Effect of Impurities on Estradiol Crystallization in a Sustained-Release Implant G.F. NEEDHAM*’, R. R. PFEIFFER’*, G.L. ENGEL**, 8. S.RUTHERFORD*§,AND D. J. ALLEN’ Received February 26, 1991, from ‘Lilly Research Laboratories, a Division of Eli Lill and Company, P.O. Box 708, Greenfield, *Present address: Lilly {esearch Laboratories, Division of Eli Lilly and Company, IN 46140. Accepted for publication January 6, 1992. §Present address: Dow Elanco and Company, 4040 Vincennes Circle, Suite 601, Indianapolis, Lilly Corporate Center, Indianapolis, IN 46285. IN 46268.
Abstract 0 During the development of a silicone rubber implant for the delivery of estradiol 17-p some batches of implants made from a
certain lot of commercial estradiol inexplicably developed surface crystals of estradiol after several days of storage. An impurity profile was obtained for 28 lots of estradiol by a newly developed HPLC method. One or more impurities may have had a role in the spontaneous crystal growth on the surface of the implants, because the one lot of estradiol that initially had surface crystals on aging produced acceptable implants after multiple recrystallizations. Attempts to isolate suspected impurities for characterization were unsuccessful. During the manufacture of the implants, temperatures sufficient to melt the estradiol (mp, 173-179 “C) were used. It was expected that, upon implant cooling, melted impure estradiol would form a thermodynamically more active (i.e.,noncrystalline) physical form. This metastable form could have migrated to the implant surface, where ambient conditions favored crystallization. Because melted estradiol of a higher purity tended to crystallize more readily, it was less likely to form a glass upon cooling. The phenomenon of surface crystallization was limited to one lot of estradiol with t h e highest level of impurities. Data from differential scanning calorimetry studies supported this conclusion.
Estradiol 17-/.3has been used in many products for a wide range of therapeutic effects, including growth promotion in beef cattle.14 During the development of a silicone rubber implant for the sustained release of estradiol in cattle, a peculiarity was observed in the in vitro release from certain batches of implants in which the only variable in the component materials was the lot of estradiol. The initial release of the drug was significantly higher for some batches of implants. Implants made from one steroid lot seemed acceptable, but when inspected visually and by scanning electron microscopy several days after manufacture, the implants were covered with a thin layer of estradiol crystals (Figure 1A). Most lots of steroid used resulted in implants that did not develop any surface crystals, even after many months of storage (Figure 1B). Because this steroid-lot-specific condition might have produced an undesirable “burst” release in vivo, an attempt was made to explain the surface crystallization phenomenon.
Experimental Section Preparation of the implants included a step in which a dispersion of estradiol in poly(dimethylsi1oxane) was heat cured onto a nonmedicated silicone polymer substrate under conditions that melted some of the steroid. To understand the observed phenomena, we determined the chemical purities and thermal properties of various lots of estradiol. Determinations of estradiol purity by gas chromatography (USP method) indicated that all of the raw materials were USP quality. The USP method, which involves chemical derivatization of estradiol, ~ ~ ~ e s drug s e s potency rather than detecting impurities. Therefore, a gradient HPLC method was developed for the detection of small and 1012 1 Journal of Pharmaceutical Sciences Vol. 81, No. 10, October 1992
variable amounts of impurities in estradiol. An Altex Spherieorb I1 octyldecylsilane column (5 pm; 250 x 4.6 mm) w m used. The mobile-phase gradient changed linearly from 50%methanol in water at injection to 90%methanol in water after 10 min and then remained unchanged during the last 10 min of the chromatography. Estradiol and related compounds were detected spectroscopically a t 280 nm. The flow rate was 1 mumin, and the injection volume was 10 pL of a 1-mg/mL solution of estradiol in methanol. Numerous lots of estradiol, purchased from three suppliers, were examined by this method. All lots were analyzed as received. One purchased lot, associated with unacceptable implants, was purified by recrystallization. The existence of polymorphsb7 provided a basis for purifying the estradiol. The steroid waa recrystallized from methanol (as the methanol solvate) and then from ethanol-water (as the hemihydrate). The phases were identified by powder X-ray diffraction. With this procedure, the total impurity level was reduced to 0.1%) when calculations were done on a normalized area percent basis. All lots of estradiol from supplier B and all but one lot of estradiol from supplier A made acceptable implants. Implants made with one lot of estradiol from supplier A (labeled A*) had crystals on the surface of the stored product and were unacceptable. When this lot of drug was purified by recrys0022-3549/92/1OOO- 1012$O2.50/0 0 1992. American PhannaceuticalAssociation
