Figure 62 Effects of tight binding on inhibition

compounds in Figure 6.2; in sequence they are: very approximately 10.3, approx. 9.3 and 8.1. For tight-binding corrected pIC50 (-log IC50) values subtract 0.3 from pKi.

The data in Figure 6.2 show that the imidazole (6.20), X=Y=CH, may be the most potent inhibitor yet described. Modelling studies indicated that additional lipophilic substituents/fusions at C4 and/or C5 could produce yet further substantial improvements—but these ideas were not pursued because ICI 207658 had been identified as a compound with great potential and, more to the point, the improvements we most sought in vivo required an approach with reduced susceptibility to oxidative metabolism at its core. We returned to that task much later.

The in vitro binding/inhibitory sequence: imidazol-1-yl>triazol-1-yl>triazol-4-yl, with 10- to 30-fold gaps was a consistent finding in our work; 5-methylimidazol-1-yl may sometimes be superior to its parent, but tight binding usually makes this uncertain. Each of these azoles attached to the naphthol-lactone framework had good potency in OI3 (shorter-term test) with ED50s of 250 to 500 pg/kg, but only the triazol-1-yl derivative retained its potency in OI2 (single dose given 20 hours earlier than the time of dosing in OI3). In this series and in most others, the triazol-1-yl compounds were equal or superior to the imidazoles in vivo despite the order of magnitude disadvantage in binding affinity. Advantages were most marked in the OI2 test. We believe relatively easy oxidation of the imidazole ring and/or the linking methylene group to be chiefly responsible for this disparity since the imidazole/triazole activity ratio is at its most extreme with the most robust frameworks. Robustness here is based partly on chemists' qualitative judgement and partly on observed plasma half-lives. In the naphthol-lactone case the imidazole had equal activity to the triazole but this still suggested easier than desired attack on the framework. Much later that idea was supported by a brief toxicology/pharmacokinetic once-per-day(u.i.d) oral dosing study on the N1-linked triazole, which returned a t1/2 of less than one hour in male rats.

The N4-linked triazole behaved worse than the imidazole in terms of its OI3/OI2 ratio (value for N4-linked triazole was >8), so it was expected to be rapidly cleared and to produce little toxicity in a similar study, which likewise involved u.i.d oral dosing. In fact it produced unacceptable liver toxicity. It seems likely that binding to haem-linked iron is disfavoured by haem-N(S-) to azole-N(S-) repulsion, see (6.21), whereas when this heterocycle is a ligand to other metallo-enzymes (not just iron-based enzymes) such repulsion could be less or even attractive if an alcohol, water or amide ligand is present—as shown for R-OH in (6.22).

One should further reflect that the average initial unbound state of the compound in vitro is essentially equal to that of an aqueous solution [though perhaps not for the highly lipophilic N-'benzyl' imidazole (6.7)], and the nitrogen lone pairs are solvated by hydrogen bonds to water—which is an excellent proton donor; so, since transfer to the enzyme-bound state involves loss of that solvation, the lack of a hydrogen bond or some equivalent in the bound state means that potency must suffer. This seems to apply in our case, but the SAR of the latest bis(4'-cyanophenyl)-methylazole Ciba-Geigy inhibitors (Lang et al, 1993), in particular the very high in vitro potency of the tetrazol-2-yl derivative CGS 45688, implies either that some H-bond replacement occurs or, perhaps more likely, there are marked conformational energetic/steric effects favouring the additional ring nitrogens in that particular compound

6.4.5 The design principles behind ICI 207658—later named Arimidex

The design of the naphthol-lactones, based on molecular modelling a wide selection of the large number of inhibitors then known, had in vitro potency as its dominant feature. But one has good reason to hope that selectivity will increase as potency increases. Testing for selectivity is time and resource consuming and can look at only a very small fraction of relevant P450s, let alone other enzymes. Effects due to receptor binding should also be considered—particularly if the drug is structurally similar to the substrate. The modelling approach by its nature has little, at least initially, to do with avoidance of metabolism, and 'adding on' this feature at some later point is seldom likely to be easy. Sustituting fluorine for hydrogen, especially in aromatic rings or as in CF3 or CF2 groups is widely practised and comes closest to a panacea—usually tolerable changes in size and lipophilicity (Edwards, 1994)—but synthetic difficulties and cost are often prohibitive and, like all supposed panaceas of the author's experience, it often seems least attractive when it is most needed.

