The significance of QT interval in drug development

Introduction

The duration of QT interval of the surface electrocardiogram (ECG) reflects the ventricular action potential duration (APD) which is determined mainly by the rapid component of the outward repolarizing current (I_Kr). This current is mediated primarily by the delayed rectifying potassium channel. Thus, the QT interval is congenitally prolonged when this current is diminished as a result of genetic mutations of this channel as for example in the Romano–Ward syndrome [1]. Reduction in this current and hence, the prolongation of the QT interval may also be acquired, resulting from electrolyte imbalance (especially hypokalaemia and/or hypomagnesaemia), endocrine dysfunction (e.g. hypothyroidism), autonomic imbalance, various disease states or most frequently, following clinical administration of drugs.

Drug-induced prolongation of the QTc interval may be followed by potentially fatal proarrhythmias. More than any other adverse drug reaction in recent times, it has been responsible for the withdrawal of many drugs from the market and yet as a surrogate of proarrhythmias, it is not well understood. Regulatory decisions have resulted in rejection of some new drugs or the restriction on the clinical use of many old and other new drugs over the last decade because of their potential to prolong the QTc interval. Therefore, there are regulatory and clinical expectations of better preapproval characterization of new chemical entities (NCEs) for this potential risk which have had a very profound influence on drug development. This paper will focus on the issues that need to be addressed during drug development, strategies aimed at identifying this risk during early preclinical and clinical phases of drug development and the regulatory assessment of the potential risk, particularly the electrocardiographic data from the clinical trials.

Because the actually measured QT interval changes with heart rate in the absence of any intervention, it is usual to correct the measured interval for changes in heart rates (RR interval) to derive a rate-corrected (QTc) interval, which is then used when evaluating the effect of an intervention. Clinically, the rate-correction applied most widely, and almost exclusively for years, is the Bazett’s correction (QTc = QT/RR^0.50), which divides the measured QT interval by the square root of the preceding RR interval. A less frequently applied rate-correction is that of Fridericia (QTc = QT/RR^0.33) which divides the measured QT interval by the cube root of the preceding RR interval. Both these corrections standardize the measured QT interval to an RR interval of 1 s (heart rate of 60 beats min⁻¹). When corrected by Bazett’s formula, on historical and epidemiological grounds, the widely accepted upper limits of normal QTc interval are 450 ms in adult males, 470 ms in adult females and 460 ms in children between 1 and 15 years of age (regardless of gender). Unless stated otherwise, the QTc interval referred to in this paper is the interval as corrected by Bazett’s formula.

Drug-induced prolongation of QTc interval is expected with class III antiarrhythmic drugs which are intended to produce their desired therapeutic benefit by blocking I_Kr, delaying ventricular repolarization and, therefore, increasing myocardial refractory period. Typical examples of these drugs include sotalol, bretylium, ibutilide, dofetilide, azimilide, sematilide, ambasilide, almokalant, N-acetyl-procainamide, fenoxedil and terikalant.

Risks from QTc interval prolongation

Excessive QTc interval prolongation in the right setting (see risk modifying factors below) can be proarrhythmic and degenerate into a potentially fatal ventricular tachyarrhythmia known as torsade de pointes (TdP), a unique polymorphic form of ventricular tachycardia which (by definition) is associated with concomitant prolongation of QTc interval [2]. TdP is triggered by the appearance of early after-depolarizations (EADs), mediated by slow inward calcium current, during the late phase 2 of the prolonged action potential. Therefore, as an extension of their pharmacological effect, this iatrogenic proarrhythmia may be expected in some individuals following the use of antiarrhythmic drugs, which possess class III (potassium channel blocking) activity. For example, the incidence of TdP is variously estimated to be 0.5% to 8.8% with quinidine [3] and 2.6% to 4.1% with sotalol [4]. The incidence is higher in combination preparations of sotalol that include a thiazide diuretic, which induces hypokalaemia [5], and lower with racemic sotalol in contrast to (+)-(S)-sotalol because of the β-adrenoceptor blocking activity of (−)-(R)-sotalol present in the former. It is plainly evident that the balance between the therapeutic antiarrhythmic effect and the potentially fatal proarrhythmic effect of QTc interval prolongation is a very delicate one, depending not only on the drug concerned and its plasma concentration but also on a number of host modifying factors. These factors include female gender, electrolyte imbalance (especially hypokalaemia), myocardial ischaemia, atrial fibrillation, congestive heart failure, bradycardias with or without heart blocks and pre-existing prolongation of QTc interval [6], to name a few.