Table CEstradlol Purlty Deterrnlned by HPLC
Supplier
Amount of Material Eluting at Indicated Retention Time (rnni)=' 12.95 13.73 14.00
14.67
17.86
A B
1.13
0.90
B
0.4 1.14 Tr Tr 0.16 1.23 Tr 2.15 1.09 Tr 0.09 1.48 Tr 0.59 1.39 0.65 1.48 Tr 0.16 0.12
97.62 100.00 99.60 97.89 100.00 100.00 99.84 97.79 100.00 96.50 98.36 100.00 99.91 97.51 100.00 98.14 97.81 98.66 97.63 100.00 99.84 99.88 100.00 99.88 96.24 97.17 99.92 97.53 97.37
A B B C A B A
A
0.11 Tr
B B
A B A A A A B B B B B A'b
LRL B A A
Tr
Tr Tr
0.14 0.64 0.28
0.12 2.39 1.17 0.08 1.33 1.28
0.34 0.30
0.29 0.95 0.25 0.37
Tr 0.32 Tr 0.39
0.60 0.30 0.35 0.34
0.67
0.69 0.28 0.30 0.65 1.28 0.48 0.69 0.51
0.63 0.72 0.79 0.72
Amount of Impurities, % 2.38 0.00 0.40 2.11 0.00 0.00 0.16 2.21 0.00 3.50 1.64 0.00 0.09 2.49 0.00 1.86 2.19 1.34 2.37 0.00 0.16 0.12 0.00 0.12 3.76 2.83 0.02 2.47 2.63
a Values are area percents of total HPLC chromatogram. Each row represents the HPLC results from a single estradiol lot. A blank space indicates that no material was detected; Tr, trace. Only one lot of estradiol from supplier A (labeled A') produced unacceptable implants. It was recrystallized and relabeled LRL.
Flgure 1-Scanning electron micrographs of the surfaces of an unacceptable implant (A) and an acceptable implant (B). The estradiol crystals are evident in panel A.
tallization (and relabeled LRL), acceptable implants were produced. The shavings from the acceptable implants showed a typical estradiol DSC melting transition (173-179 "C) (Figure 3A). Upon cooling and reheating, a broad crystallization exotherm (-14&160 "C) followed by a normal remelting endotherm was observed during the second heating cycle. Shavings from the unacceptable implants (Figure 3B) showed a normal initial thermogram. However, when a sample containing the melted estradiol material was cooled and reheated, no crystallization exotherm was observed, nor was there evidence of any melting endotherm in the range in
which estradiol melts. These results indicated that melted estradiol dispersed in the silicone had cooled to an amorphous glass under the conditions of the DSC experiments. Upon subsequent reheating, however, the acceptable estradiol samples recrystallized, but the unacceptable estradiol samples remained in the amorphous state. Although the conditions of the DSC experiments did not duplicate the thermal history of the implants during their manufacture, we interpreted the DSC results as evidence that the higher level of impurities in the unacceptable steroid lots interfered with the prompt local recrystallization of a t least some significant fraction of estradiol during the manufacture of these implants. This interference resulted in the slow crystallization of estradiol on the surface of the silicone implants. Additional support for this interpretation was the absence of surface crystals when the unsatisfactory lot of estradiol was purified and subsequently incorporated into the implants by the manufacturing procedure. Implants were kept under normal storage conditions and examined weekly by scanning electron microscopy for surface crystals. Examination of implants continued for over 1month. DSC data for shavings made from purified material incorporated into the implants were characteristic of an acceptable product. On the basis of the HPLC data, we could not determine whether the surface crystallization phenomenon was due to single, specific impurities or the total level of impurities. We believe that the effect was due to one or more unidentified impurities. Attempts to isolate and characterize impurities were unsuccessful. We have demonstrated a n example of a product in which the drug purity should be carehlly monitored on a lot-blot Journal of Pharmaceutical Sciences I 1013 Vol. 81, No. 10, October 1992
a
A
30
50
70
90
110
130
150
170
190
210
230
Degrees Centigrade
B
B
b
30
50
70
90
110
130
150
170
190
210
230
Degrees Centlgrade
Figure 2-HPLC results for estradiol obtained from two suppliers. The impurities can be seen in panel 6 . STARIF, start injection; IF, start data collection; ST, stop data collection; S, peak off scale.