Independent of modelling there are useful general principles that can be applied to seek extra selectivity. These may lead to the achievment of established selectivity requirements and perhaps also help avoid unexpected toxicity and non-pharmacologically-related side effects:

• without sacrificing much in vitro potency relative to a (notionally) iso-lipophilic comparator—flexibility, more specifically easily accessible conformational space, should be minimised; this is often a basis for improved potency.

• without sacrificing much in vitro potency relative to a (notionally) iso-lipophilic comparator—one or more low-flexibility groups may be introduced which have high steric and/or solvation demands.

• polar or more specifically hydrogen bonding atoms and groups should be preferred to lipophilic isosteres and they should be well dispersed throughout the molecule; here the potency criterion is more complex due to competitive phase effects: reducing lipophilicity has to be consistent with maintaining adequate in vivo starting state to target-bound state energy differences, steady state assumed. For a set of analogues with positive logPoctanol values this approximates, on a pIC50/logP plot, to being on the high-potency/low-logP side of a line of near unit slope (e.g. with slope near 0.7), drawn through data-points for compounds of substantial current interest.

The problem remains how to achieve at least some of these aims while maintaining or preferably improving resistance to oxidative metabolism and other clearance processes.

In this regard, one perhaps widely useful group was first recognised as a result of the attempted synthesis of tetralone (6.24), see Scheme 6.4. The tetralone and related naphthol-lactone targets had been conceived and worked on together, but a synthetic

intermediate to the tetralone, (6.23), was early on converted to the imidazole (6.25). In view of the widespread activity in compounds of this type, it was not surprising to find good potency in AR1: IC50=4 ng/ml, but the ED50 in OI3 of 0.5 mg/kg was somewhat surprising. It was more surprising when compared to the results that had been obtained several years earlier with N-4'-cyanobenzylimidazole—one of the compounds that had shown unacceptable liver effects with multiple high doses. That very simple compound had a better IC50 (<1 ng/ml), but a slightly worse ED50 of 1 mg/kg. We expected the cyanophenyl compound to be metabolically the more robust and so probably should have been the more potent in OI3 and OI2—in line with its higher enzyme inhibitory potency. In OI2 (single doses given 20 hr prior to that in OI3), both were classed as inactive (ED50>2.5 mg/kg), so their relative potency in this test was unknown; further

Sehemc ft,4

work at higher doses was not done because neither had the required 'duration' characteristics.

The unexpected potency reversal was rationalised by assuming that the 1-cyano-1-methyl-ethyl (CME) group was itself not easily attacked and in addition probably conferred some steric and electron-withdrawing protection to the benzenoid ring and its additional substituent. The modest difference in lipophilicity might then be invoked, through an increase in apparent volume of distribution and a lower rate of clearance of unchanged drug, to account for the potency reversal seen in OI3, where duration of action is moderately important. Regarding lipophilicity, the /-values for the CME group are +0.25 for octanol/water (measured) and -0.4 for hexane/water (estimated) compared to -0.40 and -0.99 (both averages of several measured values) respectively for aromatic cyano. The modest octanol/hexane difference for the CME group indicates that its presence in a molecule should not of itself seriously compromise penetration through lipid membranes; this bodes well for oral absorption and rapid distribution. Another factor assisting absorption, good solubility, could arise from the easy rotation of the strongly anisotropic CME group about the Ar-C bond in solution: this will help to keep melting points low since easy rotation in the crystal environment is highly unlikely and entropic factors therefore favour non-crystalline states, e.g. non-glassy melts and solutions. Melting points, together with logPoctanol values, are inversely correlated with log(aqueous solubilities).

That the geminal methyl groups, relative to most aliphatic groups, could have substantial resistance to oxidative attack follows from bond energetics—primary C-H bonds are the strongest—but also from a consideration of how the activation energy for hydrogen transfer is influenced by its surroundings. The forming H—OFe bond is highly polar and electron transfer to the extremely electrophilic O=FeIVporphyrin+ ' runs ahead of nuclear motion; this increases the fractional positive charge on the methylene group and the migrating H atom. The reaction-generated electric field and changes in fractional charges are opposed by the strong, electron-withdrawing field due to the cyano group (gf~0.57) and also by the weaker electron-withdrawing field associated with the aryl ring (cF~ 0.13); see (6.26).

