treatment for advanced breast cancer.1 The drug
is formulated as an oily solution and is administered as a long-
acting intramuscular injection. Key steps in the synthesis are
stereoselective 1,6-addition of an organocuprate to a steroidal
dienone followed by copper-mediated aromatisation of the
A-ring. Fulvestrant is produced as a mixture of sulfoxide
isomers, both having essentially the same pharmacological and
toxicological properties.
Several potential synthetic routes were considered that
involve assembly of the molecule in a different order. 19-
Nortestosterone, 1, 1,9-nonanediol, 7, and pentafluoropentanol,
10, were commercially available (although expensive) and
envisaged to be likely starting materials in any synthetic route.
The most likely production route to 10 started from pentafluo-
roethyl iodide which is difficult to handle due to its volatility
(bp 13 °C) and the environmental issues associated with iodine-
containing compounds. Since the supply of 10 for initial
manufacturing campaigns was uncertain, it seemed prudent to
focus on a route which used this key raw material most
economically and as late as possible in the synthesis. The
synthetic route used to make the first gram quantities of
fulvestrant involved preparation of silyl-protected 1-bromo-9-
hydroxynonane and its conjugate addition to a steroidal dienone
(Scheme 1).
Results and Discussion
Second Synthetic Route. A consideration in developing the
second synthetic route was the choice of thiol to introduce the
sulfide bond by displacement of a leaving group (e.g., mesylate
ester) from the appropriate alkyl derivative. Since pentafluo-
ropentanethiol is relatively volatile (bp 112 °C) and very smelly,
the alternative disconnection involving mercaptononanol 12 was
preferred because this compound was expected to be much less
volatile. This approach was developed into a scaleable manu-
facturing route to the required side-chain bromide 14 (Scheme
3).3 A key feature of this route is in situ formation of 12 from
isothiuronium salt 17 and its reaction with pentafluoropentyl
mesylate 11 (also a nonisolated intermediate). Bromination of
the crystalline alcohol is performed with triphenylphosphine and
bromine in acetonitrile because simpler methods (e.g., hydrogen
bromide) give unacceptable levels of side reactions. Bromide
Main areas of concern with the route were (a) the feasibility
of making pentafluoropentanol 10 on the large scale, (b) the
use of chromatography to purify several noncrystalline inter-
* Author for correspondence. Telephone: +44 (0)1625 230496. E-mail:
lyn.powell@astrazeneca.com.
(1) (a) Bowler, J.; Lilley, T. J.; Pittam, D. J.; Wakeling, A. E. Steroids
1989, 54, 71. (b) Morris, C.; Wakeling, A. E. Endocr.-Relat. Cancer
2002, 9 (4), 267. (c) Bundred, N.; Howell, A. Exp. ReV. Anticancer
Ther. 2002, 2 (2), 151.
(2) Stevenson, R.; Kerr, F. W.; Lane, A. R.; Brazier, E. J.; Hogan, P. J.;
Laffan, D. D. P. WO Patent 032922A1, 2002.
(3) Warren, K. E. H.; Kane, A. M. L. WO Patent 031399A1, 2003.
544
•
Vol. 14, No. 3, 2010 / Organic Process Research & Development
10.1021/op900315j  2010 American Chemical Society
Published on Web 03/29/2010

Scheme 1. First synthetic route
Scheme 2. Second synthetic route
14 is a liquid at ambient temperature, and its purification is
achieved with a Wiped Film Evaporator.
to the presence of traces of oxygen in the reaction vessel. It
reacts with the Grignard reagent to produce impurity 20 which
participates in the subsequent stages of the synthesis to give
the corresponding impurity 25 in fulvestrant.
Dibromononane 18 is inevitably present due to overbromi-
nation of alcohol 16. Another source of 18 is cleavage of
disulfide 12 by triphenylphosphine during the bromination step,
with subsequent bromination of the resulting mercaptononanol
12. Dibromononane 18 itself forms a diorganocopper reagent,
and this participates in 1,6-addition to two molecules of dienone
2, generating the corresponding sterol dimer impurity 24 in the
drug substance.
Quality Considerations with Manufacture of Bromide
14. 1,9-Dibromononane 18, dibromodisulfide 19, and the
extended side-chain bromide 20 are particularly troublesome
impurities encountered in the manufacture of bromide 14
(Scheme 4).
In early development, unpurified 14 behaved erratically in
the Grignard-forming step, and some batches failed to initiate.
GC analysis revealed several impurities in crude 14, including
19 (typically at about 1% w/w), and collectively these impurities
were found to inhibit formation of the Grignard reagent. This
led to the requirement for purification of 14 by distillation.