Flgure 3-DSC results for estradiol in the shavings from an acceptable implant (A) and an unacceptable implant (6).Key: (a) cooling, 5 "C/min for panel A and 10 "Clmin for panel 8 ;(b)first run, 5 "C/min; (c) second run, 5 "C/min.
basis throughout the lifetime of the product. We urge investigators to be especially careful to define purity and drug behavior when a drug is exposed to near-melting temperatures during manufacture.
3. Bequette, R. J.; Hobbs, L. G.; Scott, J. A, U.S. Patent 4 436 738, 1984; Chem. Abstr. 1984,100, 215546b. 4. Ferguson, T. H.; Needham, G . F.; Wagner, J. F. J . Controlled Release 1988,8,45. 5. Kuhnert-Brandetatter, M.; Winkler, H. Sci. Phurm. 1976, 44, suppi. 4,288. 6 . Kuhnert-Brandstatter, M.; Winkler, H. Sci. Phurm. 1976, 44, Suppl. 3, 191. 7. Kuhnert-Brandstatter,M. Phurm. Ind. 1977,39, Suppl. 4,377. 8. Kuhnert-Brandstatter, M.; Winkler, H. Sci. Pharm. 1976, 44, Suppl. 3, 177.
References and Notes 1. Nell, G. Wien. Klin. Wochenschr. 1989, 101, 398. 2. Powers, M. S.; Schenkel, L.; Darley, P. E.; Good, W. R.; Balestra, J. C.;Place,V. A. Muench.Med. Wochenschr. 1988,130,Suppl.1, S7.
1014 I Journal of Pharmaceutical Sciences Vol. 87, No. 10, October 1992
Abstract 0 During the development of a silicone rubber implant for the delivery of estradiol 17-p some batches of implants made from a
certain lot of commercial estradiol inexplicably developed surface crystals of estradiol after several days of storage. An impurity profile was obtained for 28 lots of estradiol by a newly developed HPLC method. One or more impurities may have had a role in the spontaneous crystal growth on the surface of the implants, because the one lot of estradiol that initially had surface crystals on aging produced acceptable implants after multiple recrystallizations. Attempts to isolate suspected impurities for characterization were unsuccessful. During the manufacture of the implants, temperatures sufficient to melt the estradiol (mp, 173-179 “C) were used. It was expected that, upon implant cooling, melted impure estradiol would form a thermodynamically more active (i.e.,noncrystalline) physical form. This metastable form could have migrated to the implant surface, where ambient conditions favored crystallization. Because melted estradiol of a higher purity tended to crystallize more readily, it was less likely to form a glass upon cooling. The phenomenon of surface crystallization was limited to one lot of estradiol with t h e highest level of impurities. Data from differential scanning calorimetry studies supported this conclusion.
Estradiol 17-/.3has been used in many products for a wide range of therapeutic effects, including growth promotion in beef cattle.14 During the development of a silicone rubber implant for the sustained release of estradiol in cattle, a peculiarity was observed in the in vitro release from certain batches of implants in which the only variable in the component materials was the lot of estradiol. The initial release of the drug was significantly higher for some batches of implants. Implants made from one steroid lot seemed acceptable, but when inspected visually and by scanning electron microscopy several days after manufacture, the implants were covered with a thin layer of estradiol crystals (Figure 1A). Most lots of steroid used resulted in implants that did not develop any surface crystals, even after many months of storage (Figure 1B). Because this steroid-lot-specific condition might have produced an undesirable “burst” release in vivo, an attempt was made to explain the surface crystallization phenomenon.
