Development of a commercial process to produce oxandrolone

Article abstract of DOI:10.1021/op060231b

A manufacturing scale process for the preparation of the anabolic steroid Oxandrolone was developed. Key elements included the following: the bromination of methylandrostanolone with perbromide to give the 2-bromoketone in ca. 80% yield with minimal dehydration, subsequent elimination of the bromide with Li₂CO₃/LiBr to give the 2-enone in ca. 70% yield with minimal formation of methyltestosterone, and an ozonolysis procedure to give the penultimate intermediate in ca. 90% yield. The overall yield from methylandrostanolone to Oxandrolone using the described process was 45% as compared to the original Searle yield of 8%.

Full text of DOI:10.1021/op060231b

Organic Process Research & Development 2007, 11, 378−388
Development of a Commercial Process to Produce Oxandrolone
John E. Cabaj,* David Kairys, and Thomas R. Benson
Process Research and DeVelopment, Cedarburg Pharmaceuticals, Inc., 870 Badger Circle,
Grafton, Wisconsin 53024, U.S.A.
3
Abstract:
produce multikilogram quantities of USP Oxandrolone. The
development work that culminated in the final process is
summarized herein.
A manufacturing scale process for the preparation of the
anabolic steroid Oxandrolone was developed. Key elements
included the following: the bromination of methylandrostanolo-
ne with perbromide to give the 2-bromoketone in ca. 80% yield
with minimal dehydration, subsequent elimination of the
Discussion
The original Counsell/Pappo synthesis is summarized in
Scheme 1. The need to purify the R-bromoketone 3 as well
as the enone 4 by column chromatography, the use of the
2 3
bromide with Li CO /LiBr to give the 2-enone in ca. 70% yield
with minimal formation of methyltestosterone, and an ozonoly-
sis procedure to give the penultimate intermediate in ca. 90%
yield. The overall yield from methylandrostanolone to Oxan-
drolone using the described process was 45% as compared to
the original Searle yield of 8%.
4 4
highly toxic OsO and Pb(OAc) , and the low to moderate
yields in each of the four steps of the synthesis made use of
the Searle procedure unsuitable for scale-up. With these
issues in mind we proceeded to develop a scaleable method
based on the general approach of the Searle workers.
Bromination of Methylandrostanolone. Bromination of
methylandrostanolone (2) according to the Searle procedure
Br , NaOAc, acetic acid) resulted in multiple product
2
formation as determined by NMR analysis of the crude
reaction product. The major product isolated after column
chromatography was 6, the product of Wagner-Meerwein
Introduction
(
Oxandrolone 1 is an anabolic steroidal lactone currently
being used to promote weight gain and for the relief of bone
pain associated with osteoporosis. Oxandrolone was origi-
nally developed by the G.D. Searle Company during their
investigation of the anabolic and androgenic activity of
4
5
rearrangement of the 18-methyl group (Scheme 2). A control
experiment in which 2 was treated with HBr/acetic acid
showed that dehydration was indeed caused by the acidic
media generated during the bromination. Also, there is ample
precedent for such dehydrations in similar substrates.6
Bromination with bromine in DMF, according to the Searle
2-oxasteroids. The original synthesis was developed by
1
patent, gave several products by TLC and HNMR analysis
and therefore was not pursued further.
Due to the sensitivity of the 17-alcohol to dehydration, a
protecting group strategy was briefly examined (Scheme 3).
Subjecting the corresponding 17-acetate to the Searle reaction
conditions resulted in dehydration as well, while treatment
of the 17-trimethylsilyl ether of 2 with HBr (the byproduct
of the bromination reaction) led to desilyation.
1
Counsell, Pappo, and co-workers. In our synthetic planning
we felt that the Searle approach was the most direct. Other
approaches involving the late introduction of the 17-tertiary
alcohol, which would eliminate the issue of dehydration
during the early stages of the process (Vide infra), were not
pursued due to the inherent selectivity issues that would need
to be addressed between A and D ring functionality. Also,
the raw material for the Searle approach, methyland-
rostanolone (2), is readily available from a variety of vendors
Esterification of 2 with benzoyl chloride (PhCOCl,
pyridine, DMAP, CH Cl , reflux) gave the benzoate ester 9
2 2
in 33% yield after column chromatography. When the
1
benzoate ester was treated with bromine, HNMR analysis
of the crude reaction product showed that bromine incorpo-
ration had occurred at C-2; however, a significant amount
2
at a reasonable price. However, we found that the synthesis
(
3) For more recent syntheses of Oxandrolone, see: Ferraboschi, P.; Colombo,
D.; Prestileo, P. Tetrahedron: Asymmetry 2003, 14, 2781. Desai, S. R.;
Ray, D. W., Jr.; Sayed, Y. A. U.S. Patent Appl. US2003/0109721, 2003.
S a´ nta, C.; Tuba, Z.; Mah o´ , S.; Sz e´ les, J.; Ferenczi, K.; Horv a´ th, P.; L a´ ncos,
K.; Mester, T.; Trompler, AÄ . World Patent, WO 00/77025 2000.
4) For a discussion on the origin of the regioselectivity of the bromination,
see: Djerasi, C.; Scholz, C. R. J. Am. Chem. Soc. 1947, 69, 2404. Fieser,
L. F.; Dominguez, X. A. J. Am. Chem. Soc. 1953, 75, 1704. Berkoz, B.;
Chavez, E. V.; Djerassi, C. Steroids 1962, 1323.
described by the Searle workers was not suitable for large-
scale production. We therefore developed a new process,
based on the Searle approach, which was ultimately used to
(
*
To whom correspondence should be addressed. Telephone: (262)-376-1467.
Fax: (262)-376-1068. E-mail: john.cabaj@cedarburgpharma.com.
1) (a) Counsell, R. E.; Klimnstra, P. D.; Colton, F. B. J. Org. Chem. 1962,
7, 248. (b) Pappo, R.; Jung, C. J. Tetrahedron Lett. 1962, 9, 365. (c) Pappo,
(
2
(5) Monneret, C.; Choay, P.; Khuong-Huu, Q.; Goutarel, R. Tetrahedron Lett.
1971, 19, 3223.
R. U.S. Patent 3,128,283, 1964.
(2) Approximate cost of 2 is ca. $1,300/kg.
(6) Smith, D. M.; Steele, J. W. Can. J. Pharm. Sci. 1981, 16, 68.
3
78
•
Vol. 11, No. 3, 2007 / Organic Process Research & Development
10.1021/op060231b CCC: $37.00 © 2007 American Chemical Society
Published on Web 04/20/2007

Scheme 1. Counsell/Pappo synthesis of Oxandrolone
Scheme 2. Counsell bromination of methylandrostanolone
Scheme 4. THF trapping of HBr
Scheme 5. Process summary for the production of 3 (50-L
scale)
1
0
was added as a solution at 0-10 °C. We were delighted to
find that after workup and crystallization 3 was obtained in
Scheme 3. Protecting group study of bromination reaction
8
5% yield. Very little dehydration was evident by TLC
analysis.
A brief solvent screen showed that the reaction could also
be run in CH Cl , CHCl , and diethyl ether; however, the
2
2
3
reaction is not as clean in these solvents as it is in THF (TLC
result) or (in the case of diethyl ether) prohibitively slow.
Presumably, the THF is acting as an HBr trap as shown
below in Scheme 4 thus resulting in reduced dehydration of
1
1
of starting material remained as well. Due to the low yield
in the benzoylation this approach was not pursued further.
Other protecting groups were not examined.
the 17-alcohol in this solvent. The reversible nature of this
reaction would necessitate a process change prior to scaling
this step to reactor scale (Vide infra).
Hershberg and co-workers at Schering discovered that
attempted NBS mediated oxidation of a 3-hydroxy steroid
in the pregnane series gave a mixture of brominated ketones.