Clinical manifestations of TdP, which usually is a transient tachyarrhythmia, include palpitation and, when prolonged, the symptoms arising from impaired cerebral circulation such as dizziness, syncope and/or seizures. TdP subsequently degenerates into ventricular fibrillation in about 20% of cases [7] and, not uncommonly, cardiac arrest and sudden death may be the outcome [6]. The overall mortality is of the order of 10–17% [7, 8].

QTc interval and nonantiarrhythmic drugs

Unfortunately, the potential to prolong the QTc interval and induce TdP is not confined to class III antiarrhythmic drugs. A number of class I antiarrhythmic and antianginal drugs as well as noncardiovascular drugs also carry this liability. There are now about 10 antianginal and well over 90 noncardiac drugs (Table 1), which have been reported to prolong significantly the QTc interval and/or induce TdP. In addition, there are many other drugs which have been shown to block I_Kr current in vitro. In terms of their pharmacotherapeutic classes, the noncardiac drugs include H₁-antihistamines, antidepressants, neuroleptics, antimicrobials including fluoroquinolones and antimalarials, serotonin antagonists and anticancer drugs. There are also a host of miscellaneous drugs such as probucol, cisapride, sevoflurane, bupivacaine, tacrolimus, levacetylmethadol, tiapride, amiloride and lubelozole, which are torsadogenic. In a recent survey of 2194 cases of TdP in the US Food and Drug Administration (FDA) database [8], the most common drugs implicated were cardiac (26.2%), central nervous system (CNS) (21.9%), anti-infectives (19.0%) and antihistamines (11.6%). Of the 2194 cases, 61.1% were associated with hospitalization, 27.9% were life-threatening and 9.8% were associated with a fatal outcome. The proarrhythmia was associated with a serious underlying condition in 16.2%, with drug interactions in 11.7% and with an overdose in 9.2% of the cases.

Table 1.

Non-class III drugs reported to cause QT interval prolongation and/or torsade de pointes.

Non-class III cardiovascular drugs	Non-cardiovascular drugs
Sodium channel blockers	Neuroleptics	Antidepressants	Antimicrobials	Anticancer drugs
Quinidine	Chlorpromazine	Amitriptyline	Erythromycin	Anthracycline
Disopyramide	Triflupromazine	Protriptyline	Co-trimoxazole	Aclarubicin
N-acetyl-procainamide	Promazine	Nortriptyline	Sulfamethoxazole	5-fluorouracil
Lorcainide	Perphenazine	Butriptyline	Pentamidine	Acodazole
3-Methoxy-O-desmethyl-encainide (MODE)	Fluphenazine	Desipramine	Amantidine	Adriamycin
	Prochlorperazine	Imipramine	Grepafloxacin	Tamoxifen
Ajmaline	Trifluoperazine	Lofepramine	Levofloxacin	S9788
	Triethylperazine	Clomipramine	Moxifloxacin	502U83
Antianginals	Haloperidol	Doxepin	Sparfloxacin	Arsenic trioxide
Prenylamine	Trifluoperidol	Maprotiline	Gatifloxacin	Efavirenz
Fendiline	Droperidol	Dothiepin	Clarithromycin
Lidoflazine	Penfluridol	Citalopram	Spiramycin	Miscellaneous
Bepridil	Fluspirilene	Zimeldine	Fluconazole	Vincamine
Aprindine	Risperidone	Fluoxetine	D0870	Probucol
Terodiline	Ziprasidone		Antimoniates	Glibenclamide
Perhexiline	Amisulpride	H1-Antihistamines		Epoprostenol
Amiodarone	Chlorprothixene	Terfenadine	Serotonin (5-HT2)-antagonists	Chloral hydrate
Tedisamil	Thiothixene	Astemizole	Ketanserin	Amiloride
Mibefradil	Thioridazine	Diphenhydramine	Amperozide	Bromocriptine
	Sertindole	Promethazine	Retanserin	Sevoflurane
α1/β blockers	Pimozide	Hydroxyzine	Pipamperone	Cisapride
Sotalol	Zotepine			Tacrolimus
Oxprenolol	Quetiapine	Antimalarials	Serotonin (5-HT3)-antagonists	Levacetylmethadol
Nifenalol	Olanzapine	Halofantrine	Dolasetron	Lubelozole
Indoramin		Chloroquine	Zatosetron	Tiapride
Melperone		Arteether		Tizanidine
Amosulalol				Rivastigmine
				Cocaine
Inotropic agents				Domperidone
Dobutamine				Bupivacaine

