Disappearing Polymorphs Revisited

Nearly twenty years ago, Dunitz and Bernstein described a selection of intriguing cases of polymorphs that disappear. The inability to obtain a crystal form that has previously been prepared is indeed a frustrating and potentially serious problem for solid-state scientists. This Review discusses recent occurrences and examples of disappearing polymorphs (as well as the emergence of elusive crystal forms) to demonstrate the enduring relevance of this troublesome, but always captivating, phenomenon in solid-state research. A number of these instances have been central issues in patent litigations. This Review, therefore, also highlights the complex relationship between crystal chemistry and the law.


Introduction
There is acontinual and increasing demand for crystalline molecular materials with specific, fit-for-purpose physicochemical properties. [1][2][3][4][5][6] Interest in polymorphism,c rystallization, and (in industry) in robust process development has surged over the last two decades, [7,8] as evidenced by the immense growth in knowledge concerning the design, preparation, and characterization of crystalline materials. [9] This expanding interest and demand for promising materials drives investigations of the solid form (i.e.p olymorphs,s olvates, hydrates,and amorphous materials) landscapes [8,10] of potentially relevant compounds,w ith the goal of identifying the optimally performing solid among them.
Ab road range of crystallization techniques is generally employed to search for the most stable crystal form in hundreds or (in some cases) thousands of experimental attempts. [11] New crystal forms can, however,e merge unexpectedly long after the carefully designed and executed screening experiments are completed. Such as udden emergence of an ew crystal form can be unsettling and problematic,especially in the late stages of aproduct development or even following launch, because the newly emerged form can exhibit different (possibly undesired) properties.E qually disruptive is the emergence of at hermodynamically morestable crystal form, in accord with Ostwalds Rule of Stages, [12] concurrent with the disappearance of the less-stable known forms that signal al oss of control of the production process. While it may create roadblocks in the development process or even the marketed product of the solid form of acompound of interest, the consequences of the appearance of an ew form are not necessarily negative.The serendipitous appearance of an ew form may provide as ubstance with improved characteristics.
Unfortunately,o ur current understanding of the mechanisms and processes involved in the nucleation and growth of crystals is still insufficient for precise control over the formation or disappearance of ap olymorph (or any other crystal form). [13,14] Nearly twenty years ago,D unitz and Bernstein presented an overview of the disappearing polymorph phenomenon [15] that has captivated and intrigued solid-state scientists since.I nt heir review,D unitz and Bernstein voiced their belief that crystal forms do not disappear permanently;o nt he contrary,o nce as olid form has been obtained, in principle it can always be reproduced if the right experimental conditions are met. [15][16][17][18] In the same spirit as the earlier survey,t his Review aims to discuss selected recent occurrences of disappearing polymorphs and of elusive crystal forms that have not only triggered the curiosity of researchers,but have also affected the business of pharmaceutical and health care companies.T hese examples illustrate how apparently stable polymorphs can suddenly disappear,a nd how elusive crystal forms can be prepared given the availability of conditions specifically designed to promote their formation. Theu ncontrolled loss of ac rystal form can have serious consequences,a nd there is thus an urgent need to develop methods that provide absolute control over crystal nucleation and growth, [13,14] which is still an art, rather than ar outine procedure. [19] In addition to citing examples of disappearing polymorphs from the literature and our own laboratories,the 1995 review dealt with an umber of issues that are still the subjects of debate.There have also been anumber of patent litigations in which the same issues have arisen and have been interpreted variously by the courts.W ew ill deal initially with those aspects of the subject and follow with the descriptions of an umber of recent cases of disappearing polymorphs (and other crystal forms), as well as further details on some of those previously cited.

Disappearing Polymorphs-The Concept and Misconceptions
One of us (J.B.) recently recounted the genesis of the 1995 review, [20] which was based on earlier cases in the laboratories of both Bernstein and Dunitz as well as additional examples we had encountered in the course of our involvement in the ranitidine hydrochloride litigations.I nt he twenty-year interim we have experienced numerous additional examples in which the phenomena described therein were either misinterpreted or misunderstood. Hence,w er eview some of those here.

The Concept
As we described in the section of the 1995 review headed "Vanishing Polymorphs", adisappearing polymorph refers to acrystal form that has been prepared at least once and whose existence has been established experimentally by some observation or measurement. Subsequent attempts to prepare the same crystal form by the same procedure lead to ad ifferent crystal form, alone or together with the old one.
If am ixture appears in the first instance,t hen very often in subsequent preparations the new form dominates and the old form is no longer obtained.
Thep hase rule limits to one the number of stable crystal forms that may exist under as pecific set of conditions.T he old-"disappeared"-formi sg enerally less stable than the new one under those specific conditions.I nt hermodynamic terms,i ti sm etastable,a lthough that does not necessarily imply that it would spontaneously convert into amore stable form;itonly means that it is at ahigher energy minimum than the most stable state.T oinvoke afamiliar example:diamond is metastable with respect to graphite;n evertheless,a si s widely advertised, "diamonds are forever".
Thef act that ac rystal form once existed, but is now difficult to prepare by the same method that was previously used, does not mean that it is impossible to prepare again. It has not been relegated to the "crystal form cemetery". [21] Every crystallization is ac ompetition between kinetic and thermodynamic factors.A sn oted in the last sentence of the 1995 review," it is always possible to obtain [the old form] again;i ti so nly am atter of finding the right experimental conditions"-thermodynamic and kinetic.
Recovering ac rystal form that has disappeared may require considerable time and effort and invoke some inventive and creative chemistry.T he examples given below will demonstrate the kinds of strategies that have been employed to recover crystal forms that have disappeared.

Seeds and Seeding
The1995 review also contains asection headed "Seeding". Intentional seeding is aw ell-known technique for inducing crystallization and is widely used, especially in the pharmaceutical industry.U nintentional seeding arises from the presence of small amounts-indeed, in principle one particle is sufficient-of the solid material that is present even as acontaminant. As we noted earlier, "Unintentional seeding is often invoked as an explanation of phenomena which are otherwise difficult to interpret. We shall argue in favor of this explanation, although there is no consensus about the sizeand range of activity of such seeds,which have never actually been directly observed." [15] Thes ituation this statement describes has led to considerable controversy,p articularly in the framework of patent litigations involving crystal forms.T hat controversy very much represents the clash between the cultures of science and the law,and in light of that controversy it seems appropriate, indeed compelling,t op ut the phenomenon of unintentional seeding into ap roper scientific perspective in this Review.
Virtually every chemist has at some time attempted to crystallize ac ompound. Crystallization is perhaps the classic method of purification, and the technique is one of the first mentioned in purification methods in any undergraduate organic chemistry laboratory textbook. Practicing chemists soon learn, often simply by experience,t hat it is frequently very difficult to crystallize an ewly synthesized substance, KrešoBučar  while subsequent crystallizations are considerably more facile.T he situation was documented over half ac entury ago by Wiberg in his classic text "Laboratory Te chnique in Organic Chemistry" in the section entitled "Inducing Crystallization": "When ac ompound is prepared for the first time in alaboratory,itisoften observed that it is relatively difficult to effect crystallization. However,once the compound has been obtained in the crystalline state,i ti su sually easy to effect crystallization, and it has been suggested that after initial crystallization crystal nuclei are present in the laboratory and induce crystallization". [22] In the current context those nuclei are unintentional seeds.
Many laymen are initially skeptical about aphenomenon caused by particles that cannot be seen, although very few would accept an invitation for ac asual-and unprotectedvisit to the pneumonia ward at their local hospital. The approximate limit of visual detection for the naked eye is ac rystal that weighs approximately 10 À6 g. We pointed out earlier that as peck of that size contains approximately 10 16 molecules and while there are various estimates of the size of ac ritical nucleus that could act as as eed even the largestaf ew million molecules [23] -would mean that an invisible particle could contain up to 10 10 of such unintentional seeds.
Where do these microscopic particles come from?A s noted elsewhere,d epending on our location, the air contains avast number of submicroscopic particles.For anormal urban environment there are approximately 10 6 airborne particles of 0.5 micrometer diameter or larger per cubic foot, the number being reduced by an order of magnitude in an uninhabited rural environment. Asitting individual generates roughly one million dust particles (! 0.3 micrometer diameter) per minute (a visible particle is usually ! 10 micrometers). [24] Clean rooms for various purposes (e.g.s urgery,b iological or pharmaceutical preparations,s emiconductor fabrication) employ very sophisticated technology to remove these particles and to prevent subsequent contamination. Therefore,t he possible presence of seeds of an ewly formed polymorph in al aboratory,amanufacturing facility,o ra ny location having been exposed to that form cannot be casually dismissed;indeed its presence would be hard to avoid. In his comprehensive monograph on crystallization, Mullin notes that, "Atmospheric dust frequently contains particles of the crystalline product itself,e specially in industrial plants or in laboratories where quantities of the material have been handled…Once ac ertain crystalline form has been prepared in al aboratory or plant, the working atmosphere inevitably becomes contaminated with seeds of the particular material." [23] So much for the atmosphere.What about the crystallizing medium, usually as olution?T he normal determination that dissolution has been completed is made by visual inspection. If the solution is clear to the human eye all the solute is assumed to be in solution. Mullin has also pointed out that "aqueous solutions as normally prepared in the laboratory may contain > 10 6 solid particles per cm 3 …". [23] These can be impurities or particles of the solute that have not undergone complete dissolution, and can serve as seeds for the subsequent crystallization.
Thepresence and influence of microscopic seeds and their influence on crystallization is thus well established. Nevertheless,i ti sd ifficult for many who lack practical laboratory experience to accept their existence.I nt he history of chemistry there have been many instances of inductive reasoning in understanding chemical phenomena. Thee xistence of atoms was proposed and accepted for nearly two hundred years before an atom was actually "seen". Yetn o chemist doubts the existence of atoms or the ability to make and break bonds between them.
Thep resence and influence of seeds may be invoked to explain the disappearance of one crystal form at the expense of anew form. In such acase,the unintentional seeding by the new form may be quite aggressive,p reventing the crystallization of the old form. However,there is no intrinsic reason why every system is influenced by such aggressive unintentional seeding.T here are many known examples of multicrystalline materials in which the various forms can be prepared and maintained in the known presence of other forms.A sf or polymorphism in general, every system is unique and must be individually studied and characterized to understand how to prepare and characterize each form.