Distilled 14 contains <0.1% w/w 19 and has been found to
undergo initiation reproducibly in the Grignard reaction. Di-
bromodisulfide 19 originates mainly from oxidation of 12 due
It is particularly important to control the levels of 18 and
20 in manufacture as the derived impurities in fulvestrant (24
and 25) are not significantly removed by the purification process.
The limits for levels of impurities 24 and 25 (0.8% w/w and
Vol. 14, No. 3, 2010 / Organic Process Research & Development
•
545

Scheme 3. Large-scale manufacturing route for side-chain bromide
Scheme 4. Importance of bromide quality
0.3% w/w, respectively) in drug substance are justified by
appropriate toxicity studies.
An additional source of 18 is degradation of 14 on storage.
Thermally induced decomposition of 14 presumably arises via
intermolecular alkylation of the sulfur atom followed by
bromide ion attack on the sulfonium intermediate 21 to give
several degradation products (Scheme 5). Of these, the extended
bromide 20 forms most rapidly (pathway A). A corresponding
546
•
Vol. 14, No. 3, 2010 / Organic Process Research & Development

Scheme 5. Thermal degradation of the side-chain bromide
Scheme 6. Route from registered raw materials
increase in the level of pentafluoropentyl bromide 22 is not
observed, possibly because this component is volatile. Evapora-
tion of 22 would drive the equilibrium to the right, thus
explaining why decomposition appears to favor formation of
20. The alternative mode of cleavage (pathway B) affords 18
and bis(pentafluoropentyl)thiononane 23. It is beneficial to store
the bromide 14 at 4 °C to extend its useful life.
Final Synthetic Route from Registered Raw Materials.
During the course of development, suppliers of both the steroidal
dienone 2 and bromide 14 were established, leading to a shorter
in-house manufacturing sequence (Scheme 6).
Impact of Noncrystalline Intermediates. The lack of
crystalline intermediates between dienone 2 and the drug
substance was a major concern because of the limited op-
Vol. 14, No. 3, 2010 / Organic Process Research & Development
•
547

portunity for removing impurities. Unfortunately, attempts to
prepare crystalline derivatives of the hydroxyl group at C17
were unhelpful. With successive manufacturing campaigns and
increasing scale during development, the level of impurities in
the drug substance increased (eventually reaching a total of
2.8%) w/w, and this was a major cause of concern. Whilst some
impurities could be removed (or their levels decreased) during
recrystallisation of the drug substance, the levels of others were
unchanged, and some actually increased. A detailed investiga-
tion into the origin of the troublesome impurities was necessary
before technology transfer to the commercial manufacturing
plant. In the case of impurities that cannot be removed by
recrystallisation of the final compound, it was essential to
introduce methods for controlling their formation (e.g., as
described above for the side-chain bromide).
reactivity. To achieve this, 2.0 mol equiv of acetic anhydride
was added to the solution of cupric bromide and lithium bromide
in acetonitrile. This procedure prevented formation of the
bromo-impurities but led to a slightly elevated level of the ∆6,7-
impurity 29. To counteract this, in the optimised process half
the acetic anhydride is added to the cupric bromide solution,
and the remainder is added immediately after addition of the
cupric bromide solution has finished.
Introduction of the Side Chain. Development of a low-
temperature, copper-catalysed process for 1,6-addition of the
alkyl side chain to dienone 2 is described in detail elsewhere.2
The Grignard reagent is prepared at 45 °C in about 88% yield
and with a solution strength of about 0.4 M (as measured by
titration); in routine plant manufacture, the first batch is initiated
with a small amount of specially prepared reagent (made with
iodine as initiator) whereas the second and subsequent batches
are initiated with a heel of Grignard reagent retained from the
previous batch. Since formation of the Grignard reagent is
exothermic (65 kcal/mol, equivalent to a potential adiabatic
temperature rise of 65 °C), safety considerations dictate that it
should be added in four aliquots. The organocuprate is generated
at -34 °C by addition of cuprous chloride (0.08 mol equiv).
Controlling the temperature in the range -28 to -37 °C (set
point -34 °C) during the addition of 2 affords optimum
conversion; lower temperatures lead to 1,2-addition of the side
chain due to incomplete formation of the organocuprate, and
higher temperatures result in deacylation which is wasteful of
the Grignard reagent. The reaction affords the required enone
15 as a 2.5:1 mixture of 7R- and 7ꢀ-isomers in about 90% yield.
Since the product is an oil, it is not readily purified without
resorting to chromatography. The required enone 15 and
byproducts generated during formation of the Grignard reagent
are extracted into isohexane for use in the next stage (isohexane
was chosen as a suitable solvent for extraction of the product
because it permits an easy swap to acetonitrile prior to
performing the aromatisation step).