The intervening quaternary carbon atom somewhat distances (insulates) the methyl groups from these field effects, but with an expected 'transmission coefficient' of ~0.4, one could still expect substantial protection. Similar lines of argument apply to the benzylic methylene hydrogens with the more electron withdrawing triazole, now with no 'insulating' atom, inhibiting oxidation better than imidazole: gf~0.49 and 0.35 respectively (azole cF data are from the Ph.D. thesis of D J Hall, University of Wales at Bangor, 1990). The size and hydrophilicity of the azole rings also hinder oxidative metabolism of the methylene groups, but attack is speeded by their weak resonance stabilisation of the transition state to the intermediate radicals. Fortunately the last-named effect is not dominant in these systems.

The effect of the 1-cyano-1-methyl-ethyl (CME) group on the ease of oxidation of the aryl rings is also expected to be substantial: thus for CH2-CN, gf~0.23 and cp+~0.16 (more deactivating than an aromatic chloro-substituent: cp+=0.11). The CME group is expected to be equally deactivating and even non-protonated imidazolylmethyl, and more so the triazolylmethyl group, will further reduce rates of electrophilic (oxidative) aryl substitution.

Steric hindrance around the cyano gives confidence that hydrolytic or other nucleophilic attack at this group could be minimal, and no easy metabolic release of cyanide is predicted—unlike benzyl cyanide where oxidation at the relatively unhindered benzylic C-H bonds produces a cyanohydrin, which allows easy release of cyanide and potential acute toxicity.

Guided by the above ideas on selectivity and previous SAR, we thus set out urgently to synthesise the 3,5-bis-CME analogue of (6.25) and to make triazole equivalents. The triazol-1-yl analogue of (6.25) was disappointing but ICI 207658, (6.27), (see Table 6.1), was very potent in vitro and, more importantly at that time, in vivo it was equipotent with CGS 16949A in the demanding OI2 test, both having ED50~15 pg/kg. Clearly this was very exciting!

In preliminary studies in male rats, at extremely high multiples of the effective dose in females, ICI 207658 showed no untoward effect; in particular liver triglycerides remained normal. A small increase in liver weight was consistent with the observed induction of mixed function oxidases. Adrenal and other organ weights were the same as controls. The reason for using males for this and the many other compounds investigated is that oestrogens indirectly regulate both adrenal weight and circulating triglyceride concentrations. We hoped that changes in the background levels in males would cause only minor changes in these parameters of central interest.

Problems arising from the use of different sexes are usually minor in most species except for rat: males frequently clear compounds faster than females and maximum plasma concentrations, Cmax, are often lower. The effects are multiplicative on AUC (area under curve of plasma-concentration vs. time). Thus simple ratios of effective dose in one sex to side-effect/toxic dose in the other can sometimes mislead selectivity assessments. The same cautions and others are more widely known to apply to comparisons across species.

This discussion of selectivity ratios benefits greatly from hindsight and is relevant here mainly to the Team's first development compound described in Section 6.5. The retrospectroscope also indicates that our clamour for pharmacokinetic and preliminary toxicological studies, which exceeded the capabilities of the appropriate Safety of Medicines Department workgroup to respond, contributed to some insecure conclusions. Most such studies took place later than initial in vivo selectivity studies and were not chiefly driven by the need to better assess selectivity ratios.

In the case of ICI 207658, Cmax was lower in males than females by two- to threefold, but half-lives of ~6 hr, dropping to ~4 hr at the end of the multiple-dose study were reported to be essentially the same in both sexes. The results of this preliminary study supported and expanded the basis for the Team's conclusion that ICI 207658 was a very promising compound.

Half-lives in rat usually are much shorter than in man so the above values, while being short of our target range, were not a cause for concern and it was predicted by our Safety of Medicines experts that induction of mixed function oxidases in liver was most improbable at the very low predicted human therapeutic dose. It was subsequent data from studies in dog and pigtailed macaque monkey—producing half-lives of ~8 hr and 7 to 10 hr respectively—that seemed to indicate a remarkably uniform half-life across species and led to a majority view that the compound could not be relied on to achieve our target minimum half-life in patients.