"Universal Seeding"
Thep ublicity surrounding some cases of aggressive unintentional seeding led to discussions,p articularly in legal circles,ofthe alleged phenomenon of universal seeding-that is,i ns ome cases of disappearing polymorphs,w hen the old form could not be made by the old process somehow,t here was an implication that the entire universe must be seeded. To put the matter to rest it is important to quote afootnote from the 1995 review: "The claim for universal seeding,t aken literally,isobviously absurd. After all, the universe is estimated to contain about amillimole of stars,soone seed per star (per solar system)-not much-would need about 100 kg of the compound in question (M r % 100)".
An umber of cases of aggressive seeding have attained considerable notoriety,a nd these will be described below.I n instances where various locations at considerable distance have become "infected" with an ew form within ar elatively short time,i th as been possible to trace the source of the seeding in successively affected locations.

Recent Instances of Disappearing Polymorphs and Elusive Crystal Forms
This section describes several of the most (in)famous recent cases of disappearing polymorphs and other crystal forms.I na ddition, in relation to the sudden and unexpected disappearance of aw ell-known crystal form, we consider it particularly relevant to describe cases where elusive crystal forms,believed to be non-existent, were prepared.

Ranitidine Hydrochloride
In the early 1970s,J ames Black at (then) Smith, Kline & French identified the histamine type 2(H 2 )receptor and from the preparation of as eries of H 2 -receptor antagonists developed the first antiulcer drug, cimetidine (Tagamet ), for which he won the 1988 Nobel Prize in Medicine.H 2receptor antagonists are among the miracle drugs of the 20th century.Prior to their introduction (and the subsequent entry of proton pumps) there were millions of sufferers of peptic ulcers worldwide with asignificant number of fatalities;since their introduction, the surgical procedure for removing peptic ulcers has essentially been eliminated from the modern medical school curriculum.
Thed ramatic success of cimetidine led to industry-wide efforts to develop additional H 2 -receptor antagonists.In1977, Allen &Hanbury (then apart of Glaxo Group Research, now GSK) developed ranitidine and its hydrochloride (Figure 1a), for which aUSpatent was issued in 1978. [25] Thepreparation of the hydrochloride following the multistep synthesis of ranitidine base is given in "Example 32" of the patent (Figure 1b).
Subsequent development of the drug over nearly four years involved batch scale-ups to amulti-kilogram scale in the companys pilot plant by employing essentially the chemistry described in Example 32. [10] Theb atch prepared on April 15, 1980 failed the quality control IR analysis,w hich exhibited ahitherto unobserved sharp peak at 1045 cm À1 ,and suggested the formation of an ew crystal form designated Form 2.T he subsequent four batches exhibited increasing amounts of Form 2 and the same process no longer produced the (now designated) Form 1.C onsiderable efforts to revert to the production of Form 1 by essentially the same process were unsuccessful. Thus,t his is clearly ac ase of ad isappearing polymorph. Serendipitously, Form 2 had considerably improved filtering and drying characteristics which, in addition to the novelty of the new polymorph, formed the basis for ap atent application, granted in 1985. [10] Thec rystal structures of both forms have been subsequently determined; both forms crystallize in the monoclinic P2 1 /n,s pace group, wherein the nitroethenediamine moiety of the ranitidine cations displays different conformations and degrees of disorder ( Figure 1c). [26][27][28] This is thus also an example of conformational polymorphism. [29] Glaxo launched ranitidine hydrochloride in 1984 as Zantac and by 1992 it was the worlds best-selling drug at US$ 3.44 billion per year,when sales for the next largest drug (Bayers Adalat Procardia) were about half that amount. [30] In accord with the terms of the 1984 Hatch-Waxman Act in the US,b y1 990 an umber of generic drug companies were planning to enter the market with Form 1 in anticipation of the 1995 expiration of the Form 1 patent. Attempts to make Form 1 were based on carrying out Example 32. As transpired in the course of the subsequent litigations,e ssentially all of those attempts started with commercial Form 2;hence,atthe very least Form 2-thus seeds of Form 2-were present in the environment in which attempts to follow Example 32 were being carried out.
Following numerous attempts to prepare Form 1 according to Example 32, which led almost exclusively to Form 2, the Canadian generic firm Novopharm claimed that Glaxo had never made Form 1 and sought approval from the Food and Drug Administration (FDA) to market Form 2.G laxo aimed to prevent Novopharm (and others) from entering the market with Form 2 by suing them for the infringement (actually,v irtual infringement under the Hatch-Waxman Act) of their Form 2 patent. Novopharm admitted infringement of Form 2,b ut argued that the Form 2 patent was inherently anticipated in the Form 1 patent, since their attempts to prepare Form 1 according to Example 32 led to  Novopharm contended that the experimental procedures underpinning the Form 1 patent were flawed, to which Glaxo responded that the oppositions experiments were contaminated with seed crystals and hence not af aithful reproduction of Example 32 [Clearly,t here were no seeds of Form 2 anywhere prior to April 15, 1980].
Thel egal concept of inherency in the United States implies ac onsistent result of ap rocess,t hat is,i tm ust be invariable or inevitable that one obtains the later claimed result to establish inherency.T hus,tosupport its case against inherent anticipation, in principle Glaxo had to demonstrate that Example 32 did not inevitably or invariably yield Form 2 but could in fact yield Form 1.
To do so,i nt he course of the August 1993 trial, the notebooks of David Collin, who been the first to prepare ranitidine hydrochloride were examined, cross-examined, and compared to the wording in Example 32. Collins notebooks contain three slightly different examples.A si sc ommon for al aboratory notebook, the texts are not word-for-word identical nor is any one identical to the language in Example 32, and there was much discussion over the differences and what they would mean to apracticing chemist (one "skilled in the art" in patent lexicon).
In addition, one of Glaxos witnesses,Sir Jack Baldwin of the University of Oxford, in 1993 had two of his senior postdoctoral fellows complete the entire synthesis of ranitidine base according to the Form 1 patent followed by the reproduction of Example 32 using that prepared base.T hey also obtained Form 1 three times.
Those six instances of the preparation of Form 1 according to Example 32 were sufficient to overcome the inherency argument. The Form 2 patent was found to be valid and Novopharm (and others) were restricted from marketing Form 2 prior to its anticipated expiration in 2002.
Legal footnotes. An umber of litigation cases ensued. It was surprising that Glaxo could no longer make Form 1 in the original pilot plant, but even so,a tt he time,t he concept of disappearing polymorphs and the role of unintentional seeding were treated with skepticism by those who had no personal experience of the phenomenon. Fori nstance, counsel for Novopharm included the following in his opening statement to the court: [31] "Theres also testimony in this case which is under aprotective order from athird party pharmaceutical company that did the same thing. They reproduced example 32 and got Form 2.S ow eh ave six different locations or incidents where example 32 had been reproduced to yield Form 2,not Form 1. Whats Glaxos response to this?S eed crystals,t heyre in the air.Y ou cant see them. Youc ant smell them. Youc ant taste them. Youa lso cant detect them but theyre there,a nd these seed crystals fall out of the sky,a nd theyre very intelligent because they know when youre running one of these example 32 experiments.T hey fall out of the sky and they fall in your reaction beaker and it causes not Form 1 to be produced, but Form 2,and thats why when we run this experiment todaywe get the Form 2 product and not the Form 1." "Well, Isubmit that if one believes in Santa Claus we might believe in these seed crystals,but if were beyond that, were not going to believe in these seed crystals,a nd even if you do,t he techniques that were used in these reproductions would, without ad oubt, exclude these seed crystals because these seed crystals to survive in the method that has been used for these reproductions have to defy standardc hemical principles".
Some excerpts from the cross-examination of one of Novopharms witnesses regarding seeding: Question (cross-examining attorney): "Ithink the issue of seeding is one that Iw ould have expected to come from ac rystallographer.H ave you made as tudy of the subject of seeding?" Answer:"Ive found in my experiments that Icant see any seeding effects." Question:"Youfound that you cant see.Myquestion was, have you done astudy of the science of seeding whichtakes in account the myriad of works of those who can see.H ave you made such astudy?" Answer:"Ive done aliterature search to see if atheoretical phenomena (sic) like the hypothetical theory of universal seeding could be found in all of the chemical abstract literature, and the only references If ound to something called universal seeding had to do with entries like,u niversal prevention of fungus on weed seeds by using certain different fungicides. Thats the only type of reference Icould come up with when I scanned the chemical literature." Question:" … Now,d id you go to any meeting of professional crystallographers or did you consult crystallographers to see whether there was ab ody of knowledge that you hadnt found?" …Answer:" Ir eported my negative findings to the [attorneys]."