Aromatisation of the Steroid A-Ring. Control of Bromo-
Impurities. A mixture of cupric bromide and lithium bromide
in acetonitrile was found to be a highly effective system for
the aromatisation of 15. On first scale-up of the process in the
pilot plant, 2- and 4-bromo-impurities 27a and 27b were
generated at a level of about 1%. Unfortunately, the corre-
sponding sulfoxide derivatives of 27a and 27b could not be
removed during crystallisation of the drug substance. Indeed,
their level actually increased. Reaction monitoring by HPLC
indicated that the bromo-impurities began to form towards the
end of the aromatisation and their level increased significantly
when the reaction mixture was quenched. The problem was
attributed to the susceptibility of the phenol ring towards
electrophilic bromination. We reasoned that protection of the
phenol as an acetate (compound 28) would greatly reduce its
RemoVal of Copper. The optimised process uses 2.37 equiv
of cupric bromide (the precise charge being determined by the
preference for using whole bags in commercial manufacture).
Various methods for removal of used copper reagent at the end
of the process were investigated. Removal of cuprous bromide
by filtration and multiple washes with aqueous ammonium
chloride or potassium chloride were relatively inefficient.
Advantage was taken of the highly insoluble complex with
thiourea. When aromatisation is complete, copper is precipitated
by the addition of an aqueous solution of thiourea and the
required diacetate 28 extracted into toluene. Because the reaction
mixture is very acidic at this point, it is necessary to adjust the
pH with dipotassium phosphate to avoid the risk of corrosion
to filtration equipment.
Ester Hydrolysis. The toluene extract of 28 is subjected to
a solvent swap by distillation, and the acetate groups are
removed with aqueous sodium hydroxide in methanol. At this
point there is an opportunity to remove hydrocarbons and related
impurities (carried forward from the Grignard reaction) by
extraction with isohexane. The required intermediate 26 (also
an oil) is obtained by acidification and extraction into ethyl
acetate, the preferred solvent for the next stage.
DiscoVery of a New Impurity. HPLC analysis (225 nm) of
the first batch of 26 in the Technology Transfer Campaign
revealed a new impurity at a level of 1.3% by area with RRT
0.96. An attempt to identify the impurity by LC-MS failed as
no molecular ion was observed. Laboratory use trials showed
the level of impurity and its HPLC retention time were
unchanged on processing a sample of the intermediate 26 to
pure fulvestrant. A sample of the impurity was isolated by
chromatography, and its mass spectrum (EI) was characteristic
548
•
Vol. 14, No. 3, 2010 / Organic Process Research & Development

Scheme 7. Oxidation of thiourea by cupric bromide
possible to propose an explanation for the observed variability
in sulfur levels in 26. The charge of sodium hydroxide was on
a knife edge and did not allow for variable levels of acetic
anhydride remaining after the brine washes. With some plant
batches the apparent deficiency of sodium hydroxide led to slow
or incomplete hydrolysis of diacetate 28 and, crucially, incom-
plete decomposition of sulfur to thiosulfate. Any polysulfide
remaining after the hydrolysis would regenerate elemental sulfur
upon neutralisation of the reaction mixture (to pH 5-6) with
acetic acid.
Scheme 8. Mechanism for formation of sulfur from 30
For subsequent batches, the charge of sodium hydroxide was
increased from 2.40 to 2.79 equiv, resulting in faster hydrolysis
of 28. This solved the problem, and there was no sulfur in the
isolated 26.
Physical Methods of RemoVing Sulfur. Methods of removing
sulfur from contaminated batches of 26 were sought. Treatment
of the ethyl acetate solution with 5% w/w activated carbon (20
°C for 4 h) reduced the level of sulfur by about two-thirds
although several cycles were required to produce material of
acceptable quality. The method was scaled up to mobilise
valuable material for development studies.
Oxidation to FulVestrant and its Purification. In early
development manufacture, sulfoxidation was performed with
sodium periodate in aqueous methanol. The process suffered
from variable conversion, and there was concern regarding the
need to filter off and dispose of sodium iodate. Hydrogen
peroxide (17.5% w/w) and acetic acid in ethyl acetate were
found to be much more reliable, affording the product in high
yield with less than 1% conversion to sulfone. Several recrys-
tallisations from ethyl acetate are required to reduce the content
of 7ꢀ-isomer to an acceptable level (<0.1%) and remove other
impurities. From the Validation Campaign pure fulvestrant was
obtained in 28% overall yield from dienone 2. The sequence
of crystallisation process gives an approximately 55:45 ratio
of sulfoxide diastereoisomers.