Regulatory consequences

Between June 1990 and March 2001, 11 noncardiac drugs, marketed in the United Kingdom and elsewhere, attracted significant regulatory actions because of their propensity to produce QTc interval prolongation and/or TdP. Their prescribing information was changed substantially to revise dose schedules, contraindications and/or precautions during their clinical use. These drugs included two H₁-antihistamines (terfenadine and astemizole), one gastric prokinetic agent (cisapride), one agent for urinary incontinence (terodiline), two anti-infectives (halofantrine and grepafloxacin), four neuroleptics (pimozide, thioridazine, sertindole and droperidol) and one drug for opiate addiction (levacetylmethadol). Eight of these drugs have now been withdrawn from the market. A number of new noncardiovascular chemical entities (for example, gatifloxacin, moxifloxacin and ziprasidone) have been refused approval in one or more of the major markets because of their potential ‘QTc liability’. Therefore, as an adverse effect, the significance of investigating an NCE for its potential to prolong the QTc interval during drug development cannot be overemphasized.

Not surprisingly, the regulatory focus on QTc interval prolongation by drugs has changed from one of a potentially desirable antiarrhythmic mechanism to one of potentially fatal proclivity. Central to the regulatory concern are the facts that (i) the number of noncardiac drugs recognized to cause QTc interval prolongation continues to increase (ii) many of these drugs are prescribed for otherwise relatively benign or low risk conditions, often with safer alternatives (iii) for most of these drugs, historically, their potential to prolong the QTc interval and induce TdP was not recognized for many months or years after the drug was approved and in clinical use, and finally (iv) the population at risk is greater than had hitherto been appreciated.

The regulatory concerns are particularly heightened by the facts that despite the best endeavours to characterize the risk and provide adequate prescribing information to the prescribing physicians, once the effect on the duration of QTc interval was identified, there was almost total disregard of this prescribing information in terms of the concurrent use of contraindicated drugs or monitoring of patients by ECG as recommended [9–12].

Issues for drug development

Between them, these QT prolonging drugs illustrate not only the diverse therapeutic classes implicated but also the limitations of previously conducted clinical trials in detecting this potentially fatal cardiotoxic effect. They highlight the roles of drug interactions, metabolites, dose schedules, pharmacogenetic traits and stereoselective factors that should be explored thoroughly when developing NCEs.

The division of drugs by therapeutic class alone deserves a critical cautionary comment. Drugs are often discovered to have more potent activity at pharmacological targets other than those originally intended during their development. Drugs are therefore known to cross ‘therapeutic boundaries’. Many QTc prolonging drugs belong to a specific chemical class, usually associated with one therapeutic area but have later been developed or used clinically in an entirely different therapeutic area (Table 2). Terfenadine, for example, was discovered through a CNS programme aimed at developing a new neuroleptic agent but because of its more potent secondary pharmacological effects at the H₁-antihistamine receptor, its development was re-focused to introduce the first nonsedating H₁-antihistamine. Not surprisingly, as any other neuroleptic drug might have, it too attracted considerable attention because of its effect on QTc interval and its ability to induce TdP. Introduced to the market in 1982 for the treatment of hayfever, it was a highly successful and popular drug until withdrawn from the market, or its use severely restricted, due to reports of TdP primarily resulting from drug interactions or overdoses. Sildenafil, originally intended for development as an antianginal drug, was developed for male erectile dysfunction and it is not surprising that like many antianginals, it has recently been shown to prolong cardiac repolarization by blocking the rapid component of the delayed rectifier potassium current, albeit at concentrations (IC₅₀ = 100 µmol l⁻¹) well exceeding those encountered therapeutically (1 µmol l⁻¹) [13]. Even 30 µmol l⁻¹ induced only a 15% blockade of I_Kr. The margin is even greater if one takes into account only the free fraction of the drug since sildenafil is about 96% protein-bound. Therefore, clearly, a clinically significant effect on repolarization is most unlikely during the therapeutic use of sildenafil [14]. Although there have been no reports of QTc interval prolongation or TdP following the marketing of this intermittently used drug after its approval, sildenafil does illustrate the point being made on regulatory limitations of classification of drugs by therapeutic class.

Table 2.