Ritonavir
Perhaps the most notorious recent example of ad isappearing polymorph is that of ritonavir (Figure 2a), an antiviral compound marketed by Abbott Laboratories in 1996 as Norvir in the form of semisolid gel capsules for the treatment of the acquired immune deficiencys yndrome (AIDS). Thec apsules were based on the only known crystal form, Form I,discovered during the development process.In 1998, however, anew and significantly less-soluble polymorph of ritonavir unexpectedly precipitated out in the semisolid gel capsules,t hereby leading to failed dissolution tests for the capsule. [32,33] Subsequent studies showed that the new form, referred to as Form II,e xhibited as ignificantly lower solubility in hydroalcoholic solutions than the marketed Form I. [33] In addition, it was found that Form II rendered Form I unobtainable in any laboratory to which Form II had been introduced. There was even speculation that the conversion of Form I into Form II in the laboratory was facilitated simply by the presence of an individual who had previous exposure to Form II (or the contaminants that were subsequently shown to enable the formation of Form II). As ar esult of these events,r itonavir had to be temporarily withdrawn from the market. [34] Crystallographic analyses showed that the crystal structure of Form I is characterized by ritonavir stacks resembling Polymorphism Angewandte Chemie a b-sheet structure ( Figure 2b). [32] Thes tructure is sustained by N-H(amide)···O(amide) and O-H(hydroxy)···N(thiazole) hydrogen bonds (with first-order graph set N 1 = C(4)C11) [35] ). Thec rystal structure of Form II,o nt he other hand, is comprised of heavily hydrogen-bonded one-dimensional ritonavir stacks (Figure 2b). [36] Each molecule in the stack is hydrogen bonded to two other molecules through at otal of eight N-H(amide)···O(amide), N-H(amide)···O(hydroxy), and O-H(hydroxy)···O(amide) hydrogen bonds (N 1 = C(6)C(9)C(11)C(12)). [32,35] Thehigher calculated crystal density of Form I suggested it is also the more stable crystal form. [37] In addition, as urvey of the Cambridge Structural Database (CSD) indicated that Form I ritonavir exhibits astatistically more favorable conformation of the carbamate moiety. [32] Theanalysis was in agreement with an NMR study in solution that revealed the existence of two conformers in solution in aratio of roughly 99:1. Theconformers could not be unambiguously resolved as Forms I and II,but it was noted that the observed 99:1 relationship of the two conformers is consistent with the initial discovery of as ingle polymorph, that is, Form I. [32] Theh igher stability of Form II was,i nt he end, attributed to the formation of ah ydrogen-bond pattern wherein, unlike in Form I," all of the strong hydrogen bond donors and acceptors have been satisfied". [33] This argument is consistent with the observation that Form II has ah igher melting point and heat of fusion (ca. 125 8 8C, 87.8 Jg À1 )t han Form I (ca. 122 8 8C, 78.2 Jg À1 ). [33] Arecent logistic regression hydrogen-bonding propensity study involving Forms I and II (using the CSD as data source) reported that the kinetically more favored Form I displays astatistically doubtful hydrogen-bond pattern. Specifically,it was found that Form I entails statistically improbable hydroxy-thiazoyl and ureido-ureido interactions-despite the hydrogen-bond donors and acceptors availability for the realization of high-propensity hydrogen bonds. [38] Theo rigin of Form II was initially unclear,a si tw as established that ritonavir solutions would crystallize as Form II only if seeded with Form II-even at amounts as low as 1ppm. Heterogeneous nucleation was identified as ap ossible cause of the formation of Form II.S pecifically,i t was found that ritonavir degrades in abase-catalyzed reaction to form acarbamate-bearing product ( Figure 2c)structurally related to the conformation of ritonavir in Form II. [32] It was also found that the degradation product forms very rapidly and that, consistent with its greater stability,itexhibits alower solubility than ritonavir.I tw as concluded that the degradation product had possibly crystallized out of ar itonavir bulk solution, which had experienced solvent loss,a nd then acted as as eed for Form II.
Eventually,c onsistent with the closing statement in the 1995 review,e xtensive studies demonstrated that the crystallization of Form I could be achieved under controlled conditions in laboratories that had previously been exposed to Form II. [33] Ritonavir was finally reformulated and approved in 1999 before being placed back on the market. [39] It was estimated that the company had experienced losses in revenue of over US$ 250 million. [34] Am ore recent study attempted ah igh-throughput polymorph screen of ritonavir to comprehensively explore the compounds structural diversity.T he screen included about 2000 crystallization experiments and resulted in the finding of three new crystal forms in addition to Forms I and IInamely,ametastable polymorph, at rihydrate,a nd af ormamide solvate. [40] These findings highlight the necessity of utilizing avariety of crystallization methods,inthis particular case,h igh-throughput screening, combined with carefully designed crystallization experiments for the retrieval of the relevant structures associated with the structural landscape [41] (also referred to as packing landscape [13] )o famolecule.
Public relations footnote. As noted, Abbotts initial encounter with Form II and its inability to produce Form I led to the disappearance of the drug Norvir from the market, leaving tens of thousands of AIDS patients without medication. This led to as erious public relations problem for Abbott. To allay public concern, the company held anumber of interviews and press conferences,a tw hich senior Abbot officials appeared in order to answer questions.T he transcripts were originally published on the website [42] of the International Association of Physicians in AIDS Care,but no longer appear there.S ome excerpts vividly portray the situation that can arise when ad isappearing polymorph is encountered: "There was no gradual trend. Something occurred that caused the new form to occur…There was no early warning." "We,q uite honestly,h ave not been able to pinpoint the precise conditions which led to the appearance of the new crystal form. We nowknow that the new form is,infact, more stable than the earlier form, so nature would appear to favor it…Form II is new." "We did not know how to detect the new form. We did not know howt ot est for it. We did not know what caused it. We didnt know howt op revent it. And we kept asking the question, why now?…Wedid not know the physical properties of the new form…Wedid not know howtoclean it, and we did not know how to get rid of it." "…our initial activities were directed towarde liminating Form II from our environment. Then we finally accepted that we could not get rid of Form II.Then our subsequent activities were directed to figuring out howtol ive in a Form II world." "This is why all of us at Abbott have been working extremely hardthroughout the summer [of 1998],often around the clock, and sometimes never going home at night. We have been here seven days aweek and we will continue to do so.W e have cancelled vacations and asked our families for their understanding and support. This is not an issue that we take lightly." "There were several sub-teams of three to 600 people per team working full time in different areas.W ealso called on as many resources as we could." "We tried everything.W econducted countless experiments. We reconditioned our facilities.W er ebuilt facilities and new lines.W el ooked at alternative sites.W ev isited an umber of [other] organizations around the world…to see if we could start clean in anew environment free of Form II." "In amatter of weeks-maybe five or six weeks,every place the product was became contaminated with Form II crystals." Question: "Youa re al arge multinational company.Y our scientists are obviouslys mart. How could this happen?" Answer:" Acompanys size and the collective IQs of their scientists have no relationship to this problem…This obviously has not happened to every drug. But it has happened to other drugs."