Well before any of the pharmacokinetic data were available we had discovered that ICI 207658 occupies a pinnacle of in vivo SAR space; not a single analogue came within an order of magnitude in OI2 potency terms! This is not the place to go into detail so data on just a few compounds are shown in Table 6.1. Well over a hundred analogues were made with small and sometimes larger variations at every locus where change is possible. If you can think up a related structure, we probably made it or tried to make it (one rather obvious analogue is an exception—that involving changing cyano to nitro—we never did attempt to make it, despite it being one of our listed targets for a long time). Even

Table 6.1 Potency of selected azoles.


1-(Ra-methyl)-3-CME-5-Rm -









Ra Rm











triazol-1-yl CMEa (ICI 207658) 4




imidazol-1-yl CME




triazol-3-yl CME




triazol-1-yl CH2-S-CH3




triazol-1-yl C(CH^-OH




triazol-1-yl C(=O)-CH




triazol-1-yl C(CH3)2-COCH3




aCME represents a 1-cyano-1-methyl-ethyl group bestimated values (adjusted for tight binding)

aCME represents a 1-cyano-1-methyl-ethyl group bestimated values (adjusted for tight binding)

such small changes as homologating one of the four methyl groups to an ethyl group, or converting two geminal methyls to cyclo-alkyl (3- or 4-membered), or introducing an ortho- or para-bromo substituent, or changing the positions around the benzene ring, etc. Many active compounds were identified, but none was as supremely effective as ICI 207658.

Analogues of CGS 16949A (6.18) containing one or two m-CME groups had, consistent with modelling work, significantly inferior ARl-potency.

As can be seen even from the very small data-set in Table 6.1, poor in vivo potency was rarely attributable to inadequate enzyme affinity. The early hypothesis concerning the properties of the CME group, now groups, with regard to in vivo handling of the drug stood the test of time—albeit one or two compounds with good and even excellent AR1 figures, but poor OI2 results, remain difficult to explain. Single test results may be wrong, we seldom had good reason to retest, or perhaps the anomalies relate to rat vs. human aromatase selectivity, or perhaps our analysis is flawed.


The antifungal lead had been converted to potential development candidate almost overnight, but that potential had to be assessed. At the time of the discovery of the bistriazole (6.17) no detail of the Ciba-Geigy compound was known, so AG was our yardstick and, because we were building up an ever-increasing body of data, it remained so for the first half of the resumed programme. Against that yardstick the limited in vitro selectivity data was pleasing, particularly with regard to P450scc/P450arom, where a 60-fold improvement over AG was seen. Most of the early efficacy data was very promising—and not just in rats: in monkeys dosed at 0.1

mg/kg for 10 days a near maximum achievable reduction in oestrogens was achieved. Similarly potent effects were not seen in dog—perhaps because levels of testosterone, a precursor of oestrone and oestradiol, increased 5- to 10-fold in a dose-related fashion with drug treatment. Literature reports ascribed the testosterone increases to hypothalamic aromatase inhibition so we tried to use this as a test system. Unfortunately, near maximal testosterone levels were seen only after multiple doses of 1 mg kg-1 day-1. In view of the very high blood levels achieved in dogs this relatively massive required dose seemed unlikely to correspond simply to aromatase inhibition; we therefore placed little reliance on intra-species dog selectivity ratio assessments. It is also possible that dog aromatase differs substantially from the rat and human enzymes.

In preliminary seven-day toxicity studies in rat, there were the expected increases in mixed function oxidases and increased liver weights at high doses, but, in the absence of effects on triglycerides, these were acceptable findings. The only slight concern expressed in the proposal for development was an increase in adrenal weights: in male rats at 50 mg/kg, 250-times the OI2 and placental enlargement ED50-doses; in dogs at 10 mg/kg.

The problem of rat sex differences in drug handling was substantial since both Cmax and half-life in males were a third of those in females. AUC is therefore an order of magnitude lower and if tissue levels daily fall below some critical threshold for long enough, it is possible for body systems to largely recover from 'toxic' effects. Small but significant reductions of the male rat accessory sex organ weights and testosterone and LH plasma levels at all doses down to and including the lowest tested, 0.1 mg/kg, were regarded as toxicologically inconsequential. Since other aromatase inhibitors tested had no such effects, these findings demonstrate a lack of selectivity, but in what way remains uncertain. It may be that, akin to AG, changes in P450-mediated rates/routes of hormone catabolism are occurring. Akin to this, interference with barbiturate metabolism in young male rats was evident at 1 mg/kg, but not at 0.1 mg/kg.