Strategy in Commercial Manufacture. Although the two
noncrystalline intermediates are isolated and stored as solutions,
it is assumed for charging purposes at the next stage that the
yield of both intermediates is 100%. The rationale is that
impurities present will also consume reagents. A further
simplification involves using whole batches of one intermediate
in the next stage. For containment reasons, none of the
intermediate solids in the sequence of crystallisations at the final
stage is discharged from the filter; in each case the solid is
removed from the filter by dissolution in ethyl acetate and
transferred to the crystallisation vessel. In effect, the synthetic
route can be regarded as one continuous process involving about
500 unit operations with 11 distillations and 14 phase separa-
tions. During the Validation Campaign, yields of pure fulvestrant
were consistent, and all impurities were controlled within their
specification limits. Remarkably, all this was achieved with just
one In Process Control (a test for the water content of an ethyl
acetate solution prior to the first crystallization).
of elemental sulfur. An HPLC method developed specifically
to measure sulfur levels showed that the contaminant has a
response factor approximately 4 times greater than that of 26
at the detection wavelength (264 nm).
Suspicion was immediately directed at thiourea as the source
of the problem. At the end of the aromatisation step there is
unreacted cupric bromide (an excess of at least 0.37 mol equiv
is used in the process), and this is likely to oxidise thiourea
rapidly to formamidine disulfide 30 (Scheme 7).
A literature search revealed that 30 decomposes in aqueous
solution to give thiourea, cyanamide, and sulfur (Scheme 8).
The measured half-life for this decomposition decreases ex-
tremely rapidly with increasing pH. Based on data in the
literature, the half-life of formamidine disulfide at pH 2.43 is
126 min, but at pH 4.12 it is only 4 min.4
To check the relevance of this information to the manufac-
turing process for 26, solutions ranging in pH from 1 to 6 were
prepared by adding dipotassium hydrogen phosphate to a 2%
w/w aqueous solution of 30 (from the commercially available
hydrochloride salt). Within 5 min the pH dropped in all the
samples, indicating deprotonation and potential decomposition
of 30, consistent with the above mechanism. Above pH 4.5 a
heavy yellow precipitate of sulfur was also observed. On
standing overnight, sulfur was observed in all except the pH 1
sample.
Control of Sulfur in the Process. Reasons for batch-to-batch
variation and previous nondetection of sulfur were sought, and
clues were found in the hydrolysis step. There was a correlation
between extended ester hydrolysis times (which could be
attributed to an insufficiency of sodium hydroxide) and elevated
sulfur levels. This suggested that nondetection of sulfur in
previous manufacture might have been due to its decomposition
during the base-catalysed hydrolysis step.
S8 is known to undergo ring-opening by nucleophiles such
as cyanide and triphenylphosphine in a stepwise manner.5 The
sulfur ring is also cleaved by hydroxide to give ultimately
thiosulfate.6
Polysulfide ions (e.g., S52-) are thought to be intermediates
in the above process. On the basis of this information, it is
Conclusion
(4) Rio, L. G.; Munkley, C. G.; Stedman, G. J. Chem. Soc., Perkin Trans
II 1996, 2, 159.
An alternative synthetic route to fulvestrant 6 has been
developed and successfully implemented in commercial-scale
manufacture. Key to success has been understanding the origin
(5) (a) Bartlett, P. D.; Meguerian, G. J. Am. Chem. Soc. 1956, 78, 3710.
(b) Bartlett, P. D.; Davis, R. E. J. Am. Chem. Soc. 1958, 80, 2513.
(6) Licht, S.; Davis, J. J. Phys. Chem. B 1997, 101, 2540.
Vol. 14, No. 3, 2010 / Organic Process Research & Development
•
549

of certain impurities and then devising methods for their control
or removal. The alkyl bromide 14 used to introduce the side
chain is a potential source of analogous impurities in fulvestrant,
in particular 24 and 25. Since the impurities in 14 also prevent
reliable initiation of the Grignard reaction, it is important to
purify the material by distillation. Copper-catalysed 1,6-addition
of the alkyl side chain at low temperature affords a 2.5:1 mixture
of R- and ꢀ-isomers in good yield; the success of the reaction
depends on precise control of the temperature.
Cupric bromide is a highly effective reagent for aromatising
the A-ring. The newly formed phenol is prone to bromination,
giving related impurities 27a and 27b in fulvestrant that are
not readily removed. Formation of these bromo-impurities is
prevented by including acetic anhydride in the reaction mixture.
Upon completion of the aromatisation, copper is removed by
precipitation of its complex with thiourea. During the technology
transfer phase, elemental sulfur (a previously unseen impurity)
was detected in the aromatisation product at a significant level.