Experimental Section Preparation of the implants included a step in which a dispersion of estradiol in poly(dimethylsi1oxane) was heat cured onto a nonmedicated silicone polymer substrate under conditions that melted some of the steroid. To understand the observed phenomena, we determined the chemical purities and thermal properties of various lots of estradiol. Determinations of estradiol purity by gas chromatography (USP method) indicated that all of the raw materials were USP quality. The USP method, which involves chemical derivatization of estradiol, ~ ~ ~ e s drug s e s potency rather than detecting impurities. Therefore, a gradient HPLC method was developed for the detection of small and 1012 1 Journal of Pharmaceutical Sciences Vol. 81, No. 10, October 1992
variable amounts of impurities in estradiol. An Altex Spherieorb I1 octyldecylsilane column (5 pm; 250 x 4.6 mm) w m used. The mobile-phase gradient changed linearly from 50%methanol in water at injection to 90%methanol in water after 10 min and then remained unchanged during the last 10 min of the chromatography. Estradiol and related compounds were detected spectroscopically a t 280 nm. The flow rate was 1 mumin, and the injection volume was 10 pL of a 1-mg/mL solution of estradiol in methanol. Numerous lots of estradiol, purchased from three suppliers, were examined by this method. All lots were analyzed as received. One purchased lot, associated with unacceptable implants, was purified by recrystallization. The existence of polymorphsb7 provided a basis for purifying the estradiol. The steroid waa recrystallized from methanol (as the methanol solvate) and then from ethanol-water (as the hemihydrate). The phases were identified by powder X-ray diffraction. With this procedure, the total impurity level was reduced to 0.1%) when calculations were done on a normalized area percent basis. All lots of estradiol from supplier B and all but one lot of estradiol from supplier A made acceptable implants. Implants made with one lot of estradiol from supplier A (labeled A*) had crystals on the surface of the stored product and were unacceptable. When this lot of drug was purified by recrys0022-3549/92/1OOO- 1012$O2.50/0 0 1992. American PhannaceuticalAssociation
Table CEstradlol Purlty Deterrnlned by HPLC
Supplier
Amount of Material Eluting at Indicated Retention Time (rnni)=' 12.95 13.73 14.00
14.67
17.86
A B
1.13
0.90
B
0.4 1.14 Tr Tr 0.16 1.23 Tr 2.15 1.09 Tr 0.09 1.48 Tr 0.59 1.39 0.65 1.48 Tr 0.16 0.12
97.62 100.00 99.60 97.89 100.00 100.00 99.84 97.79 100.00 96.50 98.36 100.00 99.91 97.51 100.00 98.14 97.81 98.66 97.63 100.00 99.84 99.88 100.00 99.88 96.24 97.17 99.92 97.53 97.37
A B B C A B A
A
0.11 Tr
B B
A B A A A A B B B B B A'b
LRL B A A
Tr
Tr Tr
0.14 0.64 0.28
0.12 2.39 1.17 0.08 1.33 1.28
0.34 0.30
0.29 0.95 0.25 0.37
Tr 0.32 Tr 0.39
0.60 0.30 0.35 0.34
0.67
0.69 0.28 0.30 0.65 1.28 0.48 0.69 0.51
0.63 0.72 0.79 0.72
Amount of Impurities, % 2.38 0.00 0.40 2.11 0.00 0.00 0.16 2.21 0.00 3.50 1.64 0.00 0.09 2.49 0.00 1.86 2.19 1.34 2.37 0.00 0.16 0.12 0.00 0.12 3.76 2.83 0.02 2.47 2.63
a Values are area percents of total HPLC chromatogram. Each row represents the HPLC results from a single estradiol lot. A blank space indicates that no material was detected; Tr, trace. Only one lot of estradiol from supplier A (labeled A') produced unacceptable implants. It was recrystallized and relabeled LRL.