The Team's development proposal was accepted by higher management and, only 15 months after restarting the programme, the Team had a compound in full development! With luck and rapid development we might still achieve commercial success—but the Ciba-Geigy compound, seemingly superior in potency and selectivity, was now clearly well ahead in the race. And there were still so many hurdles for the bis-triazole (6.17) to clear.

As the toxicity studies with (6.17) proceeded, the tally of adverse findings increased and our understanding of the unusual steroidogenesis in rat adrenal increased, leaving us with concerns about our ability to detect adverse changes relevant to other species, particularly man. In dogs, hypokalaemia was seen at modest doses and adrenal cortex vacuolation was slightly elevated from controls even at 0.5 mg/kg. It seemed certain that inhibition of 11-hydroxylation was to blame, but there was also no doubt that matters were made worse by the progressively higher Cmax and AUC values which follow daily dosing of any long half-life compound. In our dogs the half-life was 2-4 days so substantial accumulation would have occurred. This is the reverse situation to that described above for male rats and emphasises the importance of temporal drug level profiles to safety/selectivity assessments. Such profiles are also very important to some chronic efficacy studies: (6.17) has a half-life in pigtailed monkeys of one day so, barring enzyme induction—which is unlikely at the low doses used—there will be no gaps in drug cover and with chronic dosing a two-fold elevation of Cmax and AUC should occur. As stated previously, it is almost maximally effective with once-daily doses of 0.1 mg/kg. Contrast this with CGS 16949A: we found it to have a half-life of ~5 hr in female rats but less than 2 hr in monkeys; its large advantage over (6.17) in OI2 and still greater advantage (40-fold) in OI3 is reversed in monkeys—they require 0.1 mg/kg every 12 hr to achieve near maximal reduction of oestrogen levels.

This competitor compound mirrored our own in steadily revealing its weaknesses throughout the time of the bis-triazole (6.17) development. Our work in rats and dogs revealed increasing selectivity issues and the absence of, from our viewpoint, relevant selectivity data from both oral presentations and publications dealing with CGS 16949A (fadrozole hydrochloride), including human studies, encouraged us in due course to review our priorities.

As a business we had had many adverse experiences with long half-life compounds in chronic (6 month and more) toxicity studies. This was not to be an exception. In dogs, serum cholesterol and triglyceride levels were reduced by modest doses of the bis-triazole (6.17) and, by six months, cataracts were seen in the eyes of most dogs dosed at 7.5 mg/kg. The new fibre cells, which in mammals continuously enlarge the eye lens during life, need to synthesise their own cholesterol because they incorporate it in large amounts into their membranes and, being an avascular tissue, they cannot obtain it from the low density lipoprotein (LDL) in plasma. A prolonged shortage of cholesterol in this tissue seems to lead to cataracts. We suspect that at the observed high plasma levels in dogs, (6.17) inhibits one or more of the P450-dependent enzymes that transform lanosterol to cholesterol.

The lead was born from a poor fungal lanosterol-14-methyl demethylase inhibitor and died, 18 months into development, due, probably, to inhibition of a canine lanosterol-demethylase.

In passing it is worth noting the large number of conformers easily accessible to (6.17) and its multiple chelation possibilities—bidentate and even tridentate. Perhaps these facts contribute to its inadequate selectivity.


Inhibition of mammalian cholesterol synthesis had no precedent in aromatase inhibitors prior to the findings with (6.17), but we now urgently needed to look at possible successors to (6.17) in this new light. What had we found in that category during those 18 months? Not a lot. We found as expected that we could improve on the original naphthol-lactones (6.20)—but not sufficiently to fall into the presently required category. Potency in OI2 and PE9 had been somewhat improved, those improvements being associated with electron withdrawing substituents at C7. The better substituents, e.g. cyano, should hinder oxidative metabolism of the aromatic ring system and the proximal benzylic methylene. During this synthetic work we were surprised, following nitration or bromination, to observe amongst several products some substitution at the very sterically hindered C9 position; mostly reaction was at C7. As in the ICI 207658-like series, we made a large number of analogues but failed to make significant headway. We concluded, using an estimated cF value for the phenolic-lactone group, that extra protection of the gemdimethyls and the aromatic system was desirable. But as we saw no practical way to provide it we ceased work on this series. On reflection there were more changes we might have tried.