Paroxetine Hydrochloride
Paroxetine hydrochloride is as erotonin re-uptake inhibitor used for the treatment of depression. Thec hemical compound paroxetine was initially developed by the Danish company Ferrosan in the 1970s.B eecham (now part of GlaxoSmithKline,GSK) purchased the rights to paroxetine in 1980 and undertook development of the hydrochloride salt of paroxetine as ad rug product. Beecham developed ap rocess that produced paroxetine hydrochloride in acrystalline form that was later referred to as the "anhydrate" crystalline form. Late in 1984, however, in the course of pilot plant scale-up, the "hemihydrate" crystalline form suddenly appeared at two Beecham sites in the UK within af ew weeks of each other. Then ew hemihydrate form was designated Form 1 and the previously existing anhydrate was labeled Form 2 ( Figure 3).
Theh emihydrate was not hygroscopic and exhibited handling properties superior to those of the anhydrate.I n 1986 GSK applied in the US for apatent on the hemihydrate, which was granted in 1988. [43] Paroxetine hydrochloride was finally marketed as the hemihydrate form in 1993 under the name Paxil . [10,44] During the 1980s in the course of the development of the compound for eventual launch, Beecham investigated the properties of both the anhydrate and hemihydrate.T hey determined that in the presence of water (or humidity) the anhydrate undergoes ac onversion into the hemihydrate, aprocess that is accelerated by temperature,pressure,and the presence of seeds of the hemihydrate. [45] In Beechams experience,i tw as difficult to avoid the conversion of the anhydrate to hemihydrate in the presence of water or humidity in afacility seeded with hemihydrate.
In 1998, Apotex, aC anadian generic drug manufacturer, filed an Abbreviated New Drug Application (ANDA) with the FDAt om arket ag eneric version of the off-patent anhydrous paroxetine hydrochloride (Form 2). Theh emihydrate patent would expire in 2006. Again, under the terms of the Hatch-Waxman Act GSK opposed the Apotex request, arguing that the anhydrous form (Form 2)would convert into the hemihydrate (Form 1). GSKs argument was based, in part, on evidence that Apotex had begun development of its anhydrous product by bringing the hemihydrate into its manufacturing facility,thus providing the seeds that had been shown to be afactor in the conversion. In addition, there was contact with water in the manufacturing of the active pharmaceutical ingredient (API), in the formulation, and in the production of the final water-based coating process of the pill, as well as the use of pressure in the last processing step.
At rial was held in February,2 003 in the US Federal Court, Chicago,w ith Judge Richard Posner presiding. [46] Judge Posner is one of the most cited judges in the history of the US federal courts,h aving written over 2500 published decisions,a nd as aP rofessor of Law at the University of Chicago has published nearly 40 books.
GSKs assertion that Apotex would infringe the hemihydrate patent was based on GSK and Apotex documents showing that many of Apotexs anhydrate batches had Polymorphism Angewandte exhibited evidence of conversion:t here were batches of anhydrate that converted almost entirely into hemihydrate when stored at 40 8 8Ca nd 75 %h umidity within one month. This Apotex experience was bolstered by the results of GSKs testing of Apotexs API and its formulated tablets.F urthermore,Apotexs specification for release of bulk material was based on av isual method of comparing spectral data that could not detect less than 5-8 %ofinfringing hemihydrate in the bulk API.
In its defense,A potex argued that seeding is "junk science", not widely accepted in the scientific community and that the mechanism of the role of seeds and the conversion is not understood. Moreover,itclaimed that the supplier of the bulk API had improved the process to avoid contact with water and was storing the material in improved storage bags, less permeable to water.
In response,G SK argued that Apotex produced the tablets under conditions of normal humidity and sprayed the tablets with an 88 %water-based aqueous coating.
Judge Posner rejected Apotexs contentions concerning seeding,s tating "that there is no scientific basis for believing that seeding occurs…is obviously wrong." In his ruling on the case he found that Apotexs anhydrate converts into the hemihydrate (in accord with the earlier publication by SmithKline Beecham scientists) and that it "may continue until it reaches 100 per cent." He also found that Apotexs limit of detection of the hemihydrate in an allegedly anhydrous material was 5-8 %, but he did not rely on this finding in determining whether Apotex was infringing the GSK patent.
Nevertheless,J udge Posner ultimately ruled in Apotexs favor. Claim 1o fG SKs hemihydrate patent recited "Crystalline paroxetine hydrochloride hemihydrate". Judge Posner ruled that this was valid, but that Apotexs product would not likely infringe the patent because Apotex would not be making it intentionally,a nd not "in any commercially significant quantity".J udge Posner interpreted Claim 1a s limited to only "commercially significant amounts of hemihydrate"and explained that the concentration would have to be in the "high double digits to contribute any commercial value". He further stated that GSK had not established that Apotex would be marketing material with high double-digit concentrations of hemihydrate and that Apotex would not benefit monetarily from the hemihydrate.
Judge Posner thus found the patent valid, but also that Apotex would not infringe it. Thec ase was appealed to the US Federal Circuit Court, which handles all patent appeals in the United States.T he Federal Circuit ruled in favor of Apotex, but for different reasons than those invoked by Posner.T he Federal Circuit also found that Judge Posners claim construction was incorrect, and that the claim properly covered any amount of hemihydrate.T he Federal Circuit, however, reasoned that Claim 1must, therefore,beinvalid for inherencyb ecause if anhydrate converted into the hemihydrate now,i tm ust have converted into hemihydrate in at least small amounts in the prior art. TheU SS upreme Court refused to hear the case.
Many aspects of those rulings,a nd the way they were subsequently reported in the trade press,deal with important aspects of disappearing polymorphs and seeding.D etails are provided in the Addendum of this Review.
Avery recent study demonstrates that it is not possible to claim that all the probable crystal forms of acompound have been found or even fully characterized. Pina et al. showed that Form 2,w hich had initially been described as ah ygroscopic anhydrate,w as in fact an onstoichiometric hydrate, [47] which dehydrates and rehydrates very easily,despite alack of continuous channels in the crystal lattice.The higher stability of Form 1 was justified by ahigher number of hydrogen bonds being involved in retaining the water molecules in the crystal lattice, [47] although no quantitative estimates of the relative stabilities were calculated.

Paroxetine Methane Sulfonate (Paroxetine Mesylate)
Them arket success of paroxetine hydrochloride (worldwide sales of US$ 3.2 billion in 2001) led many pharmaceutical companies to search for additional crystal forms.O ne approach is to prepare ad ifferent salt. In the mid-1990s SmithKline Beecham (now GSK) in the UK and Synthon in the Netherlands independently succeeded in making the mesylate (methanesulfonate) salt of paroxetine.C onsidering the nature of the chemistry,i ti sn ot surprising that the procedure leading to the salt is similar in the two patents.The first US patent was issued to Synthon on 23.02.1999. Another US patent was issued to SmithKline Beecham nearly 15 months later on 16.05.2000.
How could two patents be issued for the same salt?O ne clear possibility is that they are two different crystal forms of the same salt, which should be discernible from the analytical data presented in the patent. Table 1contains acomparison of the relative data from the two patents.Clearly,the only data that can be compared are those from the IR spectra. TheI R data from the two patents are given in Table 2. Comparison of the two peak lists indicates that they do not characterize the  . .
same crystal form. In the course of litigation it became apparent that every reported synthesis of paroxetine mesylate (since the issuance of the Synthon patent) has yielded the peaks found in the SmithKline Beecham patent. There are only two explanations for this situation:1)the Synthon form is ad isappearing polymorph, having been prepared and characterized at least once,and subsequent preparations and crystallizations led to the SmithKline Beecham form, or 2) the Synthon data are in error.
As eries of litigations on the issues associated with the paroxetine mesylate reported in these two patents (and their European equivalents) failed to achieve alegal consensus on the issue.Some of those litigations are summarized in Table 3.
Although not reported in the literature,t he paroxetine mesylate case is an outstanding example that underpins the necessity to thoroughly investigate,c haracterize,a nd document APIs and speciality chemicals.

Rotigotine
Rotigotine (Figure 4a)i sanon-ergot-derived dopamine agonist initially prescribed for the treatment of Parkinsons disease,a nd later approved for moderate-to-severe cases of restless-legs syndrome.I ti sm arketed by UCB under the name Neupro ,a nd administered through at ransdermal patch to minimize the unpleasant side effects of the drug. [48] Neupro was approved by the European Medicines Agency (EMEA) for use in Europe and then by the FDA for the US market in 2007. In 2008, apreviously unknown and thermodynamically more stable polymorph emerged in the Neupro patches,inthe form of "snow-like crystals". Thenew polymorph was unanticipated and unexpected, as the drug had been established since the 1980s and no polymorphism had been observed during drug development or thereafter. [48,49] While the new polymorph (Form II)e xhibited no reduction in efficacy, physicians,distributors,pharmacists,and patients were advised to refrigerate their Neupro stocks, since refrigerated storage significantly reduced crystallization rates.U CB continued to supply Neupro in Europe,b ut specific batches were recalled and replaced by batches that were refrigerated immediately after manufacture.W hile there was "only" am inor disruption in Neupro supply in Europe,the situation was much more serious in the US,where Neupro became temporarily unavailable.A fter the FDA recommended reformulation of the drug, UCBs new Neupro formulation that did not require refrigeration was approved by the agency in 2012.
Theorigin of the more stable Form II of rotigotine is not known, but it is reasonable to speculate that the polymorphic transformation was suddenly triggered by an event, or an impurity,inthe patch or the drug itself.The initially observed and long-known polymorph (Form I)c rystallizes in the tetragonal P4 3 space group, [50] while Form II crystallizes in the orthorhombic P2 1 2 1 2 1 space group. [51] Both polymorphs feature disordered thiophene moieties and similar hydrogenbonding patterns of one-dimensional zigzag chains of O-H···N hydrogen bonds [N 1 = C(8) [35] ]. However, Form II is more dense (and thus likely more stable,a ccording to the Burger-Ramberger density rule [37] )and accompanied by aconformational change caused by the adjustment of the torsion angle between the thiophene and alkyl moieties.