We therefore tried to brominate the silyl ether 8 using these
2 2
conditions (NBS, HBr/HOAc, t-BuOH, CH Cl , RT) and
obtained 3 contaminated with succinimide in 46% yield. No
reaction occurred without the addition of HBr. Presumably
the HBr is needed to promote enol formation. Not surpris-
ingly, the silyl ether was cleaved in the process. While the
NBS method appeared to have some promise, we did not
The bromination reaction was worked up by adjusting the
pH with Na CO followed by ethyl acetate extraction and
2 3
7
water washing. The ethyl acetate solution of 3 was then
concentrated to a set volume, and the resulting white solid
was further precipitated by adding heptane at ambient
temperature. The crystallization proved effective at removing
what little dehydration product was present. Drying of the
product at ca. 40 °C gave 3 in 70-90% yield. The initially
white product cake tended to discolor upon drying (light
pink). This was not a major issue on the 50-L scale as drying
times were fairly short and the downstream chemistry worked
reliably well with this material; however, the discoloration
was a concern since we did not know its origin. We were
fairly pleased that we were able to produce enough material
using this procedure to meet the customer’s initial bulk drug
requirements. Although the brominating agent used in the
process is considerably more expensive than molecular
pursue it further since bromination with perbromides gave
far superior results.8
The use of a perbromide for this transformation was first
reported by Doorenbos in which he obtained a 41% yield of
9
3
from 2 by using pyridinium tribromide in THF. In our
initial experiment, 2 was dissolved in THF and 1.1 equiv of
the perbromide phenyltrimethyl ammonium bromide (PTAB)
(10) Chodounsk a´ , H.; Slav ´ı kov a´ , B.; Kasal, A. Collect. Czech. Chem. Commun.
(
(
(
7) Hershberg, E. B.; Gerold, C.; Oliveto, E. P. J. Am. Chem. Soc. 1952, 74,
1994, 59, 435.
3
849.
(11) For an example of THF opening with aqueous HBr, see: Segi, M.;
8) LaCour, T. G.; Guo, C.; Bhanbaru, S.; Boyd, M. R.; Fuchs, P. L. J. Am.
Chem. Soc. 1998, 120, 692.
9) Doorenbos, N. J.; Dorn, C. P., Jr. J. Pharm. Sci. 1962, 51, 414.
Takahashi, M.; Nakajima, T.; Suga, S. Synth. Commun. 1989, 19, 2431.
For a related opening with BBr , see: Kulkarni, S. U.; Patil, V. D.; Wetherill,
3
R. B. Heterocycles 1982, 18, 163.
Vol. 11, No. 3, 2007 / Organic Process Research & Development
•
379

Table 1. Bromination of 2 using PTAB in various solvents
a
percent purityb
entry
conditions
percent yield (%)
comments
1
THF/0 °C
EtOAc extraction
50-L procedure)
75
92-94
discoloration upon drying
(
2
3
DME/0 °C
DME/0 °C
THF added in extraction
EtOAc/0 °C
64
70
not determined
not determined
solubility problem in workup
solubility problem in workup
4
5
6
7
73
97
solubility problem in workup
dehydration occurs
ethyl ketal formed
2
CH Cl
2
/0 °C
not determined
88
68
not determined
90
not determined
EtOH/RT
EtOH/RT
emulsion problem in workup
CH
EtOH/8% H
EtOH/14% H
2-propanol/RT
O knockout
2
Cl
2
extraction
O/RT
O/RT
8
9
0
2
84
82
70
89
91
86
reaction complete in 5.5 h
reaction slow (24 h)
-----
2
1
H
2
1
1
1
2
MeOH/RT
n-PrOH/RT
not determined
84
-----
not determined
ketal major product
not as clean as EtOH
reaction (TLC result)
a
b
Purity not included in yield calculation. Weight percent by HPLC analysis.
bromine, the ease of handling this solid compared to volatile
bromine outweighed the cost differential.12 A summary of
the 50-L process to produce 3 (ca. 1 kg per run) is shown in
Scheme 5.
was coated with an orange residue, as was the case when 3
was dried. A significant amount of the bromoalcohol gen-
erated in the bromination reaction presumably carries through
to the wet cake. Drying of the cake (40-45 °C) can then
cause closure of the bromoalcohol back to THF and liberation
of HBr. Variability in cake colors and extent of decomposi-
tion may be attributed to differences in cake washing
efficiency. Detection of the bromoalcohol in process streams
has been complicated due to its instability to GC analysis.
Since the decomposition was occurring during the drying
of 3, we first attempted to bypass the drying by performing
a solvent displacement with DMF, the solvent used in the
next step. The elimination reaction was then run using the
standard conditions (see discussion below). However, the
overall yields of 4 from methylandrostanolone were only
As the need to produce larger quantities of Oxandrolone
grew, it became obvious that production would have to be
performed at reactor scale. Since our 50-L procedure to
produce 3 worked adequately, we initially transferred this
procedure directly to the plant without any changes. The
reaction itself worked well at this scale; however, upon
drying at 35-45 °C, the intermediate 3 uniformly turned
from white to dark green. TLC and NMR analysis of this
material indicated that 3 had dehydrated upon drying giving
1
0 as the major product.13
3
9% on two trials. The reason for the low yield is currently
unknown; however, it appears that the bromoalcohol level
was sufficiently reduced as there was no evidence of
discoloration upon drying of the product enone 4.
The second option we examined to eliminate the dehydra-
tion was performing the reaction in a solvent other than THF.
This option was more appealing since the formation of the
bromoalcohol would no longer be an issue. The results of
the solvent screen using PTAB as the brominating agent are
shown in Table 1.
While brominations in DME and EtOAc went to comple-
tion and appeared to be fairly clean by TLC, the workup
was complicated by the insolubility of 3 in the reaction
solvent (entries 2 and 4). Reaction in DME with the addition
of THF after reaction quenching did not improve the
solubility to an appreciable extent (entry 3). Interestingly,
After a brief investigation it was determined that HBr had
caused the dehydration; however, the source of the HBr was
initially unknown since a base quench was used. Also,
quenching of residual bromine with metabisulfite or thio-
sulfate did not eliminate the discoloration during drying. We
then reasoned that the source of the HBr was most likely
4
-bromo-1-butanol, which is produced from ring opening of
THF by HBr (Scheme 4). To prove this theory a sample of
the bromoalcohol was made by simply bubbling HBr gas
through THF at the same reaction temperature as the
bromination reaction (ca. 0-5 °C). Upon heating of the pro-
duct from this reaction in a vacuum oven the entire oven
2 2
dehydration occurs to a much greater extent in CH Cl than
in any of the other solvents examined (entry 5). The
bromination reaction also works in alcoholic solvents, which
made isolation via water precipitation possible. We initially
ran this reaction in ethanol alone (entry 6). By TLC analysis
a new product in addition to 3 is also formed. Addition of
water (pH 1-2) and stirring for several hours lead to collapse
(
12) The cost of molecular bromine (Aldrich, 2.5 kg) is ca. $24.80/mol of Br
The cost of technical grade pyridinium tribromide (Aldrich, 500 g) is $138/
mol of Br
2
.
2
.
(
13) Hewett, C. L.; Gibson, S. G.; Redpath, J.; Savage, D. S. J. Chem. Soc.,
Perkin Trans. 1 1974, 12, 1432.