The only blemish on the profile of ICI 207658 was its projected 'inadequate' halflife in man; in all in vitro selectivity tests conducted, including now against cholesterol synthesis, it performed superbly. None of a great many analogues was attractive. What else might be done?

There were clues to hand: androstenedione had been synthesised with deuterium or tritium in specific locations as part of the aromatase mechanistic studies. Kinetic isotope effects were seen. Of most relevance to us, the 19-trideutero compound in admixture with the non-deuterated parent showed an intermolecular isotope effect kH3/kD3 approaching 3-fold, and the first hydrogen removal is known to be rate limiting. If this ratio, or even half the ratio, applied to the half-life of a deuterated ICI 207658—in at least two of the species used previously—might we not be home and dry? A quick back-of-the-envelope calculation showed that the additional cost of even a tetradeca-deutero compound could be trivial—probably only about one penny/mg— and, with an increased half-life, there was reason to believe the daily dose might be only ~3 mg per patient.

We made the three deuterated compounds (6.34), (6.35) and (6.36), henceforth referred to as D2, D12 and D14. The hydrogens attached to the triazole ring were not changed because the SAR and metabolite studies pointed firmly against metabolism in this ring being relevant. Several antifungal triazoles had half-lives in male rats of about a week.

Similarly the hydrogens directly attached to the benzene ring were not replaced because these are rarely subject to primary isotope effects: rate-determining attack takes place initially at carbon, on the n-system, and secondary isotope effects are typically too small for the present purpose.

The D2 and D12 compounds were made despite concerns that wherever attack normally might occur, the partially caged substrate would still react at the remaining weakest point, so called 'metabolic switching', and so reduce any advantage. The likelyhood of metabolic switching in our target deuterated analogues also carried with it the risk of generating new, longer-lived, or more abundant metabolites with reduced selectivity or increased toxicity. Of course reduced toxicity is also possible: the subject has been reviewed by Pohl and Gillette (1984-5).

Intramolecular isotope effects in P450-mediated oxidative reactions, as in non-enzymic haem-based model systems, can be very large: values around 20 are known and 5-10 are normal. In contrast, the isotope effect expressed in kcat for metabolism of phenylethane or a,a-dideuterophenylethane with a rabbit liver-derived P450LM2 enzyme is only 1.28. This indicates that at least one enzymic step with a large 'commitment to catalysis' precedes hydrogen abstraction in this case (White et al, 1986). It can be argued that the effect is small in this case because of the activated nature of the secondary, benzylic C-H bonds. It is however by no means exceptional and ICI 207658 contains a related if much deactivated, more hindered and more hydrophilic part-structure. In contrast to this low value a related study on trideuteromethoxy anisole showed an in vivo isotope effect of 10.

Studies in vivo generally show less marked substrate dependence with smaller but still some substantial isotope effects. Increases in half-life of 1.5- to 2.5-fold are typical (Blake, Crespi and Katz, 1975). In such clearance processes one is dealing with multi-step events and the oxidative step is normally only partially rate determining. D B Northrop has developed a general equation, Equation [6.1], for the interpretation of isotope effects in multi-step reactions. The maximum rate, Dv, is controlled by the ratio of catalysis, R, which represents the ratio of the rate of the isotope-influenced catalytic step to the rate of the other forward steps contributing to the maximum rate.

Octanol/water partition and in vitro aromatase inhibition studies showed the expected equivalence of all isotopic species. The effective size of the more slowly vibrating CD fragment is on average very slightly smaller than a corresponding C-H fragment, but the difference in non-covalent binding properties is well below the detection limit in most biological systems.

In vivo potency and limited pharmacokinetic studies with the three compounds, mainly as single agents but sometimes as solid-solution mixtures (to avoid possible differential solubilisation and absorption of individual samples) produced somewhat confusing results. OI2 tests (necessarily using females) in head to head comparisons with ICI 207658 (DO) showed a 3-fold potency improvement for D2 and improvements of 3.5- and 2-fold for D14 on separate occasions. The result for D12 was identical to that of DO. This indicated the benzylic methylene as the main site of oxidation in female rats at very low (2, 5, 10 and 20 pg/kg), near-therapeutic doses.