DMP 543
DuPont entered the development of pharmaceutical agents for the treatment of Alzheimers disease in the 1980s. Extensive studies finally resulted in the development of DMP 543-an acetylcholine release enhancer with the desired potency, plasma duration, and brain penetration properties ( Figure 5). It became evident in the early stage of the drugs development that DMP 543 was very susceptible to poly-

Angewandte
Chemie morphism, as 17 polymorphs were produced and characterized by powder X-ray diffraction. [52] Thea uthors attributed the ability of DMP 543 to form such al arge number of polymorphs to the conformational flexibility of its pyridyl groups,although it has recently been determined that there is no statistical correlation between molecular flexibility and the tendency to polymorphism. [29] It was found that some polymorphs interconvert easily on heating. But more intriguingly,a pparently identical recrystallization conditions would not always lead to the formation of the same polymorph. A robust method for the crystallization of as ingle polymorph was finally established using an ethyl acetate/heptane solvent mixture.The first process batch, however,yielded apreviously unknown polymorph (#18). Although this procedure initially appeared robust, researchers were concerned that the preparation of the solvent mixture might not be reproducible and that differential solvent evaporation rates could, in future experiments,trigger the formation of other new and possibly unwanted solid forms.T his led to the decision to produce ac rystallization procedure based on as ingle solvent. It was then found that isopropanol reliably produced another new polymorph (#19, Form A)i nh igh yields,a nd this polymorph was finally chosen for further development.
After the initial batch of Form A was prepared, the synthetic procedure for DMP 543 was refined, and this modified procedure was utilized to prepare the second batch. Thesynthesis of the second batch began in Deepwater, NJ (USA), but was completed at adifferent location, namely the Merck Frosst Centre in Dorval, Quebec,w here the clinical trials were intended to be conducted. During the course of the synthesis,t he anthrone alkylating agent had to be purified three times (by two recrystallizations and one chromatographic purification) before the compound was finally recrystallized from isopropanol to achieve 98.5 % purity,w hich was the lowest purity grade specification acceptable for clinical trials.Asecond recrystallization of the solid, however,r esulted (to everybodys surprise) in the formation of yet another new polymorph (#20, Form B). Three subsequent recrystallizations were performed using the new form, utilizing seeds of Form A with the hope of producing alarge batch of Form A.All three recrystallization experiments yielded Form B.Attempts to prepare Form A at DuPont in Deepwater resulted exclusively in the formation of Form B-very shortly after Form B was discovered at the Merck Frosst site in Dorval. [52,53] DuPont was never again able to produce Form A. Form B turned out to be the most stable of all polymorphs according to thermal analysis and was, therefore,s elected as the preferred crystalline form of DMP 543 for production. [53] Although no new polymorphs were found in the following five years,D MP 543 was never commercialized.

LAB687
Another extraordinary account of the unpredictability of polymorphism comes from Novartis,and involves acompound internally identified as LAB687 (Figure 6a), [54] an inhibitor of the microsomal triglyceride transfer protein developed for the control of triglyceride and LDL-cholesterol levels. [55] Tw o polymorphs were found during drug development:t he original synthetic route yielded Form B,w hile ag ram-scale synthesis based on ad ifferent synthetic procedure led to the formation of Form A.
As ubsequent polymorph screen was conducted using a9 8.9 %p ure batch of Form A,w hich yielded at hird polymorph (Form C), as well as at oluene solvate.T wo solvates based on heptane methylcyclohexane were then later discovered during the development of seeding-based crystallization procedures for the reproducible larger-scale (100 g) synthesis of Forms A and C. Although Forms A and C were found to have similar intrinsic solubilities and physical stabilities, Form C was selected for further development because it had better filtration and flow properties than Form A.Remarkably,once Forms A and C were discovered, Form B could not be reproduced. Anew polymorph, Form D, unexpectedly appeared when the crystallization process for Form C was scaled up for am ulti-kilogram synthesis.O ddly, once Form D emerged, Forms A and C could no longer be prepared. It was assumed that Form D appeared due to achange in the impurity profile of the LAB687 batches during the implementation of an ew and more-efficient phase I synthetic route for one of the LAB687 intermediates.Indeed, this route led to the formation of ad imeric urea byproduct (i.e.a ni mpurity,F igure 6b)t hat was not present in batches obtained using other synthetic routes,a lthough no direct  correlation could be determined between the formation of this impurity and the appearance of Form D.

Sulfathiazole
Theu nexpected formation of polymorphs of the antimicrobial compound sulfathiazole (Figure 7a)i sa lso triggered by impurities or reaction byproducts.F ive polymorphs of sulfathiazole have so far been identified (Figure 7b-f) [56][57][58][59][60] in addition to over one hundred solvates. [61] Thes olid-state chemistry of the sulfathiazole polymorphs has been thoroughly analyzed [62,63] and numerous researchgroups reported different methods for the preparation of each polymorph. [62] It is,however, also reported that some methods cannot be used with confidence for the formation of phase-pure batches of the targeted sulfathiazole polymorphs,s ince different research groups reported different outcomes using the same crystallization conditions. Ar ecent review [64] addressed the widespread belief that aparticular sulfathiazole polymorph is consistently accessible from agiven solvent, and suggested that this view was in fact misleading. Thea uthor shared his personal experience from over 2000 crystallization experiments involving sulfathiazole and revealed that all five known polymorphs can be obtained from nearly every solvent used in his comprehensive studies. It was also stated that the polymorphs are enantiotropically related and that the solids treatment after crystallization is most critical in determining the polymorphic outcome. Indeed, the sensitivity of the sulfathiazole solid towards post-crystallization treatment is in keeping with the inconsistent results reported in the literature,which shows that the same (or very similar) crystallization conditions can lead to the formation of different sulfathiazole polymorphs.
Despite the inconsistencies in the reported crystallization outcomes,i th as been established that the crystallization of sulfathiazole from water follows Ostwalds rule of stages, [12] whereby the least-stable polymorph (Form I)a ppears first, and transforms into Forms II and III before the transformation ends with the appearance of the thermodynamically stable Form IV. [63,65] Notably,i th as also been demonstrated that the transformation (and disappearance) of the metastable Form I can be suppressed by the presence of impurities in the crystallization solution. More specifically,acompound that forms as ab yproduct of the sulfathiazole synthesis (i.e. ethamidosulfathiazole) was used to show that concentrations as low as 10 mol %s tabilize metastable Form I,w hile amounts of 0.5-1.0 mol %y ielded am ixture of the four Forms I-IV.Pure solutions,aswell as those with an impurity content of 0.01 mol %, were shown to yield Form IV. [65] The ability of ethamidosulfathiazole to stabilize Form I was attributed to its capacity to integrate into the growing crystal faces of Form I without disrupting crystal growth. Thegrowth of Forms II, III,and IV,onthe other hand, becomes inhibited once ethamidosulfathiazole becomes attached to their growing crystal faces.
Although sulfathiazole polymorphs do not tend to "disappear for good" as claimed for some other polymorphs described in this Review,t he sulfathiazole system seems unique because of its complexity and sensitivity,a nd high-

Angewandte
Chemie lights the need to understand how byproducts obtained from the synthesis of at arget compound can profoundly affect its solid-state chemistry.