380
•
Vol. 11, No. 3, 2007 / Organic Process Research & Development

Scheme 6. Plant process summary for the production of 3
Table 2. Examination of the elimination conditions to
form 4
entry
conditions
4/11a
b
1
2
3
4
5
6
7
CaCO
3
, DMA , reflux
2/1 (100% crude yield)
c
c
c
KOH, MeOH, H
DBU, CH Cl , rt
collidine, 150 °C
2
O, 80 °C
multiple product formation
multiple product formation
multiple product formation
2
2
Scheme 7. Counsell elimination to generate 4
Li
Li
2
CO
CO
3
, DMF, H
2
O, 100-120 °C 1/5 (68% crude yield)
c
2
3
, DMF, 110-115 °C
multiple product formation
LiCl, DMF, reflux
ca. 2/1c
a
b
c
Determined by HPLC. DMA ) N,N-Dimethylacetamide. TLC result.
Table 3. Optimization of LiBr promoted elimination
of this less polar component to the desired bromide. Running
the reaction in the presence of water reduces the level of
this compound and gives fairly high quality material (entry
entry
conditions
4/11a
1
2
3
4
5
6
7
8
9
a
LiBr, Li
10-120 °C
LiBr, K CO
10-120 °C
LiBr, Na CO
10-120 °C
LiBr, CaCO , DMF,
2
CO
3
, DMF,
, DMF,
, DMF,
15/1 (91% crude yield)
8). We have tentatively assigned this less polar compound
1
1/1b
as the diethyl ketal of 3 but have not yet characterized it as
such. Presumably the presence of water leads to hydrolysis
of the ketal to give the ketone. Doubling the amount of water
slows the reaction substantially (entry 9). The sterically more
encumbered isopropanol gives slightly poorer quality material
after precipitation with water (entry 10). Use of methanol
gave mainly the dimethyl ketal of methylandrostanolone
2
3
1
1/1b
2
3
1
1/1b
3
1
10-120 °C
LiBr, Li
2
CO
3
, DMA,
46/1 (97% crude yield)
1
10-120 °C
multiple product formationb
LiBr, Li
110-120 °C
2
CO
3
, DMSO,
(entry 11) while n-propanol led to a more complicated
LiBr, pyridine, DMF,
10-115 °C
O, multiple product formationb
3 2
LiBr, Li CO , DMF, H
1/1b
reaction mixture (entry 12). Thus, an ethanol/water mixture
appeared to be the most promising solvent system for this
reaction and was therefore used for scaleup.
1
2
110-120 °C
LiBr, DMF, 110-120 °C
mainly dehydration product
While PTAB works as a brominating agent in ethanol,
the cost of production quantities of pyridinium tribromide
b
1
4
Determined by HPLC. TLC result.
(
PyHBr
3
) is approximately half the cost of PTAB. Also,
is soluble in ethanol thus making
unlike PTAB, PyHBr
3
convenient solution charging possible. We initially ran this
reaction in the absence of water and obtained a 69% yield
of 3 (98 wt % purity); however, this result was not
reproducible as lower purity 3 was obtained when this
reaction was repeated (88 wt % purity). Removal of ethanol
temperature of 106 °C was reached (ca. 80 °C over operation
temperature). Given this stability data, we felt comfortable
running this reaction at reactor scale.
Formation of Enone 4. Counsell and co-workers at Searle
2 3
reported that treatment of 3 with LiCl/Li CO in refluxing
(
ca. 1/3 of original 14 mL/g) after water precipitation gave
DMF gave a 93% yield of crude 4 contaminated with
approximately 5 to 10% methyltestosterone (11). After
1
a 91% yield of 3 but with substantially lower purity (75%
by weight). As was the case when PTAB was used for the
bromination, a higher spot on TLC was formed during the
reaction. This spot collapsed to 3 upon addition of aqueous
sodium thiosulfate (pH 1-2), which serves to quench residual
perbromide. When the reaction is run in ethanol/water
column chromatography and recrystallization they obtained
4 in only 39% yield. In our hands, we obtained a 60% yield
of crude 4 with a 4/11 ratio of only 2/1 (area percent) as
shown in Scheme 7. Since we needed kilogram quantities
of this intermediate, column chromatography, as a means of
purification, was not an option. Also, initial recrystallization
attempts did not improve the ratio to a useful level.
Since the Searle procedure was not suitable for scaleup
we examined different conditions to perform this transforma-
tion (Table 2). Treatment of 3 with both inorganic and
organic bases led to multiple product formation or poor ratios
mixtures (8% H
those of the 3 made using the 50-L procedure (82% yield,
2 wt % purity). The plant process that was used for the
2
O), the yield and purity are comparable to
9
production of 3 in 20-30 kg batches is summarized in
Scheme 6.
Initially we had ruled out alcoholic solvents, since we
were concerned about reaction between the alcohol and the
perbromide as is observed with bromine. However, addition
of pyridinium tribromide (PyHBr ) to ethanol lead to only a
(entries 1-4). Addition of water to help solubilize the Li
2
-
15
CO gave 11 as the major product (entry 5). Use of Li CO
3
2
3
3
in the absence of LiCl also led to multiple product formation
mild 4 °C exotherm over 2.5 h. Also, differential scanning
calorimetry analysis on pyridinium tribromide showed no
significant events (exothermic or endothermic) until a
while use of LiCl in the absence of Li CO gave a poor 4/11
ratio (entry 7).
It was not until the effect of the Lewis acid was examined
that the real breakthrough in this transformation occurred.
2
3
3
(14) The cost of PyHBr is $184/kg versus $370/kg for PTAB.
2
We first examined the use of MgBr in place of LiCl and
(
15) Urben, P. Bretherick’s Handbook of ReactiVe Chemical Hazards, Fourth
Edition; Kluwer Academic Publishers: 1999.
discovered that the ratio of 4/11 was improved to 10/1 (area
Vol. 11, No. 3, 2007 / Organic Process Research & Development
•
381

Scheme 8. Process to produce enone 4
2 3
Table 4. 4/11 ratio as a function of Li CO particle size
entry D[v, 0.10], µm D[v, 0.50], µm D[v, 0.90], µm 4/11a
1
2
3
3.74
2.92
200.93
14.96
17.52
627.39
51.01
26.69
827.73
13/1
99/1
1/1
percent HPLC). We were able to improve this ratio to 15/1
a
Determined by HPLC.
8
by using LiBr as the Lewis acid and Li
2
CO
3
as the base. A
screen of other bases in combination with LiBr was
undertaken. The results are summarized in Table 3. Yields
were only determined when synthetically useful 4/11 ratios
were obtained.
Scheme 9. Upjohn dihydroxylation to produce 5
Use of other carbonate bases as well as pyridine resulted
in a 1/1 mixture of the two regioisomers (entries 2-4 and
7
). The elimination can also be successfully performed in
DMA (dimethylacetamide) but not in DMSO. The addition
of water to help solubilize the Li CO (as in entry 5 of Table
) resulted in multiple product formation (entry 8). Finally,
when the reaction was run in the absence of base, the main
2
3
2
1
2
product was the17-gem-dimethyl olefin 12 (entry 9).
Other factors that presumably reduce the amount of
available carbonate such as the presence of heptane (lower
solubility of Li
less solvent, and decreased agitation tend to gave higher
levels of 11. Tight control over particle size of Li CO has
2 3
CO ) from insufficient drying of 3, use of
2
3
1
7
therefore been established.
Oxidation of 4 to 5. Oxidation of the enone 4 according
to the Searle patent (OsO /Pb(OAc) ) gave the acid 5 which
was converted into Oxandrolone using NaBH . The overall
yield from 4 using these conditions was only 7% after
chromatographic purification. The extremely low yield in
this reaction made it necessary for us to develop an
alternative oxidation protocol.
Presumably, the acidic nature of the reaction medium leads
to dehydration and methyl migration in this case, as in the
formation of 6. It should be noted that 12 is also one of the
impurities generated under normal processing conditions and
is removed after crystallization.