Sulfur, which is evidently produced by oxidation of thiourea
followed by hydrolytic cleavage of the resulting disulfide, is
readily transformed into water-soluble byproducts by increasing
the charge of aqueous alkali during the ester hydrolysis step.
Intermediates in the manufacture of fulvestrant are oils, and
they are processed further without any purification. The final
oxidation step affords fulvestrant (a crystalline compound) plus
significant amounts of accumulated organic-soluble impurities
from previous stages. Multiple crystallisations are required to
achieve the desired purity.
first batch of Grignard reagent in full-scale manufacture was
prepared on a larger scale and with an excess of magnesium
so that a heel (186 L) could be left in the reactor to initiate the
next batch. The following procedure describes the procedure
for the second and subsequent batches in a manufacturing
campaign.
Magnesium raspings (9.76 kg, 401 mol) and THF (738 L)
were added to the heel (186 L of approximately 0.4 M solution
plus excess magnesium) from the previous batch, and the
mixture was heated to 45 °C. Bromide 14 (160.4 kg, 401 mol)
was added in four portions; a temperature rise of 4-10 °C after
each addition indicated formation of the Grignard reagent. The
addition was followed by a line wash of THF (34 L). The
mixture was cooled to 10 °C, and excess magnesium was
allowed to settle out. The solution of Grignard reagent in THF
(minus the heel) was transferred to a cryogenic vessel via a
dip-leg, diluted with tetrahydrofuran (170 L) and cooled to -34
°C. Cuprous chloride (2.0 kg, 20 mol) was added, followed by
a solution of 17ꢀ-acetoxyestra-4,6-dien-3-one (2) (82 kg, 261
mol) in THF (350 L). The transfer of this solution took 3.5 h,
and the temperature was maintained at -34 °C throughout. The
addition was followed by a line wash of THF (35 L). The
reaction was quenched by adding a solution of acetic acid (70
kg, 1165 mol) in THF (79 L); the addition took 3 min and
resulted in a temperature rise of 15-20 °C. The mixture was
warmed to 20 °C, transferred to a standard reactor with a line
wash of THF (39 L), and then diluted with water (570 L).
Topanol (2 kg) was added to inhibit peroxide formation, and
THF was removed by distillation under atmospheric pressure.
The residual solution was cooled to <30 °C, diluted with more
water (250 L), and extracted with isohexane (410 L). The
mixture was passed through Harborlite filter aid followed by a
wash of isohexane (125 L), and the phases were separated. The
organic phase was washed with a solution of potassium chloride
(100 kg) in water (400 L). The resulting solution of 15 in
isohexane was used directly in the next stage. The yield of 15
(an approximately 2.5:1 mixture of 7R- and 7ꢀ-isomers) was
estimated by HPLC analysis to be 90-95% (RRT of 7R-isomer
1.00 and 7ꢀ-isomer 1.24). The strength of the solution (both
isomers) was approximately 27% w/w. For charging purposes
at the next stage, the yield was assumed to be 165.6 kg (100%).
Two batches of 15 were combined for use in the next stage.
7r-[9-(4,4,5,5,5-Pentafluoropentylsulfanyl)nonyl]estra-
1,3,5(10)-triene-3,17ꢀ-diol (26). A solution of 15 (assumed
content 331 kg, 521 mol as a mixture of 7R- and 7ꢀ-isomers)
was heated to remove most of the isohexane by distillation under
atmospheric pressure. The residue (approximately 500 L) was
diluted with acetonitrile (1420 L) and distillation was continued
until the batch temperature reached 85 °C (residual volume
approximately 600 L). The residual solution cooled to 20 °C
and its volume was adjusted to 1000 L with acetonitrile. A
solution of cupric bromide (275 kg, 1231 mol), lithium bromide
(75 kg, 864 mol) and acetic anhydride (61 kg, 597 mol) in
acetonitrile (1056 L) was prepared and added over 3 h to the
solution of 15 in acetonitrile, maintaining the temperature at
about 20 °C. A further portion of acetic anhydride (45 kg, 441
mol) was added followed by a line wash of acetonitrile (235
L) and the mixture was held at 20 °C for a further 4 h. The
Experimental Section
General. Starting materials, reagents, and solvents were
obtained from commercial suppliers and used without further
purification. IR spectra were recorded using a Thermo Electron
Avatar FT-IR instrument. Melting points were obtained with a
Mettler Toledo DSC822e instrument. NMR spectra were
obtained using a Bruker DPX 400 instrument; 1H spectra were
measured with reference to an internal standard of tetrameth-
ylsilane at 0 ppm, and 13C spectra were measured with reference
to the DMSO signal at 39.5 ppm. HPLC analyses were
performed with an Agilent 1100 instrument using the following
conditions: analysis of 15, Spherisorb ODS2 column (4.6 mm
×250 mm, 5 µm) at 20 °C and 220 nm with a flow rate of 2.0
mL/min and a mobile phase consisting of CH3CN and water
(49:1 v/v); analysis of 26, Spherisorb ODS2 column (4.6 mm
×250 mm, 5 µm) at 20 °C and 220 nm with a flow rate of 2.0
mL/min and a mobile phase consisting of CH3CN and water
(9:1 v/v), analysis of sulfoxide isomers of 6, Spherisorb CN
(4.6 mm ×250 mm, 5 µm) at 20 °C and 280 nm with a flow
rate of 2.0 mL/min and a mobile phase consisting of dichlo-
romethane, isohexane, and ethanol (33:66:1 v/v). Strengths (%
w/w) of isolated products (or their solutions) were measured
by HPLC analysis (or GC analysis in the case of bromide 14)
using a purified reference standard.