Progesterone
Crystal structure prediction (CSP) has enjoyed ad ecade of increasing success rates and an expanding range of applications in the study of molecules of increasing complexity. [66,67] Ar ecent study suggested that progesterone (Figure 8a)m ay be au seful model compound with which to test advances in CSP,a sas uitable "real" system of pharmaceutical interest. [68] Steroids are the basis of fundamental hormones and many drugs;t hey have rigid structures,a nd progesterone is ar elatively simple example of as teroid. Beyond that, steroids are generally relatively easy to crystallize and they often exhibit polymorphism. Progesterone was chosen as am odel compound for CSP based on the above criteria, with as upposedly well-understood polymorphic system documented in the scientific literature stretching back 70 years.
Thepredictive studies correctly identified the two known polymorphs (Forms I [69] and II, [70] Figure 8b,c) in enantiomorphic space groups.C SP studies also strongly indicated that there were an umber of more-stable centrosymmetric structures with one at the global minimum. This lowestenergy structure was ultimately crystallized by mixing natural (nat)p rogesterone with its enantiomer ent-progesterone. When this work was initiated, centrosymmetric structures were not considered because they do not exist naturally for this chiral molecule,but the results from the CSP studies were so compelling that the experimental search for ar acemic structure was undertaken. During the course of this study it became apparent that the metastable low-melting-point Form II of nat-progesterone was an example of adisappearing polymorph, as attempts to make it soon failed and attempted crystallizations became erratic in their polymorphic outcome. Attempts were then made to template progesterone with as tructurally related steroid, pregnenolone (Figure 8a), which resulted in the successful preparation of Form II along with the concomitant formation of a1:1 progesterone: pregnenolone cocrystal. [71] Themetastable form produced by these means was stable for periods ranging from hours to two or three months! About this time it transpired that the University of Innsbruck had samples of both polymorphs in an archive. Their metastable polymorph had not transformed in over 50 years!T he chemical analysis of the long-lasting Form II sample revealed impurities that were not present in the sample of Form I,t hus highlighting the potential role of contaminants/additives in stabilizing metastable crystal forms. Attempts were made to analyze the impurity profiles associated with the archived samples,w ith av iew to doping crystallizations involving modern much purer commercial material. These were unsuccessful owing to the complexity of the impurity profiles associated with the old samples. [72] 3.10. The Elusive Form II of Aspirin Aspirin (acetylsalicylic acid, Figure 9a)i sawidely used analgesic, and its antiplatelet activity makes it ac ommonly prescribed long-term preventative agent for reducing the risk of heart attacks and strokes. [73] Although first synthesized  more than 150 years ago,o nly one polymorph [74] of aspirin (Form I)w as known until 2005. Thep otential polymorphism of aspirin was extensively studied and debated in the 1960s and 1970s, [75] but the existence of another polymorph was not definitely established with certainty. [76] Astudy from the early 1980s reported that aspirin, if crystallized in the presence of aspirin anhydride (Figure 9a), exhibits the characteristic aspirin Form I diffractogram, along with weak additional peaks. [77] It was concluded that these are not likely to correspond to an ew aspirin polymorph, but rather belonged to an impurity in the form of ac ocrystal or solid solution composed of aspirin and salicylic acid. [77,78] It took morethan 30 years to discover that these "impurity" peaks are actually related to another polymorph, Form II.
Ac omputational study published in 2004 predicted that the known crystal Form I is the most stable,b ut it also predicted the existence of as econd polymorph that exhibits virtually the same crystal lattice energy as the known form. [79] Thee lusiveness of this polymorph was ascribed to its predicted low shear elastic constant, which suggests that the polymorph exhibits an energetic low barrier to transformation into Form I.Alater study concerning aspirin cocrystals led to the discovery of the new polymorph, Form II. [80] The Form II solid was obtained in the course of an attempt to cocrystallize aspirin with levetiracetam in a1 :1 ratio,a nd although its structure was derived from crystallographic data of lower quality, [80,81] it was in good agreement with the computationally predicted crystal structure of the elusive lowenergy polymorph. Forms I and II of aspirin are structurally (and energetically) very similar. [79] Both forms feature centrosymmetric aspirin dimers held together by the carboxylic acid homosynthon. Thed imers form two-dimensional layers parallel to the crystallographic c axis.T he two forms differ in the relative positions of the neighboring dimer layers, as well as the symmetry elements between them (Figure 9b,c).
Soon after the initial discovery of Form II,t wo studies provided evidence that the initially reported Form II is in fact an inter-grown phase [82] containing domains with structural features corresponding to both Forms I and II. [83,84] It should be noted that no solid exhibiting the structural features of Form II alone could be isolated at that point. It was also reported that the inter-grown phase could only occasionally be prepared by recrystallization of freshly synthesized aspirin from acetonitrile or tetrahydrofuran. Interestingly,c ommercial samples crystallized under the same conditions only yielded crystals of Form I.I tw as then discovered that the diffractograms of the inter-grown crystals displayed weak peaks belonging to the aspirin anhydrate,thus indicating that the anhydrate might have played as ignificant role in the formation of the inter-grown phase. [78] Further investigations demonstrated that the inter-grown crystals can be regularly prepared using aspirin anhydrate as seed material. It was also shown that seed quantities of up to 10 wt %yield inter-grown crystals with as ubstantial presence of Form II domains. Finally,t he seeding experiments have also established that phase-pure batches of the elusive Form II can be reliably prepared if aspirin solutions are seeded with 15 wt %o ft he anhydrate. [78] 3.11. The Elusive (caffeine)·(benzoica cid) Cocrystal Recently,m olecular cocrystals have been attracting the attention of pharmaceutical and materials scientists,primarily because of their potential ability to alter the physicochemical properties of molecular compounds [85][86][87][88][89][90][91][92][93][94][95][96][97] while maintaining the pharmaceutical activity of the active pharmaceutical ingredient. This deepened interest in cocrystals has led to the development of increasingly sophisticated crystallization techniques, [98,99] which are generally used during cocrystal screening in an integrated fashion in the search for new (co)crystal forms of adrug candidate.When cocrystallization attempts fail, it is difficult to know whether the failure was due to poorly chosen experimental conditions or because the cocrystal simply cannot form based on thermodynamic considerations,t hat is,t he cocrystal lattice energy is higher than the lattice energy of the cocrystal components.
Caffeine is one of the most utilized model compounds in studies of pharmaceutical cocrystals,a nd has been shown to engage in cocrystallization with aw ide variety of carboxylic acids; [100,101] despite this,t he literature is replete with reports that caffeine does not form ac ocrystal with benzoic acid. [102,103] Fore xample,arecent study describes the efforts of four research groups to prepare the elusive cocrystal using neat grinding,liquid-assisted grinding,and solution-mediated phase transformation. [104] After all attempts failed, CSP methods [105,106] were employed to evaluate the potential existenceo fa1:1( caffeine)·(benzoic acid) cocrystal, and showed that the formation of the target cocrystal is indeed thermodynamically favored.
TheCSP work also aimed to identify appropriate heteronuclear seeds for the cocrystallization of caffeine and benzoic acid. It was presumed that ahigh kinetic barrier hindered the formation of the cocrystal, and it was proposed that this barrier could be overcome by introducing ah eteronuclear seed, which matched the target cocrystal structurally or epitaxially.F luorinated benzoic acids were used to form cocrystals based on molecular assemblies similar in shape and size to those present in the putative (caffeine)·(benzoic acid) cocrystal.
Therationale behind the use of fluorinated benzoic acids lay in the relatively small size difference between the van der Waals radii of hydrogen and fluorine. [107] Thes trategy was successful:h eteronuclear-seeding experiments yielded the target cocrystal in all four laboratories where "seedless" cocrystallization attempts previously failed ( Figure 10). Interestingly,o nce the heteronuclear seeds were introduced to al aboratory,t hey-or the seeds of the product cocrystalcontinued to act as long-lasting laboratory "contaminants" that enabled cocrystallization even when present at undetectably low levels,a no bservation consistent with Wibergs earlier observation. [22] Theq uest for the (caffeine)·(benzoic acid) cocrystal demonstrates the utility of CSP calculations in assessing the likelihood of cocrystal formation. At the same time,the study stresses that current cocrystal screening methods need to be improved to eliminate the occurrence of false negative results that could impede the development of functional multicomponent crystalline materials.T his study highlights the Polymorphism Angewandte gaps in our current understanding of the nucleation process of cocrystals and of how laboratory contaminants may affect the outcomes of crystallization experiments.
Ar elated study recently demonstrated that seeding is indeed ap ractical method of crystallizing anticipated solids that are inaccessible at ambient conditions.S pecifically, am onohydrate of the neurotransmitter g-aminobutyric acid was obtained at high pressure and subsequently recovered at ambient conditions at which its crystallization was unsuccessfully attempted in numerous trials.T he high-pressure polymorph could then be consistently used to produce the elusive hydrate at ambient conditions. [108]

Recovering Disappeared Polymorphs
As noted above,a ta ny particular temperature and pressure,t he Gibbs Phase Rule permits the existence of only one thermodynamically stable polymorph of asubstance. However,kinetic stability allows the coexistence of more than one form. It is,therefore,possible in principle to prepare and maintain an umber of crystal forms at ambient conditions without limitation.
In many of the cases of disappearing polymorphs described above,t he form that disappeared was the only one known until an ew form appeared-often as ar esult of the same procedure that previously led to the now absent form. In most cases this means that among the known forms, the new form is the thermodynamically preferred, but not necessarily the most stable form under those conditions.Since the disappeared form had been prepared and characterized (often many times and over long periods of time), it must occupy its own definite region in phase space,e ven if it becomes very difficult to prepare it again. Thea lmost inescapable conclusion from this situation is that the most practical strategy for recovering adisappearing polymorph is to employ kinetic crystallization methods rather than thermodynamic crystallization methods.The dominance of anew form is often also aresult of aggressive seeding by that form, thus indicating that those seeds must be assiduously avoided to prepare the disappeared form. Thefollowing examples will demonstrate these principles.
One of the first detailed studies of conformational polymorphism involved the study of dimorphic p,p'-dichlorobenzylideneaniline. [109,110] Both formsw ere grown from ethanol solutions.T he metastable triclinic form initially crystallized as transparent needles with a4long axis parallel to the needle axis.Cleaving the crystals perpendicular to the needle axis would induce atransformation to the orthorhombic form that could be detected by an increasing cloudiness of the crystal and ac oncomitant loss of single crystallinity.O ver ar elatively short period of time (i.e.afew weeks), as the amount of the orthorhombic form increased in the laboratory, it became increasingly difficult to obtain the triclinic form. Them ethod that finally produced the triclinic form quite consistently (but not always!) was to prepare am aximally saturated solution in boiling ethanol (a beaker seemed to work better than an Erlenmeyer flask) and to immediately place the solution in aclosed desiccator freshly charged with CaCl 2 and minimizing the contact of the solution with the laboratory atmosphere that no doubt contained seeds of the orthorhombic form. This provided ak inetically biased crystallization, combining the high degree of supersaturation with the fairly rapid cooling and the desiccating power of the CaCl 2 .W hile the triclinic form could be made quite consistently by this method, the eventual solid-to-solid transformation could not be prevented.
We noted above the notion that alaboratory can become seeded with astable form and render it extremely difficult, if not impossible,t op repare the metastable form in that same laboratory environment. Essentially two solutions are possible-but, again, not always successful-to this situation: 1) move to another laboratory (another distant geographical location may be required) or 2) thoroughly clean the laboratory.W ed escribe an example of each of these solutions.
In 1972, one of us (J.B.) prepared the dimethyl analogue of the dichlorobenzylideneaniline described above,a nd found the cell constants to be identical to those reported by Bürgi et al. about four years previously. [111] When we were ready to carry out the crystal structure analysis afew months later the crystals had deteriorated, so the compound was recrystallized using the same ethanol solvent as the previous batch. This resulted in an ew polymorph. Over the next two years, numerous syntheses and recrystallizations that followed failed to yield the original crystal form, although at hird polymorph did appear.
At that time we were moving into anew laboratory afew kilometers distant from the old one where the original experiments had been done.W eh ired an ew student (by telephone) and instructed her to use newly purchased reagents and virgin glassware in the new lab.T he original polymorph was prepared on her first attempt.
Theo ther option to attempt to recover ad isappeared polymorph is to cleanse the laboratory of the culpable crystal form. Such as trategy was employed by Nielsen and Borka with benzocaine:picric acid. [112] Them aterial was used as ap harmacopeial standard in the 1960s.T here is ah igher melting (162-163 8 8C) form that was obtained from the lower melting (132 8 8C) form by drying the latter at 105 8 8Cfor at least one hour or by vacuum drying/sublimation. Once the higher melting form was obtained, the lower melting form could no longer be prepared. In the authors words:" As am atter of curiosity,i to ught to mentioned that once the stable modification was obtained, the metastable modification could no longer be isolated…It was found that after discarding all samples, washing the equipment and laboratory benches and waiting for 8-12 days,t he low-melting modification could be isolated again. This has now been repeated several times in our laboratories." [112] In 1999, we initiatedat horough reinvestigation of this system based very much on hot-stage microscopy, [16] and in addition to the two 1:1c omplexes we confirmed an earlier reported 2:1complex and ahydrate of a1:1 complex. [113,114] In as et of carefully designed experiments we first prepared the low-melting 1:1f orm from as aturated aqueous solution at 80 8 8C, since the hot-stage experiments indicated that the lowmelting form is the stable one above this temperature.T hus, we avoided the presence of seeds of the high-melting form, which was subsequently prepared by an on-aqueous geldiffusion crystallization with both components dissolved in a3:1 CHCl 3 :CH 3 OH solvent mixture.T he 2:1c omplex was obtained over ap eriod of four weeks from a1 :1 mixture in isopropyl alcohol. The1 :1 hydrate was obtained after 48 h from as aturated aqueous solution in as ealed virgin flask at 20 8 8C.
The1995 review described the joint experience (in Zürich and Beer Sheva) with p'-methylchalcone.I nt he 1920s,t he compound was investigated by Weygand for nearly ten years using thermomicroscopy and summarized in a1929 review. [115] Form any years it competed for the title of "world record holder" for the number of reported polymorphs (albeit lacking structure determinations) of am olecular compound with 13 forms.W eygand distinguished seven modifications (called "main forms") as monotropically related with ah igh probability.I no ur hands,i na ccord with the finding of Weygand, once the seeds of the most-stable highest melting form (m.p.758 8C) are present in the laboratory it is virtually impossible to obtain any of the other forms by standard solution crystallization techniques.
It is well known that the polymorphic form may be influenced by the reaction mixture,s ince the material is crystallizing from ad ifferent solution environment. [17,116] p'-Methylchalcone [17] is prepared by as imple condensation reaction, so that some synthetic conditions-at least the solvent and the temperature-may be readily varied. We carried out the base-catalyzed condensation reaction using the appropriate ketone and aldehyde under nine conditions (three solvents:m ethanol, ethanol, and 2-propanol;t hree temperatures:2 0 8 8 C, 4 8 8C, and À13 8 8C) and obtained five thermodynamically unstable forms directly from the reaction mixture. [17] As ac onsequence of their instability,t hey were not easy to handle or characterize,but we did obtain sufficient evidence to positively identify and distinguish them.
In many of these cases of disappeared polymorphs,i fthe old form can be obtained, it often transforms to the new,and presumably thermodynamically more stable,f orm. That situation is by no means universal. Fori nstance,i nt he case of ranitidine hydrochloride,i ns pite of the difficulty of preparing Form 1 in the presence of Form 2 seeds,t he two forms can exist side by side essentially indefinitely,since there is no simple mechanism for the transformation between them.
Recent studies have demonstrated the utility of engineered surfaces [117,118] and heteronuclear seeds [119,120] in crystallizing specific polymorphs and discovering new ones.I ti s possible that such approaches could be utilized to recover polymorphs that had apparently "disappeared".
Thev ariety of circumstances and conditions for these examples demonstrates that each molecule and each multicrystal form is unique.R ecognizing the phenomenon of disappearing polymorphs and learning to overcome and control it requires ac ombination of considerable skill on the part of the chemist with the acquisition of an intimate familiarity with and understanding of the crystal chemistry of the compound in question.