4
4
4
Based on the results in Table 3 the reagent combination
We initially investigated a two-step oxidation procedure
in which the intermediate diol would be isolated and then
converted to the acid via diol cleavage. When 4 was
2 3
of LiBr/Li CO was used for scale-up with DMF as the
solvent. The current operating conditions for this transforma-
tion are summarized in Scheme 8. Thus the bromide 3 is
2 3
heated in the presence of LiBr and Li CO until the reaction
18
dihydroxylated using the Upjohn method the diol 12 was
19
obtained in 82% yield. Oxidative cleavage with NaIO
4
gave
is complete. Acetic acid is added to adjust the pH and
dissolve the lithium salts. After ethyl acetate extraction and
azeotropic drying, heptane is added to precipitate the enone
5
fairly cleanly in 100% recovery contaminated with
inorganic salts (Scheme 9). Although we were able to
produce 5 efficiently via this route, we decided that the
extreme toxicity of OsO warranted the development of an
4
alternative oxidation method.
16
4
in 60-80% yield (93-97 wt % purity). The process
summarized above was used to produce 10-15 kg batches
of the key enone 4.
We initially examined accessing the diol 12 via an epoxide
2 3
Interestingly, the particle size of the Li CO was crucial
2 2
opening strategy (Scheme 10). Treatment of 4 with H O
to the success of the reaction. We first became aware of this
during a 100-gal reactor scale campaign in which an
approximately 1/1 ratio of 4/11 was obtained. Comparison
2
0
cleanly gave the epoxide 13 in 87% yield. However,
(
17) The formation of 11 during this reaction is not well understood; however,
Djerassi has shown that the rearrangement of 2,2-dibromoandrostan-17R-
ol-3-one to the 2,4-dibromo regioisomer is mediated by HBr (Djerassi, C.;
Scholz, C. R. J. Am. Chem. Soc. 1947, 69, 2404). In our system, it is
of the particle size of the Li
produced shows an interesting trend. In earlier laboratory runs
ACS grade Li CO (Aldrich catalog number 25,582-3,
9+%) that typically has a smaller particle size than bulk
Li CO (Aldrich catalog number 20,926-0, -40 mesh, 99%)
was used and an acceptable ratio of 4/11 was obtained (Table
, entries 1 and 2). In the failed production run the particle
size was much larger (entry 3).
2 3
CO and the ratio of 4/11
2
3
conceivable that reaction of the liberated HBr with the Li
on the degree of suspension of the base and ultimately to the particle size
of the Li CO . The larger particle size Li CO is most likely not as efficient
2 3
CO is dependent
9
2
3
2
3
2
3
at sequestering the HBr as material with a smaller particle size. Therefore,
rearrangement to the 4-bromo regioisomer is more prevalent when larger
particle size Li CO is used. Elimination of HBr in the 4-bromo regioisomer
2 3
would give 11. Also, the fact that this is a kinetic effect was proven when
the 4/11 ratio of a sample of 4 did not change (98/2) upon re-exposure of
the enone to the reaction conditions.
4
(
16) We initially isolated 4 by water knockout from DMF; however, drying of
the DMF wet cake was difficult, and recrystallization from ethyl acetate/
heptane was typically needed to provide adequate product purity.
(18) Pappo, R.; Jung, C. J. Tetrahedron Lett. 1962, 12, 365.
(19) VanRheenen, V.; Kelly, R. C.; Cha, D. Y. Tetrahedron Lett. 1976, 24, 1973.
(20) Pelc, B.; Hodkova, J. Collect. Czech. Chem. Commun. 1967, 32, 410.
382
•
Vol. 11, No. 3, 2007 / Organic Process Research & Development

Scheme 10. Attempted epoxide opening of 13
Table 5. Effect of temperature on ozonolysis of 4
entry temperature (°C)
percent yielda
purityb
1
2
3
4
5
-40 to -50
-30 to -40
-10 to -20
0 to -10
83
95
99
90
89
90
92
88
ambient
not determined multiple product
formationc
Scheme 11. Original Searle ozonolysis to produce
oxandrolone
a
b
Isolated yield after filtration and drying. Area percent as determined by
c
HPLC. TLC result.
Table 6. Ranging of temperature at plant accessible
temperatures
entry
temperature (°C)
percent yield
percent puritya
1
2
-15 to -10
-10 to 0
93
92
90
92
a
Determined by weight percent HPLC.
Scheme 14. Plant process for the production of 5
Scheme 12. Ozonolysis with dimethylsulfide workup
We initially ran the ozonolysis in methanol and used a
more traditional methylsulfide workup of the ozonide and
discovered that the reaction went to completion but gave
multiple products by TLC analysis (Scheme 12). Presumably
the keto dialdehyde 16 and the desired ester 15 were the
major components of the reaction mixture. Further oxidation
Scheme 13. Ozonolysis with caustic workup
4
of this mixture with NaIO gave mainly acid 5 by TLC
analysis. Other ozonide cleaving agents (P(OCH
Zn, etc.) were not tried.
3 3 3
) , PPh ,
4
attempted oxidative epoxide opening using Pb(OAc) or
21
periodic acid led to multiple product formation. Attempted
epoxide opening to generate the diol under acidic (H SO
HClO , or MeSO H) or basic (NaOH/DMSO) conditions also
led to multiple products by TLC analysis.
While the two-step oxidation method was somewhat
successful, it was cumbersome and gave product that
contained multiple unidentified impurities. While considering
different methods to workup the initial oxidation product of
the ozonolysis we were attracted to a procedure described
by Bailey for the workup of ozonolysis products using
2
4
,
4
3
With our lack of success in using the epoxide as an
entry to Oxandrolone we turned our attention toward an
ozonolytic approach.22 A thorough search of the patent
literature revealed that in fact this approach was used by a
group at Searle to synthesize Oxandrolone.23 The Searle
24
aqueous sodium hydroxide. This procedure would eliminate
the need for odiferous cleavage agents and would allow us
to purify the acid during the workup via standard acid base
patent describes the ozonolysis of 4 in CCl
4
to give the
chemistry. Also, further oxidation with NaIO would be
4
mixed anhydride 14, which is converted directly to Oxan-
unnecessary, as the sodium salt of the acid is the initial
drolone after reduction of the aldehyde and cyclization
product formed after base quenching. The fact that a group
at Searle had used this workup for a similar ozonolysis also
bolstered its attractiveness from a process standpoint.25 We
were extremely happy to find that ozonolysis of 4 in
methanol at -30 to - 40 °C, followed by quenching with
an aqueous solution of sodium hydroxide at ca. -15 °C, gave
the acid 5 in 90% yield after pH adjustment with HCl
(Scheme 13). This was a major breakthrough since we were
then able to perform the desired oxidation without using
(Scheme 11).
Alternatively, the Searle group obtained the methyl ester
1
5 when the ozonolysis was performed in a mixture of
methanol and methylene chloride. The use of the carcino-
genic CCl and the presumably thermal decomposition of
4
the potentially dangerous ozonide or peroxide intermediate
made the Searle procedure unsuitable for scaleup.
(
(
21) Nagarkatti, J. P.; Ashley, K. R. Tetrahedron Lett. 1973, 21, 4599.
22) For a review of production scale ozonolysis, see: Van Ornum, S. G.;
Champeau, R. M.; Pariza, R. Chem. ReV., submitted for publication.
23) Nysted, N.; Pappo R. U.S. Patent 3,109,016, 1963.
(24) Bailey, P. S.; Bath, S. S. J. Am. Chem. Soc. 1957, 79, 3120.
(25) Sollman P. B. U.S. Patent 3,281,431, 1966.
(
Vol. 11, No. 3, 2007 / Organic Process Research & Development
•
383

Scheme 15. Reduction and cyclization to produce Oxandrolone
Scheme 16. Preparation of seco-acid standard
comparable to that obtained using the 50-L procedure, we
were confidant that the reaction could be run at the higher
temperature. Also, as mentioned above, recrystallization
of 5 was part of the normal processing so as to ensure
high product purity prior to the final chemical step of the
process.