Flgure 1-Scanning electron micrographs of the surfaces of an unacceptable implant (A) and an acceptable implant (B). The estradiol crystals are evident in panel A.
tallization (and relabeled LRL), acceptable implants were produced. The shavings from the acceptable implants showed a typical estradiol DSC melting transition (173-179 "C) (Figure 3A). Upon cooling and reheating, a broad crystallization exotherm (-14&160 "C) followed by a normal remelting endotherm was observed during the second heating cycle. Shavings from the unacceptable implants (Figure 3B) showed a normal initial thermogram. However, when a sample containing the melted estradiol material was cooled and reheated, no crystallization exotherm was observed, nor was there evidence of any melting endotherm in the range in
which estradiol melts. These results indicated that melted estradiol dispersed in the silicone had cooled to an amorphous glass under the conditions of the DSC experiments. Upon subsequent reheating, however, the acceptable estradiol samples recrystallized, but the unacceptable estradiol samples remained in the amorphous state. Although the conditions of the DSC experiments did not duplicate the thermal history of the implants during their manufacture, we interpreted the DSC results as evidence that the higher level of impurities in the unacceptable steroid lots interfered with the prompt local recrystallization of a t least some significant fraction of estradiol during the manufacture of these implants. This interference resulted in the slow crystallization of estradiol on the surface of the silicone implants. Additional support for this interpretation was the absence of surface crystals when the unsatisfactory lot of estradiol was purified and subsequently incorporated into the implants by the manufacturing procedure. Implants were kept under normal storage conditions and examined weekly by scanning electron microscopy for surface crystals. Examination of implants continued for over 1month. DSC data for shavings made from purified material incorporated into the implants were characteristic of an acceptable product. On the basis of the HPLC data, we could not determine whether the surface crystallization phenomenon was due to single, specific impurities or the total level of impurities. We believe that the effect was due to one or more unidentified impurities. Attempts to isolate and characterize impurities were unsuccessful. We have demonstrated a n example of a product in which the drug purity should be carehlly monitored on a lot-blot Journal of Pharmaceutical Sciences I 1013 Vol. 81, No. 10, October 1992
a
A
30
50
70
90
110
130
150
170
190
210
230
Degrees Centigrade
B
B
b
30
50
70
90
110
130
150
170
190
210
230
Degrees Centlgrade
Figure 2-HPLC results for estradiol obtained from two suppliers. The impurities can be seen in panel 6 . STARIF, start injection; IF, start data collection; ST, stop data collection; S, peak off scale.
Flgure 3-DSC results for estradiol in the shavings from an acceptable implant (A) and an unacceptable implant (6).Key: (a) cooling, 5 "C/min for panel A and 10 "Clmin for panel 8 ;(b)first run, 5 "C/min; (c) second run, 5 "C/min.
basis throughout the lifetime of the product. We urge investigators to be especially careful to define purity and drug behavior when a drug is exposed to near-melting temperatures during manufacture.
3. Bequette, R. J.; Hobbs, L. G.; Scott, J. A, U.S. Patent 4 436 738, 1984; Chem. Abstr. 1984,100, 215546b. 4. Ferguson, T. H.; Needham, G . F.; Wagner, J. F. J . Controlled Release 1988,8,45. 5. Kuhnert-Brandetatter, M.; Winkler, H. Sci. Phurm. 1976, 44, suppi. 4,288. 6 . Kuhnert-Brandstatter, M.; Winkler, H. Sci. Phurm. 1976, 44, Suppl. 3, 191. 7. Kuhnert-Brandstatter,M. Phurm. Ind. 1977,39, Suppl. 4,377. 8. Kuhnert-Brandstatter, M.; Winkler, H. Sci. Pharm. 1976, 44, Suppl. 3, 177.
References and Notes 1. Nell, G. Wien. Klin. Wochenschr. 1989, 101, 398. 2. Powers, M. S.; Schenkel, L.; Darley, P. E.; Good, W. R.; Balestra, J. C.;Place,V. A. Muench.Med. Wochenschr. 1988,130,Suppl.1, S7.
1014 I Journal of Pharmaceutical Sciences Vol. 87, No. 10, October 1992