3-Oxo-7r-[9-(4,4,5,5,5-pentafluoropentylsulfanyl)nonyl]estr-
4-en-17ꢀ-yl Acetate (15). The reactor and associated equipment
were rigorously dried, and the process steps were carried out
under a nitrogen atmosphere. To initiate the first batch in full-
scale manufacture, a 0.4 M solution of Grignard reagent (about
37 L) was prepared separately using iodine as initiator. The
550
•
Vol. 14, No. 3, 2010 / Organic Process Research & Development

solution was then added over 1 h to a stirred mixture of toluene
(1000 L) and a solution of thiourea (150 kg, 1970 mol) in water
(1650 L) cooled to 10 °C. The addition was followed by a line
wash of acetonitrile (330 L) and toluene (70 L). The mixture
was cooled to 0 °C, and the pH was adjusted to about 3 by the
addition of dipotassium hydrogen phosphate (200 kg, 1148 mol).
After stirring for 1 h at 0 °C, the precipitated copper/thiourea
complex was removed by filtration, and the filter cake was
washed with toluene (1315 L). The phases in the filtrate were
separated, and the toluene phase containing diacetate 28 was
washed with water (100 L) at 2 °C and then with 10% w/v
sodium chloride solution (3 × 1000 L) at 60 °C. The toluene
solution was concentrated by distillation under reduced pressure
(200 mbar) to about 760 L. Methanol (990 L) was added to
the residual solution, and distillation was continued at atmo-
spheric pressure until 660 L of distillate had been collected.
The solution was cooled to 25 °C and diluted with more
methanol (600 L). 47% w/w Sodium hydroxide solution (124
kg, 1457 mol) was added, followed by a line wash of methanol
(55 kg), and the mixture was stirred at 30 °C for 5 h. When
hydrolysis was complete, the aqueous methanolic solution was
extracted with isohexane (3 × 890 L) to remove nonpolar
impurities carried through from the Grignard reaction. The
aqueous methanolic solution was diluted with more methanol
(100 L) and neutralised with glacial acetic acid (74 kg, 1232
mol) followed by a line wash of EtOAc (65 kg). Methanol was
removed by distillation at atmospheric pressure, leaving a
residue of about 500 L that was partitioned between water (437
L) and EtOAc (1190 kg) at 23 °C. The organic phase was
concentrated by distillation at atmospheric pressure to about
600 L, providing Fulvestrant PHS as an approximately 50%
w/w solution in EtOAc which is suitable for use directly in the
next stage. The yield of 26 (a mixture of 7R- and 7ꢀ-isomers)
was estimated by HPLC analysis to be 80-85% (RRT of 7R-
isomer 1.00 and 7ꢀ-isomer 1.05). For charging purposes at the
next stage, the yield was assumed to be 308 kg (100%).