A very limited pharmacokinetic study compared the compounds at 1 mg/kg with historical data. Plasma levels of D12 in female rats were followed to 70 hours post dosing and showed no detectable isotope effect. A similar result applied to D2 from samples taken at 1, 2 and 8 hours post dosing in males—the timepoints at which data were available from the historical study of DO in males. Only D14 showed some effect: in males followed to 24 hr post dosing: the Cmax and AUC increased by 60% (up to 8 hr), but with no detectable change in half-life.

In part because of the limitations of the severely resource-constrained pharmacokinetic study, we tried in a very few animals to use mass spectrometric analysis to follow intra-individual handling of mixtures of compounds. In male rats dosed with a solid-solution of DO with its D2 and D14 analogues, and using the historical data on DO for comparisons, the apparent isotope effects interpreted as half-lives were 1.0 to 1.2 for the D2 compound (poor data due to plasma-related peaks) and 2.0 for D14. The same isotope effect, 2.0, was seen for a binary solid-solution of DO and D14 compounds. A similar experiment in one dog yielded an apparent isotope-induced increase in half-life of 2.1-fold up to 12 hr post dosing, but decreasing beyond this time to an average of 1.7-fold over the full 24 hours of the experiment.

Being encouraged by these sighting experiments, but realising their extreme limitations and the possible future need for more extensive work, we carried out a detailed analysis of the likely kinetic scheme for the overall process. Surprisingly, this revealed that results from these, at-first-sight, ideal experiments, involving intra-individual temporal changes in concentration ratios of compounds, cannot be unambiguously interpreted in terms of individual clearance rates or related half-lives. We were therefore left with insufficient solid evidence of benefit from deuteration and the undoubted penalties of increased compound costs, analysis costs and uncertainties with regard to Registration Authority views and delays. The approach was abandoned.

The search for a better candidate continued mainly with cis-tetrahydronaphthalenes like (6.12), stilbenes related to stilboestrol (6.3) and corresponding reduced analogues with a 1,2-diarylethane framework. In many cases, compounds with a pyridine ring replacing one of the benzene rings, e.g. (6.37) (racemate), had excellent activity both in AR1 and in OI2. None of these compounds was satisfactory in all respects. In some the half-life was too long—longer than or similar to the bis-triazole (6.17) in the dog was now a near automatic bar to progression—while others had inadequate selectivity: like bis-triazole (6.17), the pyridine (6.37) substantially lowered serum cholesterol levels—by then a totally unacceptable encumbrance. Whether the effects on aromatase and cholesterol synthesis could be separated through resolution was not investigated. Time


had almost run out. We had had to progress many compounds, first through larger scale synthesis, then often into semi-chronic and chronic tests before finding them unsatisfactory.

Janssen also were now forging ahead with the very impressive but racemic triazole R76713, (6.38). Published information showed potent activity in volunteers, so they too were now far ahead of us and our limited selectivity data on the compound gave us no comfort whatever. We had to make a choice—now. That choice was by now almost inevitable: ICI 207658 was associated with only temporary accumulation following multiple large doses in dogs and it had an excellent selectivity profile. Selectivity had again come close to the top of the Team's priorities. Once again life comes pretty much full circle and ICI 207658 was entered into development under the number D1033.

Long term toxicity studies revealed no significant additional findings to those seen with shorter exposures, but comprehensive pharmacokinetic studies revealed that the preliminary half-life estimates in rat had been in error. Due to a plasma-associated material interfering with the assays, those estimates were generally too long; for example, the initial t1/2 in males is now known to be ~2 hr and a half-life of 2.3 hr was observed after long-term dosing. It is therefore almost certain that the isotope work did achieve its objectives. But had we pursued this line into development we may have been faced with a supra-optimal half-life in patients—see below.