Regulatory utility of classification of drugs by a ‘therapeutic class'.

Drug	Original therapeutic class	Clinically developed for use as
Imipramine	Antihistamine and sedative	Antidepressant
Chlorpromazine	Sedative	Neuroleptic
Terodiline	Antianginal	Urinary incontinence
Terfenadine	Neuroleptic	H1-antihistamine
Melperone	Neuroleptic	α-blocker
Sotalol	β-blocker	Class III antiarrhythmic
Nifenalol	β-blocker	Class III antiarrhythmic
Sildenafil	Antianginal	Erectile dysfunction
Amiodarone	Antianginal	Antiarrhythmic
Butyrophenones	Analgesic	Neuroleptic
Iproniazid	Anti-tuberculous	Antidepressant (MAOI)
Ethmozine	Neuroleptic	Class I antiarrhythmic

Limitations of clinical trials

None of the older drugs had declared its potential to prolong the QT interval during the types of clinical trials conducted in the past. Of particular regulatory concern has been the interval from first approval of almost all these drugs to the first identification of their proarrhythmic or QTc prolonging potential. Apart from halofantrine, this has ranged from 2 to 3 years for astemizole to as much as 17 years for pimozide. The proarrhythmic potential of halofantrine was strongly signalled during its clinical trials. Among the drugs developed recently, both sertindole and levacetylmethadol were shown to prolong QTc interval during clinical trials. However, a great reliance is being placed on the ability of clinical trials – both in healthy volunteers given high single doses or multiple doses over very short periods during Phase I or in patients receiving ‘normal’ doses during Phase II/III – to uncover this risk. The crucial question from a regulatory perspective is how efficient and reliable these clinical trials are in achieving this objective, given the patient population enrolled, background noise arising from spontaneous intraindividual variability in QTc interval and the relatively low frequency of the clinically significant drug-induced cardiac effect [15].

The present efficacy-orientated approach is primarily responsible for failure of clinical trials to detect the risks of TdP. The numbers of patients exposed are not large enough nor are all the patient subgroups likely to receive the drug during its uncontrolled clinical use (and in fact at a much greater risk) represented in these preapproval clinical trials. These include those with predisposing diseases or those receiving drugs with a potential for pharmacokinetic or pharmacodynamic interactions. Thus, the scope for detecting drug-drug or drug–disease interactions in clinical trials is also very limited. And yet, the experience has shown that these are among the most important risk factors! Equally importantly, it is now recognized that the risk can vary from day to day depending on intercurrent event or intervention.

The frequency of TdP or prolongation of the QTc interval to a proarrhythmic threshold (500 ms) varies with the class of drugs. Not unexpectedly, it is the highest with class III antiarrhythmic drugs. For noncardiac drugs, the frequency is unknown and can vary from approximately 1 in 100 (e.g. halofantrine) to 1 in 50000 (e.g. terfenadine), depending on clinical circumstances. Overall, however, the frequency of this effect with noncardiac drugs is difficult to estimate. This is hardly surprising since the diagnosis of this particular toxicity requires an ECG monitoring facility which is either not available in general practice or when available in a local hospital, not utilized appropriately. Since TdP can be transient, its diagnosis in a patient presenting with symptoms suggestive of TdP, such as dizziness or syncope, requires immediate access to a cardiac rhythm recording facility. Even in asymptomatic patients, despite the requirements included in prescribing information, there is a general lack of appropriate patient monitoring by ECG. More importantly, however, the effect is often not recognized as iatrogenic and it is grossly under-reported (reporting rate is of the order of 10–20%) even when recognized as drug-induced. This was well exemplified by the events preceding the withdrawal of terodiline [16]. In all likelihood, the frequency of TdP or QTc interval >500 ms is relatively low (<0.1%) and below that which can be confidently detected by the size of the clinical trials database that is usually included in the regulatory submissions. The frequency is sufficiently low that the risk has usually been uncovered hitherto only through spontaneous reports during the postmarketing use of the drugs concerned. Although low, it is nonetheless unacceptable given the nature of the disease under treatment in many cases and the potential for a fatal outcome – that is, an adverse risk/benefit ratio. Halofantrine and arsenic trioxide best illustrate the careful need to balance the potential risk against the potential benefits.