Outlook
It should be apparent from the content of this Review that the mere existence of polymorphs and polymorphic transformations is virtually impossible to predict, and that uncontrolled polymorphic conversions can have asevere impact on the development of molecular materials for potential APIs and speciality chemicals.Bearing in mind the advances made in understanding some of the vagaries associated with the solid state,i ti ss ometimes difficult to comprehend why and how new polymorphs still emerge (while others disappear) long after crystal-form screens presumably have been completed. Thepoint is that it can never be stated with certainty that the most stable form has been found;a tb est it can be determined which of the known forms is the most stable.A s the evidence above clearly shows,anew (and most often more-stable) form can appear at any stage in the history of ac ompound (or life-cycle of adrug).
Today we have access to highly sensitive analytical instruments and automated polymorph screening platforms. High-throughput salt and polymorph screens were "not on the radar" 30-40 years ago,and the whole ethos surrounding drug development was entirely different. In the experiences of one of the authors of this review (R.W.L.) polymorph Polymorphism Angewandte Chemie screens were actually carried out in the past by default, but were not designated as such. During the course of research and development of an API, al arge team of highly skilled chemists would fine-tune ap rocess from an initial milligram scale synthesis through scale-up,o nt ot he pilot plant, and ultimately into production. Within reason, time and resources were not the issue.I nt erms of the pharmaceutical industry, many drugs currently on the market and still highly profitable were developed and produced in this "classic" way.
What has changed?The key differences affecting product development these days are the compressed time scales for drug development (surely at odds with the notion of kinetics and crystallization?) and far fewer skilled process chemists available to develop APIs in the manner described above.The onus today is on efficiency and taking advantage of technical developments that have appeared in the last decade or so. Advances have been made based on the chemistry of the particular compound in question, not least in the development of automated polymorph screening platforms themselves,b ut also in the increasing sensitivity and precision of automated analytical instrumentation, and in situ analysis and algorithms for pattern recognition (e.g.c omparisons of diffraction patterns or Raman spectra).
There is no standard strategy or foolproof recipe for the search for crystal forms.Acombination of carefully designed manually performed crystallization experiments combined with automated high-throughput screens can reduce,b ut not totally eliminate the likelihood of unexpected polymorphic transformations if both highly pure and impure materials are used. All readily accessible "significant" byproducts obtained in the synthesis of at arget compound should be considered for use as seeds and additives.T here is an eed to screen for possible polymorphic transformations under stressed conditions (e.g.e xtreme humidity,t emperature,a nd pressure) to nurture confidence in the robustness of ap roduct once it leaves the realm of ac ontrolled laboratory space.F urthermore,the screens and solid-state studies should be considered throughout the lifetime of ad rug or speciality chemical to allow for the potential materialization of solid phases that take considerable amounts of time to nucleate. CSP [67,105,106] methods are now becoming progressively more capable of guiding the search for new crystal forms, [66,[121][122][123][124][125][126][127] and should be considered in solid-form screening processes,a long with knowledge-based hydrogen-bond propensity calculations, [128][129][130][131] whenever possible.S ince CSP was recently also successfully utilized to determine the crystal structure of as ub-micrometer-sized crystallite (present in picogram amounts) in ab ulk consisting of ad ifferent polymorph, [132] it is viable that CSP could aid the structural characterizations of small molecular impurities that potentially act as crystal seeds for the formation of unanticipated crystal forms.
Processes associated with drug and speciality chemical development have changed radically in the last decade or two. Skills associated with particle engineering,d rug design, and even relatively "unfashionable" skills such as those concerned with filtration and drying have advanced dramatically.All of these process modifications involve perturbations that can potentially lead to new solid phases.
Whilst advances in technologies associated with form screening,a nalytical instrumentation, and in silico approaches have come al ong way,w ea re often forced to return to the question of the level of better understanding of the fundamentals associated with crystal nucleation and growth. Will some (or many) of us still "get surprised" by the vagaries of polymorphism and crystallization in 20 years time-or will it be all sorted out by then?There are still many challenges and surprises in store.