Escalating dose studies in male volunteers confirmed our expectation of good absorption, rapid distribution and high bioavailability, so the compound progressed into the first patients. The benefits we had confidently expected to find materialised and with a half-life of two days it fitted our target for optimum use long-term in post-menopausal women. No serious side effects, enzyme induction or inhibition—barring aromatase—have been observed and no indications of a lack of selectivity have been seen at either the 10 mg or 1 mg u.i.d. doses investigated (Plourde, Dyroff and Dukes 1994). Since the lower dose gives >95% inhibition of aromatisation in biochemical studies in patients, and equivalent anticancer efficacy to the higher dose, this smaller quantity, corresponding to approximately 15-20 pg/kg, was chosen as the recommended dose for use of Arimidex in postmenopausal breast cancer.

Arimidex was launched in the U.K. on 19th September 1994 and so became the first of the fourth-generation, potent, highly-selective aromatase inhibitors to achieve commercialisation.

Mature results from large randomised clinical trials show that Arimidex treated patients have a significant survival benefit over patients treated with another endocrine agent. It is the first aromatase inhibitor to show such an advantage.


The author extends his thanks to Mr Mike Large and Mr Chris Green for their contributions to much of the chemistry described, to his biological colleagues headed by Dr Mike Dukes for their superb work and often heroic efforts, and to the host of others who make essential contributions to the appraisal of any potential new addition to the drug armamentarium. I also thank Dr Dukes for helpful comments and criticisms of the manuscript. Thanks and apologies are extended to the many other chemists who made contributions to the Team's endeavours but whose work has here been so scantily reported.


Akhtar, M., Njar, V.C.O. and Wright, J.N. (1993) Mechanistic Studies on Aromatase and Related C-C Bond Cleaving P-450 Enzymes. Journal of Steroid Biochemistry and Molecular Biology 44, 375-387.

Blake, M.I., Crespi, H.L. and Katz, J.J. (1975) Studies with Deuterated Drugs. Journal of Pharmaceutical Sciences 64, 367-391.

Brodie, A., Brodie, H.B., Callard, G., Robinson, C, Roselli, C. and Santen, R. (eds.) (1993) Recent Advances in Steroid Biochemistry and Molecular Biology: Proceedings of the Third International Aromatase Conference; Basic and Clinical Aspects of Aromatase. Journal of Steroid Biochemistry and Molecular Biology, Volume 44(4-6).

Castagnetta, L., D'Aquino, S., Labrie, F. and Bradlow, H.L. (eds.) (1990) Steroid Formation, Degradation and Action in Peripheral Tissues. Annals of the New York Academy of Sciences, Volume 595.

Djerassi, Carl (1992) The pill, pygmy chimps, and Degas' horse: the autobiography of Carl Djerassi. New York: BasicBooks.

Edwards, P.N. (1994) Uses of Fluorine in Chemotherapy. In Organofluorine Chemistry: Principles and Commercial Applications, edited by R.E.Banks, B.E.Smart and J.C.Tatlow, pp. 501-541. New York: Plenum Press.

Henderson, D., Philibert, D., Roy, A.K. and Teutsch, G. (eds.) (1995) Steroid Receptors and Antihormones. Annals of the New York Academy of Sciences, Volume 761.

Lang, M., Batzl, Ch., Furet, P., Bowman, R., Hausler, A. and Bhatnagar, A.S. (1993) Structure-activity Relationships and Binding Model of Novel Aromatase Inhibitors . Journal of Steroid Biochemistry and Molecular Biology 44, 421-428.

Plourde, P.V., Dyroff, M. and Dukes, M. (1994) Arimidex®: A potent and selective fourth generation aromatase inhibitor. Breast Cancer Research and Treatment, Special Issue: Aromatase and its Inhibitors in Breast Cancer Treatment, edited by A.M.H.Brodie and R.J.Santen, 30(1), 103-111.

Pohl, L.R. and Gillette, J.R. (1984-85) Determination of Toxic Pathways of Metabolism by Deuterium Substitution. Drug Metabolism Reviews 15(7), 13351351.

Schenkman, J.B. and Greim, H. (eds.) (1993) Cytochrome P450. Handbook of Experimental Pharmacology, Volume 105.

Wakeling, A.E. (1990) Novel Pure Antioestrogens, Mode of Action and Therapeutic Prospects. Annals of New York Academy of Sciences 595, 348-356.

White, R.E., Miller, J.P., Favreau, L.V. and Bhattacharyya, A. (1986) Stereochemical Dynamics of Aliphatic Hydroxylation by Cytochrome P450LM2. Journal of the American Chemical Society 108, 6024-6031.

0 0

Post a comment