Trials conducted during the clinical development of a drug typically include 1500–3000 highly selected patients showing relatively little pharmacokinetic or pharmacodynamic variability. These are unlikely to identify the potential of a drug to induce TdP. A database of 1500 patients will barely detect an event which occurs at the rate of 1 in 1000, and almost certainly would miss one that occurs with a frequency of 1 in 5000 or less (α error of 0.05 and β error of 0.05). Long-term safety studies do not include adequate ECG monitoring at peak plasma concentrations of the drug or its metabolites.

If the background incidence of an adverse event (e.g. TdP or QTc interval prolongation to proarrhythmic levels) to be detected is of the order of 1 in 1000 and the incidence of the same event to be detected, when drug-induced, is 1 in 1000, the number of patients required in the safety database would have to approach approximately 20000. Clearly, it is impractical, and indeed undesirable, in the case of a highly novel or effective medicine, to have to complete such a large clinical trial programme (in terms of number of patients and/or the duration of exposure) before an application for a marketing authorization is filed. Thus, to assess the clinical risk of proarrhythmias, the regulatory agencies must rely upon the surrogate marker, namely QTc prolongation provided it is adequately investigated and quantified during drug development. The principles of development include testing a drug at more than its intended therapeutic dose to define a potential dose–response relationship and to also do so in the presence of its metabolic inhibitors, another method of stressing the system to define any potential effects of the drug on cardiac repolarization.

QTc interval and drug interactions

A number of drugs such as terfenadine, astemizole, pimozide, cisapride and levacetylmethodol have had their potential to prolong the QTc interval and induce TdP uncovered as a result of drug–drug interactions [17–24]. These five drugs (and remarkably, many others with a potential to prolong the QTc interval) are metabolized by CYP3A4. The activity of this enzyme is also highly susceptible to liver disease. Not surprisingly, TdP in association with the clinical use of these drugs has been observed most frequently following their concurrent use with inhibitors of CYP3A4 such as azole antifungals and macrolide antibiotics [25, 26]. Other risk factors are liver disease (e.g. with terfenadine) and diabetes (e.g. with cisapride).

QTc interval and drug metabolites

Pharmacological studies of terfenadine, astemizole and cisapride best illustrate the role of metabolites. For terfenadine, its carboxylic acid metabolite, fexofenadine, is active as an antihistamine but is devoid of the I_Kr blocking property inherrent in the parent compound [27, 28]. Fexofenadine has now been developed for clinical use and is already on the market. This contrasts with the two metabolites of astemizole, namely desmethylastemizole and norastemizole. All of the three astemizole-related moieties are active as antihistamines, but the potential to prolong the QTc interval is predominantly present in astemizole and desmethylastemizole [29]. Indeed, the latter metabolite is slightly more potent in terms of cardiotoxicity and has a much longer elimination half-life compared with astemizole. At steady state following therapeutic doses, the plasma concentration of this major metabolite can be 30-fold higher than that of astemizole. In one patient with astemizole-induced TdP, concentrations were 7.7–17.3 ng ml⁻¹ for desmethylastemizole and <0.5 ng ml⁻¹ for astemizole [30]. Arguably, the proarrhythmic activity of astemizole during its clinical use probably derives largely from the presence of this metabolite in circulation. Norcisapride, the main metabolite of cisapride, has been shown to have a much lower proarrhythmic potential, if at all, than the parent drug [31]. Thus, as with fexofenadine, norastemizole is currently being developed clinically to replace astemizole and similarly, norcisapride to replace cisapride.

QTc interval and dose schedules

Since prolongation of QTc interval is a concentration-dependent type A adverse reaction, its frequency can be greatly reduced by an appropriate dosing regimen of the drug concerned. Overdose with astemizole or terfenadine is often associated with cardiac arrhythmias [32, 33]. Astemizole was originally approved at a 10 mg daily dose, but it has a long half-life, requiring many days before steady state is achieved. Given that desmethylastemizole with its cardiotoxic potential has a much longer half-life than astemizole, the perils of recommending a loading dose of astemizole soon became evident. A recommendation to administer astemizole at a 30 mg daily loading dose for 1 week followed by 10 mg daily had to be re-revised to remove the loading dose recommendation following reports of cardiac arrhythmias [34]. Pimozide is another drug which has a half-life of approximately 55 hours in most individuals. This is hightly variable, being as long as 150 hours in some patients even in the absence of any inhibitors of its metabolism. It was introduced originally at a starting dose of 2–4 mg daily with a slow upward titration to a maximum daily dose of 10 mg. Subsequently, the starting dose was increased to 20 mg daily, the slow titration schedule was removed and the maximum daily dose was increased to 60 mg. Trials investigating the use of pimozide in schizophrenia in the USA had to be suspended in 1981 following the sudden deaths of two patients during acute titration of pimozide to 70–80 mg daily doses [35]. Following reports of QTc interval prolongation and TdP, the dosing schedule was re-amended to recommend an initial starting dose of 2 mg daily with a very shallow dose titration to a maximum daily dose of 16–20 mg. In the USA, pimozide is not approved for use in schizophrenia.