7r-[9-(4,4,5,5,5-Pentafluoropentylsulfinyl)nonyl]estra-
1,3,5-(10)-triene-3,17ꢀ-diol (Fulvestrant) (6). An approxi-
mately 50% w/w solution of 26 (assumed content 308 kg, 521
mol as a mixture of 7R- and 7ꢀ-isomers) in EtOAc was diluted
with more EtOAc (695 kg) and glacial acetic acid (188 kg, 3131
mol). Aqueous hydrogen peroxide (17.5% w/v, 203 kg, 1045
mol) was added, followed by a line wash of water (60 kg), and
the mixture was stirred at 23 °C for 8 h. A further portion of
EtOAc (555 kg) was added, and excess hydrogen peroxide was
destroyed with a solution of sodium sulfite (100 kg) in water
(1080 L). The mixture was neutralised with a solution of 47%
w/w aqueous sodium hydroxide (279.6 kg) in water (525 kg),
keeping the temperature below 30 °C. The mixture was passed
through Harborlite filter aid followed by a wash of EtOAc (140
kg), and the phases were separated. The organic phase was
washed with water (615 L); to assist phase separation the
mixture was again passed through Harborlite filter aid followed
by a wash of EtOAc (140 kg). The organic phase was dried by
heating to reflux, collecting water in a binary separator, and
then was passed through Harborlite filter aid (to remove
insoluble inorganic material) followed by a wash of EtOAc (665
kg). The EtOAc solution was concentrated by distillation at
atmospheric pressure to about 920 L, then cooled to 10 °C over
10 h with seeding to promote crystallisation. The solid was
collected in a pressure filter and washed with cold EtOAc (208
kg). The crude product was removed from the pressure filter
by dissolution in hot EtOAc, and the resulting solution was
subjected to another crystallisation cycle. In total, four crystal-
lisations were carried out to achieve the required purity (<0.1%
7ꢀ-isomer). With each successive crystallisation, the EtOAc
solution was concentrated to a predetermined volume that
decreased in proportion to the estimated weight of product
present (based on laboratory experiments and HPLC analysis).
After the final crystallisation the product was dried in a stream
of nitrogen at 60 °C. The yield of pure fulvestrant at 100%
w/w strength was 88.4 kg (mean of five batches) which
represents 28% overall yield from dienone 2. Ratio of sulfoxide
A to sulfoxide B ) 46:54 by HPLC analysis (the retention times
of sulfoxide A and sulfoxide B are approximately 20.0 and 23.5
min); mp 104-112 °C by DSC analysis (sulfoxide A and
sulfoxide B have mp 102 and 117 °C).
1H NMR (400 MHz, 300 K, CDCl3) δ 0.78 (3H, s),
0.95-1.82 (23H, m), 1.91 (1H, ddd, J ) 12.2, 2.9, 2.7 Hz),
2.03-2.40 (8H, m), 2.56-2.90 (6H, m), 3.75 (1H, m), 6.56
(1H, d, J ) 2.37 Hz), 6.60-6.66 (1H, m), 6.61 (1H, s), 6.76
(1H, s), 7.12 (1H, d, J ) 8.62 Hz); 13C NMR (100 MHz, CDCl3)
δ 11.1, 14.7, 22.6, 22.7, 25.1, 27.3, 27.4, 28.6, 28.8, 29.0, 29.3,
29.4, 29.6 (t, JCF ) 21.9 Hz), 30.6, 33.3, 34.7, 37.0, 38.3, 42.1,
43.4, 46.5, 50.9, 52.5, 82.0, 113.1, 115.4 (tq, JCF ) 253 Hz, 38
Hz), 116.2, 118.9 (qt, JCF ) 285 Hz, 36 Hz), 127.0, 131.2,
136.9, 154.1.
2-(9-Hydroxynonyl)-2-thiopseudourea Hydrobromide (17).
A mixture of 1,9-nonanediol (140 kg, 874 mol), toluene (1646
L), and 48% w/w hydrobromic acid was heated under reflux
(about 93 °C) for 7 h. After cooling to 75 °C, the layers were
separated, and the upper organic phase was washed with water
(125 L) at 70-75 °C. The organic solution of 9-bromonony-
lalcohol was concentrated by distillation to 820 L and cooled
to ambient temperature. A solution of thiourea (63.5 kg, 834
mol) in isopropanol (387 kg) was added, and the mixture was
heated under reflux (about 86 °C) for 20 h. The reaction mixture
was cooled to 0 °C over 5 h and stirred at 0 to -4 °C for 4 h.
The solid was collected by filtration, washed with octane bp
ca. 120 °C (210 kg) and dried in a stream of cold nitrogen.
The yield of 17 at 100% w/w strength was 238.5 kg (91%).
9-(4,4,5,5,5-Pentafluoropentylsulfanyl)nonyl alcohol (13).
All process operations described below were performed under
a nitrogen atmosphere, and water was degassed with nitrogen
before use.
Triethylamine (86.1 kg, 852 mol) was added to a stirred
solution of 4,4,5,5,5-pentafluoropentanol (119 kg, 669 mol) in
acetonitrile (284.5 kg) at 20 °C. To the resulting mixture was
added a solution of methanesulfonyl chloride (87.0 kg, 760 mol)
in acetonitrile (214 kg), maintaining the temperature at 20 °C.
Analysis showed mesylation to be complete after about 2 h.