Addendum
In the final section of this Review,wewill present excerpts from legal proceedings associated with the paroxetine hydrochloride case,w ith the aim of highlighting the complex relationship between science and the law.A lthough no scientific aspects of polymorphism are formally discussed, we describe some additional important aspects of the disappearing polymorph phenomenon.
As previously described, the Paxil (paroxetine hydrochloride) litigation (SmithKline Beecham versus Apotex) continued for nearly eight years before the US Supreme Court refused to hear the case.T he issues of seeding and inherency, both intimately connected with disappearing polymorphs were central to the case.Inanumber of instances, the positions of the witnesses (as on record from trial testimony) or of the judge (from his opinion in the first instance) were not always precisely quoted or correctly interpreted in secondary publications.I nt he interest of informing the reader of how these issues can be interpreted and misinterpreted as well as of setting the record straight, we present some of that testimony and the way it was subsequently interpreted. Please note:one of us (J.B.) was awitness at trial on behalf of SmithKline Beecham.
It will be recalled that Judge Posner found valid the patent for which the independent Claim 1i ss imply "Crystalline paroxetine hydrochloride hemihydrate". However,h ea lso found that Apotex would not be infringing that patent because they would not be marketing an anhydrate API with "high double-digit" percentages of the hemihydrate and would be gaining nothing from the quantities of the hemihydrate that he found would be in the Apotex product. In what follows we relate the events subsequent to Judge Posners ruling and how that ruling and the decisions from the Federal Circuit (the venue for patent appeals) were interpreted by the trade press and perhaps some of the community not familiar with the details.
With regard to seeding,t here was undisputed evidence that Branford Chemicals (the Apotex subsidiary that actually manufactured the API) started their research on the compound with the hemihydrate,sothat their facility was seeded with Form 1.J udge Posner found that there would be conversion of anhydrate into hemihydrate: "Some conversion from anhydrate to hemihydrate is likely to occur in aseeded facility in which the anhydrate is exposed to air;BCIs plant is seeded;and the anhydrate manufactured there is exposed to non-dehumidified air before it leaves the plant. The evidence is sufficient to support an inference that . .
BCI will be making at least tiny amounts of the hemihydrate if it is permitted to manufacture the anhydrate." And Judge Posner then related to the amount of hemihydrate and the ability to detect it: "In sum, Ia mn ot persuaded that Apotex will produce an anhydrate that has sufficient hemihydrate to be detectable by the methods in use in 1985." 1985 was the date of the application for the patent on the hemihydrate,b ut it is unclear why he limited his analysis to 1985, as patent law does not limit the method of detection for infringement purposes to those available at the time of the invention, especially in this case where the claim simply recites "Crystalline paroxetine hydrochloride hemihydrate". As discussed above,t he Federal Circuit ultimately reversed Judge Posner on his non-infringement determination, but ultimately held the claim invalid as inherently anticipated.
In the course of the trial, Judge Posner asked counsel for SmithKline Beecham whether as ingle crystal of the hemihydrate in anhydrate API would infringe the patent. The response was that in principle yes (in accord with the formal reading of the patent law), but that the SmithKline Beecham case was built on aconsiderable body of evidence that many of Apotexs batches had converted and that, therefore,itwas highly likely that subsequent batches would exhibit conversion-to an extent much greater than as ingle crystal and proven as detectable by methods available in 2003.
On appeal, the Federal Circuit agreed with SmithKline Beechams claim construction: "[N]othing in the 723 patent limits that structural compound to its commercial embodiments," and thus overturned Judge Posner on that issue.M oreover, the Federal Circuit found that Apotex would be infringing the Form 1 patent: "Thus,r eading Claim 1i nt he context of the intrinsic evidence,t he conclusion is inescapable that the claim encompasses,w ithout limitation, PHC [Paroxetine Hydrochloride] hemihydrate-a crystal form of paroxetine hydrochloride that contains one molecule of bound water for every two molecules of paroxetine hydrochloride in the crystal structure." In other words,i np rinciple,e ven as ingle crystal would infringe the patent. Thus,t he Federal Circuit agreed with SmithKline Beecham that Apotex would infringe the patent.
Upon appeal to the Federal Circuit, two of Apotexs arguments are relevant in the current context. First Apotex attacked the decision acknowledging conversion as clearly erroneous,and, second regarding seeding,they argued: "In sum, the district courts apparent fascination with the seeding theory led it to af inding that smacks of alchemy,n ot chemistry." At the time of the trial there was still amajor misconception about the meaning of a" disappearing polymorph", as evidenced by the courtroom exchange: Question: "Okay. And, Dr.B ernstein, under your theory that once hemihydrate was made,s eeds of the hemihydrate would contaminate any further paroxetine hydrochloride that was made,t hen anyone practicing the 196 Ferrosan patent in the United States would produce the hemihydrate after the HP23 and 24 batches [SmithKline Beechams first batches of Form 1]were sent to the United States,correct?" Answer: "No thats not correct. That doesnt represent my point of view.Asyou pointed out earlier,the last sentence in my paper says once you have prepared it, you ought to be able to prepare it again. So,Iam not saying that the anhydrate cant be prepared again. What Ihave said is that after the anhydrate is prepared, if there are-if there are seeds around and water,then it is highly likely that that will convert, and Ithink you have to make ac lear distinction between the preparation and the conversion process.T hose are two different processes.Inever said it couldnt be prepared." In April 2004, the Federal Circuit reversed Judge Posners ruling and found that "any amount of crystalline paroxetine hydrochloride hemihydrate without further limitation…will infringe Claim 1." However, it found Claim 1w as invalid under alegal argument that the clinical trials that SmithKline Beecham had carried out constituted a" public use." Thec oncurring appellate judge in that decision also expressed the view that Claim 1w as invalid because one crystal form converted into amore stable crystal form without human intervention, and was therefore a"naturally occurring process". However, the two other judges dismissed this view because the crystal compound was as ynthetic, man-made compound, and thus is a "composition of matter" that is eligible for patent protection.
This decision was appealed to the full court en banc (fifteen judges) in June 2004. Thef ull Court reversed the "public use" issue and remanded the case back to the original panel. In April 2005, the panel again found the claim invalid, but this time on the grounds of inherent anticipation.
"Because the record contains clear and convincing evidence that the production of PHC anhydrate in accordance with the 196 [original Ferrosan anhydrate] patent inherently results in at least trace amounts of PHC hemihydrate,this court holds that the 196 patent inherently anticipates Claim 1ofthe 723 patent…" It will be recalled that in US legal terms inherency means invariably or inevitably.Hence,toreach such aconclusion the Federal Circuit would have to be convinced that Form 1 existed before the SmithKline Beecham inventors said they first detected it late in 1984. To examine the evidence for "at least trace amounts of PHC hemihydrate"i ti sn ecessary to return to the trial transcript and Judge Posners examination of experts from both sides on this question.
Thef ollowing is the related transcript excerpt of the dialogue between Judge Posner and SmithKline Beechams expert: Question: "Ihave afew questions,Dr. Bernstein. Yousaid on Friday that there was no hemihydrate before December, 1984. In fact, you said today that you are absolutely convinced there was none.Idont understand that. Itake it you mean there was no detectable hemihydrate,dont you?" Answer: "Well, your Honor,from the history subsequent to December 84, when there were locations in whicht here was definitely hemihydrate and there was water available,there was almost, there was ah igh, very high probability of conversion. So the fact that there was never any evidence of the hemihydrate prior to 1984 and no evidence of conversion prior to December of 1984 indicates to me that it didnt exist, and its In fact, unchallenged data on accelerated stability tests of the anhydrate at 40 8 8Ca nd 75 %r elative humidity over aperiod of several months in 1982 had exhibited no evidence of any conversion into hemihydrate.E ven Judge Posner summarized in his decision: "First, ab atch of anhydrate manufactured by Ferrosan in 1980, though stored in ah ot and humid place (the greater the heat-short of the melting point, of course-and the humidity, the likelier is conversion from the anhydrous to the hemihydrous form), had three years later still not converted to the hemihydrate form, suggesting that it had not been seeded and hence that there were no seeds as late as 1980. And [pilot plant batch]H P22, manufactured just weeks before HP23, contained no detectable hemihydrate,w hereas HP23 was entirely hemihydrate." Judge Posner summarized the testimony of Terry Threlfall, Apotexs witness on this issue: "Dr.T erence Threlfall, Apotexs expert on polymorphism, testified to the contrary of Bernstein that anhydrous and hemihydrous forms of paroxetine can coexist happily.There is support for this conjecture in SmithKline Beechams own evidence,ofwhich more later,that some of Apotexs anhydrous product contains small amounts of hemihydrate without conversion of the rest. In other words, as Threlfall testified, amixture of anhydrate and hemihydrate can be an equilibrium, in which event the earliest batches of paroxetine manufactured by Ferrosan may have contained undetectable quantities of the hemihydrate.Inlight of this evidence,Dr. Bernsteins absolute certainty that hemihydrate did not exist before December 1984 is not tenable.N oo ne knowsw hen the hemihydrate form of paroxetine came into existence,a lthough it is ar easonable inference that it did not exist in adetectable amount until then." Theo pinion that the hemihydrate did not exist before December 1984 may be untenable in Judge Posners view,but there were absolutely no data or scientific evidence that it ever existed before.M oreover,i nl ight of the later observations about the tendencyf or conversion and the total lack thereof in the 1980 batch strengthened that conviction. With regard to the existence or non-existence of crystal forms,i t seems incontrovertible and contrary to the norms of scientific reasoning that one can not claim to have ac rystal form for which there is-and up to ac ertain date has never beenabsolutely no physical or chemical evidence for its existence. Especially in light of the results of the accelerated stability tests,t he fact that after the date of its appearance some batches did not convert cannot serve as evidence that it existed prior to its actual positive discovery late in 1984.
Furthermore,J udge Posner was discussing the situation after December 1984 in which it was likely that there were seeds of hemihydrate in locations where it have been prepared. Thel ack of conversion, even in the presence of seeds does not prove his point for the period prior to December 1984. Conversion depended on the amount of water (or humidity), temperature,a nd pressure.I ft he combination of those factors was not sufficient, conversion would not take place.Inthe 1980 accelerated stability tests on an umber of samples,they almost certainly would have been sufficient.
This Addendum and some of the legal examples cited in the text provide evidence for the complex relationship-"uneasy bedfellows" in one view-between science and the law,m ore specifically between the scientific method and scientific reasoning on the one hand and legal reasoning on the other hand. This dynamic relationship has been addressed in anumber of monographs [133][134][135][136][137][138] and will no doubt continue to generate debate from practitioners and scholars of jurisprudence and scientists.
Forthe scientific community,the standards have been well stated by Peter Huber in the closing paragraph of one of his treatises on the subject: "The best test we have of certainty is good science-the science of publication, replication, and verification, the science of consensus and peer review; the science of Newton, Galileo, and Gauss,Einstein, Feynman, Pasteur and Sabin;the science that has eradicated smallpox, polio,a nd tuberculosis;t he science that has created antibiotics and vaccines.Oritisatleast, the best test of certainly so far devised by the mind of man." [133] . .