QTc interval and stereoselectivity

Stereoselectivity in the pharmacological activity of a number of drugs acting at cardiac pharmacological targets is well known. For example, usually only one of the enantiomers of β-adrenoceptor blockers and dihydropyridine calcium channel blockers is pharmacologically active – either exclusively or dominantly. Stereoselectivity in activity at potassium channels has also been described for the enantiomers of some drugs. Examples include (+)-(R)-bupivacaine [36, 37] and (+)-(R)-halofantrine [38, 39]. The proarrhythmic activity of terodiline has been shown to reside in (+)-(R)-terodiline [40]. This is not surprising since it is structurally closely related to prenylamine whose proarrhythmic activity resides in its (+)-(S)-isomer [41]. Indeed, terodiline was marketed as an antianginal drug as long ago as 1965 before it was re-developed in mid-1980s for urinary incontinence following the observation of frequent and severe urinary retention associated with its cardiovascular use [16]. Although there is no information on its other isomers, (−)-(4S,6S)-acetylmethadol (levoacetylmethadol) has now been reported to be highly torsadogenic [42].

There is now an increasing trend to ‘chiral switches’ –the development of single enantiomers of previously marketed racemic drugs. This strategy is not without unforeseen potential risks. In view of its shorter half-life, the clinical availability of (R)-fluoxetine might represent a great advantage for clinical use of the drug in individuals in whom a greater dosing flexibility is required, e.g. in the elderly. This enantiomer may also be less prone to inhibit CYP2D6 compared with (S)-fluoxetine, which may also prove to be an advantage in the elderly who are likely to be receiving other CYP2D6 substrates. However, results from early clinical trials aimed at developing (R)-fluoxetine for clinical use suggest that the risk/benefit ratio of this enantiomer may warrant careful re-evaluation. Its use in about 2000 patients raised concerns over its potential to prolong the QTc interval at the highest dose administered. Indeed, this unexpected finding led to the termination of clinical development of this enantiomer [43].

QTc interval and pharmacogenetic factors

In the context of QTc interval prolongation by drugs, pharmacogenetic influences can be significant at both pharmacokinetic and pharmacodynamic levels. At a pharmacokinetic level, the metabolism of a number of QTc prolonging drugs, especially the neuroleptics, antidepressants and cardiovascular agents, is predominantly under the control of CYP2D6, a major drug metabolizing enzyme that is polymorphically expressed in the population. These include sertindole [44], thioridazine [45], risperidone [46], indoramin [47], nortriptyline [48] and terikalant [49]. The QTc interval prolongation following the administration of terikalant has been shown to correlate with CYP2D6 metabolic capacity [49]. In addition, it appears that metabolism of terodiline and prenylamine may also be controlled by CYP2D6 [16]. The significance of this polymorphic metabolism lies in the fact that extensive metabolizers can be converted into impaired or poor metabolizers by the presence of liver disease or during concurrent administration of drugs that inhibit drug metabolizing enzymes (most frequently and vividly observed with CYP3A4). Arising from its CYP2D6-mediated metabolism, thioridazine is now contraindicated in patients known to have decreased levels of CYP2D6. At a pharmacodynamic level, mutations of potassium channels, resulting in diminished repolarization reserve, are common [1, 50] and these result in an increased susceptibility of the patient to proarrhythmias. Although congenital prolongation of QT interval was thought at one time to be a diagnostic requirement for the presence of these mutations, there is now incontestable evidence that these mutations may be clinically silent, and many of the affected individuals have a normal ECG phenotype [51, 52] but are nevertheless at an increased torsadogenic risk. Female gender is also at a greater risk [53] which is further heightened during menstrual flow [54].