17 (190.8 kg at 95.0% w/w assay, 606 mol) was dissolved in
degassed water (545 L) at 40 °C and the solution was added
over 15 min to the mesylate ester solution prepared above,
followed by a line wash of water (90.5 L). Sodium hydroxide
solution (47% w/w, 310.3 kg, 3646 mol) was added over about
Vol. 14, No. 3, 2010 / Organic Process Research & Development
•
551

3 h at 40 °C, and the reaction mixture was heated under reflux
(72 °C) for 80 h. After cooling to 40 °C, the layers were
separated. The upper organic phase was diluted with octane bp
ca. 120 °C (680 kg), washed at 40-45 °C with water (181 L),
then with a solution of 36% w/w hydrochloric acid (77 kg, 759
mol) in water (362.5 L), and finally with water (2 × 181 L).
The organic solution was concentrated by distillation under
reduced pressure to 1075 L, cooled, and stirred at 10 °C for
1.5 h. The solid was collected by filtration, washed with octane
bp ca. 120 °C (270 kg) precooled to 10 °C, and dried in a stream
of cold nitrogen. The product required no further purification.
The yield of 13 (mp 40-42 °C) was 167.6 kg (82%).
nitrile (1:1 v/v, 2 × 324 L) and then with acetonitrile (162 L)
to remove the bulk of the triphenylphosphine oxide (all the
washes were carried out at 30 °C). The isohexane solution was
concentrated by distillation under reduced pressure (490 mbar
initially, reducing to 50 mbar) at <60 °C, leaving the crude
product as an oil (95.0% w/w by GC). The yield of crude 14 at
100% w/w strength 168.6 kg (87%). The product was purified
by distillation in a Wiped Film Evaporator (WFE) with a jacket
temperature of about 160 °C and a pressure of 2-3 mbar.
1H NMR (400 MHz, 300 K, CDCl3) δ 1.31 (8H, m), 1.43
(2H, m), 1.58 (2H, m), 1.89 (4H, m), 2.17 (2H, m), 2.51 (2H,
t), 2.59 (2H, t), 3.41 (2H, t); 13C NMR (68 MHz, 300 K, CDCl3)
δ 20.2 (t, J ) 3.4 Hz), 28.0, 28.5, 28.6, 28.9, 29.4, 29.4 (t, J )
24.5 Hz), 29.1, 31.1, 31.7, 32.6, 33.8, 115.7 (tq, J ) 251.5 Hz,
37.6 Hz), 119.1 (qt, J ) 285.4 Hz, 36.3 Hz).
1H NMR (400 MHz, 300 K, CDCl3) δ 1.30 (10H, m), 1.58 m
(4H, m), 1.65 (1H, s), 1.90 (2H, m), 2.17 (2H, m), 2.50 (2H,
t), 2.59 (2H, t), 3.60 (2H, t); 13C NMR (68 MHz, 300 K, CDCl3)
δ 20.2 (t, J ) 3.4 Hz), 25.6, 28.7, 29.0, 29.4, 29.4 (t, J ) 23.5
Hz), 29.2, 29.3, 31.1, 31.76, 32.60, 62.85, 116.2 (tq, J ) 253
Hz, 37.2 Hz), 118.9 (qt, J ) 285.6 Hz, 36.2 Hz).
Acknowledgment
We thank Ian Jones for providing interpretation of NMR
spectra. We also thank Brian Cox for helpful discussions
regarding the origin of elemental sulfur as an impurity in one
of the processes. We acknowledge the efforts of numerous
process chemists, analytical chemists, and process engineers in
ICI and AstraZeneca during the early stages of development
and also of production personnel in achieving successful scale-
up of the processes.
9-(4,4,5,5,5-Pentafluoropentylsulfanyl)nonyl Bromide (14).
Bromine (96.0 kg, 601 mol) was added over 2 h to a slurry of
triphenylphosphine (158.4 kg, 603 mol) in dry acetonitrile (290
L) at 20 °C. The mixture is stirred for 1 h to complete the
formation of dibromotriphenylphosphorane. Alcohol 13 (165.3
kg at 98% assay, 482 mol) was dissolved in acetonitrile (290
kg) at 40 °C, and the solution was added to the brominating
agent over 1 h at 23 °C followed by a wash of acetonitrile (30
kg). The mixture was stirred at 23 °C for 2 h to complete the
reaction. Triethylamine (78.0 kg, 771 mol) was added to the
mixture over 1.5 h, followed by a wash of acetonitrile (30 kg).
The mixture was diluted with water (162 L) and concentrated
by distillation under reduced pressure (200 mbar) at <60 °C to
750 L. The residual solution was extracted with isohexane (425
kg). The isohexane solution was washed with aqueous aceto-
Supporting Information Available
NMR spectra of fulvestrant. This material is available free
of charge via the Internet at http://pubs.acs.org.
Received for review December 4, 2009.
OP900315J
552
•
Vol. 14, No. 3, 2010 / Organic Process Research & Development