The issue of ozone scrubbing was not an issue on a
glassware scale where relatively small amounts of ozone
were used; however, reactor scale ozonolysis required
treatment of the effluent stream to destroy the excess ozone.
We found that an aqueous solution of sodium metabisulfite
to which the effluent stream was directed served this purpose
4 4
OsO and Pb(OAc) , which not only are highly toxic but
also present disposal problems. A temperature study was then
done to determine the optimum temperature for running the
ozonolysis (Table 5). As can be seen from the table, running
the reaction at temperatures of -50 to -30 °C gave high
quality 5 (entries 1 and 2), while running the reaction at
temperatures of 0 to -20 °C produced somewhat lower
quality product (entries 3 and 4). Thus the optimum operating
temperature for 50-L glassware production was determined
to be -30 to -40 °C since this temperature could be easily
maintained and the product purity was high.
(<1 ppm ozone detected after the quench solution).
With respect to the potential instability of the presumed
hydroperoxide intermediate 17, an ARSST (Advanced
Reactive System Screening Tool) on a sample of the
ozonolysis reaction mixture showed that the hydroperoxide
intermediate was stable at 25 to 100 °C, a temperature
approximately 100 °C higher than our operating tempera-
We also examined alternate reaction solvents for the
oxidation. Ozonolysis in ethanol led to multiple product
26
ture. A summary of the plant procedure that was used to
formation, while running the ozonolysis in ethyl acetate, CH
Cl , or isopropanol led to slow conversions presumably due
to the insolubility of 4 in these solvents. Reaction in a mixed
solvent system of CH Cl /acetic acid led to multiple product
2
-
2
2
2
formation. The conditions outlined in entry 2 were used to
produce 5 in 1-2 kg batches in 50-L glassware (73-78%
yield, 85-93 wt % purity). The lower yield and purity on
glassware scaleup are unknown at present; however, the
incorporation of a recrystallization of crude 5 routinely
produced material with a weight percent purity of greater
than 96%.
As was the case for production of intermediate 4, the need
to produce larger quantities of the acid 5 necessitated the
development of a reactor scale ozonolyis procedure. The
main issues that needed to be addressed before reactor scale
production could begin were the cooling limitations of the
reactors, the destruction of excess ozone, and the potential
thermal instability of the peroxide intermediate.
produce up to 10 kg batches of recrystallized 5 is shown in
Scheme 14. Thus, to the enone 4 in MeOH was added ozone
at -15 to 0 °C until the reaction was complete. NaOH was
then added to generate the sodium salt of 5. After pH
adjustment the product was filtered, dried, and recrystallized
to give pure 5.
Production of Oxandrolone 1. The final processing step
to produce Oxandrolone is shown in Scheme 15. While the
procedure itself is the most straightforward of the steps,
several points are worth mentioning.
Initially we performed the reduction in water alone, but
filtration of the product (after acidification) was extremely
slow. This was not the case in the water/ethanol mixed
solvent system. In 50-L glassware the reduction was per-
4
formed by adding solid NaBH portionwise to the solution
of the sodium salt of the carboxylic acid. Since solid addition
in a reactor setting was not possible, the commercially
The lowest achievable temperature at Cedarburg’s plant
is approximately -15 °C. From our initial development work
we determined that, while the product purity was 5-10%
lower when the reaction was run at higher temperatures, we
felt confident that the purity would be suitable for our needs
(Table 5). Furthermore, when acid 5 with a purity as low as
76 wt % was taken through the final two process steps
4
available caustic solution of NaBH was used instead with
no deleterious effects. Acidification of the sodium carboxy-
acceptable quality Oxandrolone was produced. A brief
ranging study was run to confirm the results in Table 5 and
better define the temperature range to be used in production
(26) In particular, when a sample of the reaction mixture was heated from 25 to
100°C at 2°C/min the internal temperature did not increase over the input
(Table 6). Since the yield and quality of the final product
temperature. Also, the internal pressure did not increase higher than the
background pressure caused by the boiling of the methanol solvent.
produced at a temperature range of -15 to 0 °C were
384
•
Vol. 11, No. 3, 2007 / Organic Process Research & Development

Scheme 17. Preparation of regioisomeric lactone impurity
Table 7. Levels of 17 and 19 present in crude and
recrystallized Oxandrolone
Summary and Conclusion
A scaleable process for the kilogram scale synthesis of
Oxandrolone was developed. Key elements in the process
include the following: a method to brominate methyland-
rostanolone without significant dehydration of the tertiary
alcohol, conversion of bromide 3 to the enone 4 with good
wt %
purification state
Oxandrolonea
wt % 18
0.17
wt % 20
0.16
crude
first
recrystallization
second
recryallization
97.98
99.36
0.10
0.02
∆-1/∆-4 selectivity, and an efficient reactor scale ozonolysis
100.1
none
detected
none
detected
procedure to produce the penultimate intermediate acid 5.
Work to improve the overall yield and to streamline the
process is underway.
a
As determined by HPLC.
Experimental Section
late salt (pH ca. 1-2) gives the seco-acid 18, which cyclizes
to give crude Oxandrolone. The ethanol/water wet cake is
then recrystallized from a 1/1 mixture of methanol/water (by
volume, ca. 15 L/kg of each). Recrystallization from ethyl
acetate, isopropanol, and acetone/water gave poor recoveries
and required the use of larger volumes of solvent. The
process outlined in Scheme 15 was used to produce multi-
kilogram batches of USP Oxandrolone in 81-97% overall
yield from 5 in a purity of 99-100 wt %.
General. All raw materials were used as supplied by
vendors. All reactions were performed under nitrogen in
glassware or glass-lined reactors as described below. All of
the reactions described were monitored by quantitative TLC
analysis using ethyl acetate/heptane or methanol/CH Cl
2 2
mixtures. Product purity was determined by reversed phase
HPLC using a Phenomenex Licrosphere RP 18 column (UV
3
detection at 198 nm) with a water/CH CN gradient as the
Synthesis of Potential Impurities. Several potential
impurities were synthesized for HPLC identification in the
final product. The most obvious potential impurity is the
seco-acid 18. A sample of this compound was made by
saponifying Oxandrolone with NaOH in methanol followed
by acidification with acetic acid (Scheme 16).
As was discussed above, a major impurity in the synthesis
of enone 4 was methyltestosterone (11, Scheme 8). Since
crystallization of 4 does not entirely remove 11, a certain
percentage (ca. 2-6% by area) was carried into the ozo-
nolysis reaction. We therefore thought it would be prudent
to process authentic 11 through the ozonolysis and reduction
sequence to determine the level of the regioisomeric lactone
mobile phase (2 mL/min). NMR data were obtained using a
Varian 400 MHz instrument. Chemical shifts are reported
in ppm. IR data were obtained using a Bruker Vector 22
FTIR. MS data were obtained using a Finnigan model TSQ
7000 instrument. Melting points were determined using a
Thomas Hoover Unimelt capillary melting point apparatus
and are uncorrected. Ozonolyses were performed using an
ozone generator from Pacific Ozone Technology (model
SGA-44, 122 g ozone/h output).
Preparation of 17-R-Methyl-5-R-androstan-17-â-ben-
zoyl-3-one (9). To a nitrogen purged three-neck 100 mL
round-bottom flask was added 2.01 g (6.61 mmol) of 2, 10
2 2
mL of CH Cl , 0.64 mL (7.9 mmol) of pyridine, 0.84 mL
2
0 present in the final product. Thus, ozonolysis of meth-
(7.2 mmol) of benzoyl chloride, and 81 mg (0.66 mmol) of
DMAP. After 2 h at ambient temperature no TLC analysis
inidicated the presence of only starting material. The reaction
mixture was heated at reflux for 9 h and then allowed to stir
at ambient temperature overnight. An additional 0.32 mL
(4.0 mmol) of pyridine, 0.42 mL (3.6 mmol) of benzoyl
chloride, and 41 mg (0.33 mmol) of DMAP were added,
and the reaction was stirred for an additional 4.5 h at reflux.
An additional 0.12 g (0.98 mmol) of DMAP was added, and
the reaction was stirred at reflux for an additional 19 h.
Starting material remained as determined by TLC analysis.
An additional 5.0 mL (62 mmol) of pyridine and 0.84 mL
yltestosterone followed by reduction of the resulting keto-
acid 19 and cyclization gave lactone 20 (Scheme 17). The
trans A/B ring juncture was assigned based on the Jax-ax and
J
ax-eq couplings of 12 and 4 Hz, respectively. The amounts
of the seco-acid 18 and lactone 20 impurities present in a
representative production lot of final product are shown in
Table 7. As can be seen from the table, one recrystallization
from methanol/water reduces the level of these two impurities
to e0.1%. A second recrystallization reduces them to below
the detection limit and gives Oxandrolone with a purity of
1
00 wt % (USP specification 98-102 wt %).
Vol. 11, No. 3, 2007 / Organic Process Research & Development
•
385

(
7.2 mmol) of benzoyl chloride were added, and the reaction
Preparation of 3 using Pyridinium Tribromide. To a
nitrogen purged 1 L four-neck round-bottom flask equipped
with a mechanical stirrer, thermocouple, addition funnel, and
nitrogen inlet were added 25.0 g (82.2 mmol) of 2, 100 mL
of absolute ethanol, and 25 mL of distilled water. To the
slurry were added 36.0 g (107 mmol) of pyridinium
tribromide in 200 mL absolute ethanol over 2 h at 24 to
was heated at reflux for an additional 3 h. TLC analysis
indicated only a trace of starting material remaining. The
solution was cooled to ambient temperature, and 10 mL of
water were added to the slurry. The mixture was extracted
with CH
were then washed successively with 1 M HCl (2 × 10 mL),
saturated NaHCO
(1 × 10 mL), and water (1 × 10 mL).
After drying over Na SO the solution was concentrated to
2
Cl
2
(2 × 10 mL). The combined organic extracts
3
3
27 °C. The orange color of the PyHBr persisted through
2
4
the addition. The orange slurry thinned to a fine orange
suspension, and eventually an off-white slurry formed. After
stirring for an additional 3.25 h, the reaction was complete
as determined by TLC analysis. To the off-white slurry were
added 6.0 g (24 mmol) of sodium thiosulfate pentahydrate
in 100 mL of water over 10 min (pH ca. 1). The snow white
slurry was then stirred for 30 min at room temperature. To
the slurry were then added 22.0 g (208 mmol) of sodium
carbonate in 200 mL of water over 5 min (pH 7-8). The
resulting slurry was stirred for 15 min and filtered. The white
cake was then washed successively with 100 mL of a cold
mixture of water/ethanol (0-5 °C, 2/1 (v/v)), 3 times with
125 mL of water, and once with ca. 100 mL of heptane.
The white solid was then dried at 50-60 °C in a high
vacuum oven for 19 h to give 24.0 g of 3 as an off-white
solid (77% yield, 92 wt % purity).
an oil. TLC analysis indicated the presence of pyridine salts.
The oil was then dissolved in ca. 50 mL of MTBE and ca.
5
mL of ethyl acetate. The solution was then washed
successively with ca. 20 mL of 1 M HCl, 40 mL of saturated
NaHCO , and finally ca. 20 mL of water. The organic layer
was then dried over Na SO and concentrated to give 3.52
3
2
4
g of a solid. A 1 g amount of the crude product was
chromatographed using 10% ethyl acetate/heptane as the
eluent to give 0.33 g of purified 9 as a white solid. Mp 198-
2
7
2
1
02 °C. H NMR (CDCl
3
): δ 8.0 (dd, J ) 8.0, 1.5 Hz, 2H),
.5 (t, J ) 8.0 Hz, 1H), 7.4 (t, J ) 8 Hz, 2H), 2.3 (m, 5H),
.1 (m, 2H), 1.2-1.8 (m, 13H), 1.6 (s, 3H), 1.1 (s, 3H), 1.0
13
(
s, 3 H), 0.9 (m, 2H). C NMR (CDCl
3
): δ 211.9, 165.7,
1
3
1
7
32.4, 131.9, 129.3, 128.2, 91.4, 53.7, 48.9, 47.0, 46.7, 44.6,
8.5, 38.1, 36.5, 35.9, 35.7, 32.2, 31.5, 28.8, 23.8, 21.5, 21.0,
4.9, 11.5 ppm. IR (KBr pellet): 1710, 1450, 1290, 1115,
Preparation of 17-â-Hydroxy-17-R-methyl-5-R-androst-
1-ene-3-one (4). To a 5-L four-neck round-bottom flask
equipped with a thermoprobe, nitrogen inlet adapter, and
mechanical stirrer were added 200 g (0.52 mol) of 3 and
1.2 L of DMF. To the slurry at room temperature were
-
1
+
21 cm . MS (CI-CH
4
): m/z 409 (M + H ), 287 (M -
+
BzOH + H ).
Preparation of 2-R-Bromo-17-R-methyl-5-R-androstan-
7-â-ol-3-one (3) Using Phenyltrimethylammonium Tri-
1
bromide. To a nitrogen purged 50-L four-neck round-bottom
flask equipped with an addition funnel, temperature probe,
nitrogen inlet adapter, and stirrer apparatus were added
2 3
added 42.6 g (0.58 mol) of ACS grade Li CO followed by
76.2 g (0.89 mol) of LiBr. The slurry was then heated to
110-115 °C and held at this temperature for 3 h after which
point TLC analysis indicated that the reaction was complete.
The slurry was cooled to an internal temperature of 35 °C,
and 1.4 L of ethyl acetate were added. An aqueous solution
of acetic acid (62 mL of acetic acid in 600 mL of water)
was then added over ca. 0.5 h. The reaction mixture was
then stirred at room temperature for ca. 8 h and then held at
0-10 °C in the refrigerator over the weekend. The reaction
mixture was then warmed to room temperature, the layers
were separated, and the aqueous layer was extracted three
times with 1.2 L of ethyl acetate. The combined organic
extracts were washed three times with 1.0 L of water and
then concentrated to a total volume of ca. 500 mL. To the
resulting slurry were added 1.8 L of heptane over 0.5 h at
room temperature. The slurry was held at an internal
temperature of 0-10 °C for ca. 2 h. The slurry was then
filtered, washed twice with 200 mL of 0-10 °C heptane,
and dried to constant weight in a high vacuum oven at ca.
45 °C to give 128 g (81% yield) of 17-â-hydroxy-17-R-
methyl-5-R-androst-1-ene-3-one (4) as an off-white solid
(90.3% weight percent purity, mp (uncorrected) 153-
6
50 g (2.14 mol) of methylandrostanolone (2) and 5.8 L of
THF. The solution was cooled to an internal temperature of
-10 °C. To the solution was added a THF solution of
0
phenyltrimethylammonium tribromide (884 g, 2.36 mol, 2
L of THF) over ca. 1 h while maintaining an internal
temperature of 0-10 °C. When the addition was complete
the resulting thick slurry was stirred for an additional 1 h at
0
-10 °C at which point TLC analysis indicated that the
reaction was complete. To the slurry was added an aqueous
solution of Na CO (377 g in 3.9 L of water) over ca. 1 h at
-10 °C. Additional water (4 L) was added, and the mixture
2
3
0
was allowed to warm to room temperature. The layers were
then allowed to settle for ca. 1 h and then separated. The
aqueous layer was extracted twice with 4 L of ethyl acetate,
and the combined organic extracts were washed with 4.5 L
of water. The organic layer was then concentrated using a
vacuum to a volume of ca. 2 L. To the resulting white slurry
at ca. room temperature were added 4.6 L of heptane over 2
h. The slurry was cooled to ca. 0-10 °C and held at this
temperature for ca. 2 h. The slurry was then filtered, washed
three times with 600 mL of 0-10 °C heptane, and dried to
a constant weight under a high vacuum at ca. 40 °C to give
1
a
154 °C, literature 155-156 °C ).
Preparation of Epoxide 13 (1,2-R-Epoxy-17-R-methyl-
5-R-androstan-17-â-ol-3-one). To a nitrogen purged 100 mL
three-neck round-bottom flask were added 2.01 g (6.66
mmol) of 4, 15 mL of methanol, and 0.86 mL (7.6 mmol)
6
1
1
21 g (76% yield, mp 195-197 °C, uncorrected, literature
96-198 °C ) of 2-R-bromo-17-R-methyl-5-R-androstan-
7-â-ol-3-one (3) as a pink solid. This material was used
1a
without purification in the elimination reaction.
2 2
of 30% aqueous H O . The solution was cooled to an internal
386
•
Vol. 11, No. 3, 2007 / Organic Process Research & Development

temperature of 16 °C at which point 0.20 mL (1.2 mmol) of
M HCl was added via syringe over ca. 15 s. The solution
darkened, and the internal temperature rose to 27 °C after
the addition. The solution was cooled further with an ice
bath and held at an internal temperature of ca. 0-15 °C for
over 1.5 h at an internal temperature of 5-7 °C (Note: there
may be vigorous evolution of hydrogen during the first and
second additions). When the addition was complete, the
solution was stirred for an additional 1 h at which point the
reaction was judged to be complete by TLC analysis. To
the solution containing the sodium salt of the seco-acid 22
was carefully added a 6.0 M solution of HCl (0.9 L of 37%
HCl, 0.9 L of water) over 4 h at an internal temperature of
1-9 °C (Note: the addition will generate hydrogen and be
highly exothermic). The resulting thick slurry was stirred at
1-3 °C for 4.75 h at which point the reaction was judged to
be complete by TLC analysis. As the seco-acid cyclizes, the
slurry thins and becomes easier to stir. The slurry was then
filtered and washed with a 0-10 °C solution of 1/1 EtOH/
water (664 mL of each), followed by 4.2 L of water and
finally 3.2 L of heptane. The solids were dried via high
vacuum at 40-50 °C to a constant weight to give 973 g of
crude Oxandrolone as a white solid (94% yield, weight
percent purity ) 100%).
6
1
h. Thin layer chromatography indicated that the reaction
was complete at this point. At an internal temperature of
2 °C, 5 mL of saturated Na SO were added followed
by an additional 10 mL of H O. The slurry was cooled to
-10 °C and stirred for ca. 30 min. The slurry was then
filtered, washed with ca. 20 mL of water, and dried at 40-
1
2
3
2
0
5
0 °C in a high vacuum oven to give 1.82 g (87% yield) of
19
the known epoxide 13 as a white solid.
Preparation of 17-â-Hydroxy-17-R-methyl-1-oxo-1,2-
seco-A-nor-5-R-androstan-2-oic Acid (5). Compound 4
(250 g, 0.828 mol) was dissolved in 2.5 L of MeOH, and
the resulting solution was cooled to -30 to -40 °C. Ozone
was then bubbled into the solution through a sparge tube.
After 6 h the reaction was judged to be complete by TLC
analysis. The solution was allowed to warm to -10 °C, and
an aqueous solution of NaOH (109 mL of concentrated
Recrystallization of Oxandrolone. To a 50-L four-neck
round-bottom flask equipped with a stir assembly, thermo-
couple, condenser, nitrogen inlet adapter, and heating mantle
were charged 962 g of crude Oxandrolone and 13 L of
methanol. The slurry was heated to an internal temperature
of 65 °C. To the solution 48.8 g of activated carbon were
added. The slurry was heated at 65-70 °C for 1.25 h and
hot filtered into a clean 50-L four-neck flask through a
methanol-moistened pad of celite. The original flask was
rinsed with 1.5 L of warm methanol. The filtrate contained
in the 50-L flask was heated to an internal temperature of
61 °C. Water (13 L) was then added to the solution at 61-
62 °C over 3.5 h. The slurry was then cooled to 35 °C over
6 h. The slurry was then pressure transferred to a clean 50-L
four-neck round-bottom flask in a cooling tub using 1.5 L
of water as a rinse (Note: this operation was incorporated
due to the potential for flask breakage upon attempted lifting
of the flask from the heating mantle into a cooling bath. The
slurry was then cooled to an internal temperature of 10 °C
over 1 h and held at 2-10 °C for 2 h. The slurry was then
filtered through a fritted funnel and washed consecutively
with a 5 °C 2/1 mixture of water/methanol (392 mL water,
200 mL methanol), 800 mL of water, and 2 L of heptane.
The highly crystalline solid was then dried to a constant
weight via high vacuum at 43-50 °C to give 817 g of USP
grade Oxandrolone (85% yield, 100.2% pure by weight, mp
(uncorrected) 221-225 °C, USP standard (uncorrected)
NaOH in 3.13 L of H
2
O) was added dropwise over 2 h. An
8
°C exotherm was observed during the addition. The
reaction mixture was allowed to warm to ambient temper-
ature at which point methanol was removed under reduced
pressure (2.8 L removed). The aqueous solution of the
sodium salt was washed with MTBE (2 × 1.0 L) and then
acidified to a pH of 4 using 6.0 M HCl. After stirring for
ca. 1 h, the slurry was filtered and washed with water (2 ×
300 mL) and 200 mL of n-heptane. The product was then
dried to constant weight via high vacuum at 40-60 °C to
give 211 g of 5 as an off-white solid (79% yield).
Recrystallization of 5. To a 50-L four-neck round-bottom
flask equipped with a condenser, thermocouple, nitrogen inlet
adapter, 2-L addition funnel, and heating mantle were
charged 2.96 kg of crude 5 and 20.7 L of methanol. After
heating the solution of 5 to an internal temperature of
5
5
9 °C, 25.7 L of water were added over a ca. 2 h period at
9-65 °C. The slurry was cooled slowly to room temperature
and then to 0-10 °C and held at this temperature for ca. 2
h. The slurry was filtered and washed with 1 L of a 0-5 °C
2/1 mixture of water/methanol (by volume), followed by
portionwise washing of the cake with an additional 6 L of
the cold 2/1 mixture. The highly crystalline white solid was
then dried to constant weight at 45 °C via high vacuum to
give 2.11 kg of 5 (86% yield, 100% pure by weight, mp
1
b
1b
(
uncorrected) ) 176-179 °C, literature 166-173 °C ).
212-216 °C, literature 235-238 °C ).
Preparation of Oxandrolone (17-â-Hydroxy-17-R-
Preparation of Seco-acid Impurity 18 (17â-Hydroxy-
17∝ -methyl-1-ol-1,2-seco-A-nor-5∝ -androstan-2-oic Acid).
To a 1-L four-neck round-bottom flask equipped with a stir
assembly, thermocouple, addition funnel, and nitrogen inlet
adapter were added 40.00 g (130.7 mmol) of Oxandrolone,
240 mL of methanol, and 240 mL of water. To the slurry at
ambient temperature was added 15 mL (290 mmol) of 50%
aqueous NaOH via addition funnel. After approximately 15
min the solids were completely in solution, and TLC analysis
of a pH 4 buffer/EtOAc quenched reaction aliquot (10%
methyl-2-oxa-5-R-androstane-3-one). To a 50-L four-neck
round-bottom flask equipped with a stir assembly, thermo-
couple, nitrogen inlet adapter, and 2-L addition funnel were
charged 1.10 kg (3.42 mol) of recrystallized 5, 8.4 L of
absolute ethanol, and 8.4 L of distilled water. The slurry
was cooled to an internal temperature of 8 °C at which point
a 25% solution of NaOH (198 mL 50% NaOH, 198 mL
water) was added over 16 min at an internal temperature of
8
-9 °C. To the resulting yellow solution of the carboxylate
were added 188 g (5.0 mol) of NaBH in four equal portions
4
2 2
MeOH/CH Cl eluent) indicated that the saponification was
Vol. 11, No. 3, 2007 / Organic Process Research & Development
•
387

complete. The methanol was then removed via rotary
evaporation, and the resulting aqueous solution of the
carboxylate salt was washed twice with 100 mL of MTBE.
To the aqueous solution of the carboxylate salt were added
mL of water were added. To the resulting thin suspension
were added 2.21 g (58.4 mmol) of solid NaBH portion-
wise at 4-5 °C over 20 min. After 1 h the reduction was
judged to be complete by TLC analysis (10% MeOH/CH
Cl eluent). To the reaction was then added 20 mL of 6 M
4
2
-
1
2
2 mL of acetic acid in 100 mL of water over 30 min at
3-26 °C. An additional 100 mL of water were added to
2
HCl at 5-15 °C (pH 1-2, paper) over approximately 30
improve the stirring of the thick slurry. The pH of the slurry
was 7-8 as determined by pH paper. The thick slurry was
then filtered, washed twice with 50 mL of water, and dried
to constant weight via high vacuum at 25-35 °C to give
min. After approximately 3.5 h the cyclization was judged
to be complete by TLC analysis (10% MeOH/CH Cl eluent).
2 2
Ethanol (110 mL) was removed via rotary evaporation, and
the resulting white slurry was held at approximately 5 °C
for 45 min. The slurry was then filtered and washed
successively with 30 mL of 0-10 °C ethanol/water (1/1 by
volume), water (5 × 25 mL), and finally heptane (50 mL).
The white solid was then dried overnight on a high vacuum
pump at 40-50 °C to give 8.8 g of the crude lactone 20 as
a white solid (74% yield). To a 500-mL round-bottom flask
were added 8.6 g of crude 20 and 26 mL of ethyl acetate.
While the resulting slurry was heating, 20 mL of additional
ethyl acetate were added portionwise to try and dissolve the
crude solid. The undissolved solid material was filtered off
with the aid of an ethyl acetate moistened celite bed, and
the filtrate was concentrated to a wet solid (23.0 g total
weight). To the solid was added ethyl acetate (14 mL), and
the slurry was heated. An additional 50 mL of ethyl acetate
were added in 5 mL portions to completely dissolve the
steroid. While the solution was cooling, 75 mL of heptane
were added dropwise over approximately 15 min. After
the addition was complete 75 mL of solvent were removed
via rotary evaporator, and the resulting slurry was cooled
with an ice-water bath. After approximately 1 h of cooling
the slurry was filtered and washed twice with 15 mL of a
0-10 °C heptane/ethyl acetate mixture (3/1 by volume). The
solids were then dried overnight on a high vacuum pump at
approximately 35-40 °C to give 6.1 g of recrystallized 20
20.3 g of 18 as a white powder (52% yield, relative retention
time 0.19). TLC analysis indicated that a trace of Oxan-
1
drolone was present in this sample. H NMR (DMSO): δ
3
1
1
1
3
.21 (AB
q
, J ) 12 Hz, 2 H), 2.40 (dd, J ) 16, 4 Hz, 1 H),
.95 (m, 1 H), 1.80-1.10 (m, 16 H), 1.08 (s, 3 H), 0.75 (m,
13
H), 0.72 (s, 3 H), 0.51 (s, 3 H). C NMR (DMSO): δ
75.2, 79.7, 63.2, 50.4, 45.1, 44.8, 39.8, 38.4, 36.8, 35.8,
5.6, 31.4, 31.2, 27.2, 26.1, 23.2, 20.4, 14.0, 11.4 ppm. IR
-
1
(KBr pellet): 3250, 1710 cm . MS (ESI/MS): m/z 347 (M
+
+
H
Na), 325 (M + H ), 307 (M- OH), 289 (M - OH -
O).
2
Preparation of Regioisomeric Lactone 20 (17â-Hy-
droxy-17∝ -methyl-4-oxa-5∝ -androstane-3-one). To a 1-L
four-neck round-bottom flask with a stir assembly, thermo-
couple, sparge tube, and nitrogen inlet adapter were added
14.8 g (49.0 mmol) of methyltestosterone and 300 mL of
methanol. The solution was cooled to an internal temperature
of -31 °C and then purged with argon gas via the sparge
tube. After sparging with argon for approximately 2 min,
the solution was sparged with ozone for a total of 1.5 h at
-
29 to -31 °C. The solution was then sparged with argon
for 5 min to remove the excess ozone. TLC analysis indicated
that the reaction was complete. The argon sparge was
continued for an additional 30 min. To the solution was
added 7 mL of 50% aqueous NaOH in 170 mL of water
over approximately 30 min at -30 to -10 °C. The solution
was slowly warmed to ambient temperature over ap-
proximately 2 h. Methanol (250 mL) was then removed via
rotary evaporation. The aqueous solution of the carboxylate
was then washed twice with 50 mL of MTBE, and the
aqueous layer was then stored overnight at 0-10 °C. The
next day the solution was warmed to 14 °C, and 26 mL of
1
as a white powder. H NMR (DMSO): δ 4.07 (s, 1 H), 4.03
(dd, J ) 12, 4 Hz, 1 H), 2.58 (m, 1 H), 2.42 (m, 1 H), 1.67
(m, 4 H), 1.48 (m, 6 H), 1.33 (m, 2 H), 1.15 (m, 3 H), 1.07
(s, 3 H), 0.90 (m, 1 H), 0.82 (s, 3 H), 0.75 (m, 4 H). 13
C
NMR (DMSO): δ 171.2, 83.6, 79.6, 50.6, 48.6, 38.9, 38.2,
35.3, 35.1, 32.1, 31.2, 28.4, 27.0, 26.4, 26.1, 22.9, 20.3, 14.1,
-
1
11.7 ppm. IR (KBr pellet): 3500, 1712 cm . MS (APCI):
+
+
m/z 307 (M + H ), 289 (M - H
2
O + H ).
6
M HCl were added dropwise (final pH 1-2, paper). The
Acknowledgment
product oiled out initially and then solidified. The slurry was
stirred for 2 h at ambient temperature. The slurry was then
filtered, washed with water (6 × 25 mL, pH of last wash
The authors thank Steve Guntz and the entire Cedarburg
production staff for their guidance in transferring the
laboratory procedure to the plant, Jeff Hutchison for initial
scale-up work, Maria Banach and Rada Dubishinsky for
assay development work, Dr. Gary Juneau (NMR Analysis
and Consulting) for NMR analysis, and Dr. Richard Pariza
for help in preparing this manuscript.
3
5
-4, paper) followed by heptane (50 mL), and dried at 45-
5 °C to give 13.8 g of the keto-acid 19 (87% yield), which
was used directly in the reduction. To a 500-mL four-neck
round-bottom flask equipped with a stir assembly, thermo-
couple, addition funnel, and nitrogen inlet adapter were added
1
2.53 g (38.91 mmol) of the crude keto-acid 19, 95 mL of
ethanol, and 95 mL of water. The slurry was cooled to
°C, and 2.3 mL (43 mmol) of 50% aqueous NaOH in 5
Received for review November 7, 2006.
OP060231B
4
388
•
Vol. 11, No. 3, 2007 / Organic Process Research & Development