QTc interval and pharmacodynamic susceptibility

There is a wider appreciation now of other nongenetic clinical conditions with increased pharmacodynamic susceptibility to proarrhythmic effect. QTc interval prolongation (and possibly increased QT dispersion) are associated with, and have been identified as risk factor(s) for malignant ventricular tachyarrhythmias in a variety of diseases. These include sudden deaths (usually labelled as sudden unexplained cardiac deaths) and a number of cardiovascular as well as noncardiovascular ‘natural’ diseases – for example, cardiomyopathy [55–57], cardiac failure [4, 58], myocardial infarction [59], sudden infant death syndrome [60], diabetic autonomic neuropathy [61–63], hypoglycaemia [64], cirrhosis [65] and a number of other conditions associated with autonomic failure [66, 67]. Cardiac failure is typically associated with down-regulation of potassium channels [68] and the concurrent presence of this cardiac disease obscures, or is often used to reject, the iatrogenic origin of proarrhythmias in patients receiving QT prolonging drugs. It is interesting to note that despite urinary incontinence, 27 of the 69 patients who experienced terodiline-induced proarrhythmias were receiving diuretics and 33 were in receipt of other cardioactive medications [16]. Clearly, the population at risk is much larger than had hitherto been recognized (only patients with cardiac disease). Factors contributing to this are increased longevity, frequent polypharmacy with risks of interactions and comorbidity. It is interesting to note that of the 2194 cases in the FDA database [8], 92.8% were reported between 1989 and 1998 in contrast to only 7.2% between 1969 and 1988. Increased publicity and clinical awareness of the problem cannot by themselves account for this dramatic rise in reporting.

Cisapride represents an example of a unique drug with a target population that may already be at a greater risk of developing QTc interval prolongation and TdP. This risk was compounded by heavy use of contraindicated drugs despite regulatory warnings [11]. Cisapride is a gastric prokinetic drug approved in the UK in 1988 and in the USA in 1993. The maximum daily dose was 20–40 mg in the UK and 80 mg in the USA, in divided doses. Cisapride was indicated for the relief of symptoms of impaired gastric motility secondary to disturbed and delayed gastric emptying associated with diabetes, systemic sclerosis and autonomic neuropathy. As at 31 December 1999, the FDA database had included 341 reports of arrhythmias (of which 80 had a fatal outcome) associated with cisapride. A further 23 deaths were reported in the first 3 months of 2000, and in March 2000 severe limitation in its use in the US was announced. It is now available only for specific clinical eligibility criteria and for a limited-access protocol. The routine clinical use of cisapride in other markets was also discontinued or severely restricted. In view of the pharmacodynamic susceptibility of patients with (diabetic) autonomic neuropathy, it is not surprising that among the 159 cases of QTc prolongation or TdP associated with cisapride reported to the Uppsala Pharmacovigilance Centre as of 1999, about half reported no interacting medication, and by May 2000 six of the 20 deaths associated with cisapride occurred at doses of 40 mg or less in absence of known interactions [69]. Of course, the possibility of concurrent presence of silent mutations of potassium channels, having contributed either exclusively or partly to the risk, in these patients cannot be ruled out since diabetes itself is known to be associated with mutations of potassium channels [70].

Addressing QTc interval during drug development

The ‘unexpected’ association of this potentially fatal cardiotoxic effect with drugs of such diverse pharmacotherapeutic classes as antianginals and H₁-antihistamines prompts one to question what strategies can be employed during the preclinical and clinical development of new drug candidates to alert the manufacturers, the regulators and the prescribers to this risk. Given the serious public health consequences of this concentration-related adverse reaction and that it can cost anything of the order of £400 million to bring an NCE to the market, the need to identify this risk at a much earlier stage of the drug development is obvious. Indeed there are opportunities at almost every stage of the preclinical and clinical development of an NCE to investigate it for its potential to prolong the QTc interval and/or induce TdP.

In view of the many high profile drugs that had attracted considerable regulatory attention during the period from 1990–96 as a result of their potential to prolong the QTc interval and induce TdP, the Committee for Proprietary Medicinal Products (CPMP) of the European Union (EU) adopted two significant documents in December 1997. One of these was the CPMP document ‘Points to Consider: The Assessment of the Potential for QT Interval Prolongation by Non-cardiovascular Medicinal Products’ [71]. The recommendations contained within this document are not mandatory but they do represent a strategy by which EU regulators would like to see an NCE investigated for its potential to induce proarrhythmic prolongation of the QTc interval. The other was the CPMP ‘Note for Guidance on the Investigation of Drug Interactions’ [72]. Regulatory expectations summarized in these two documents have profoundly influenced drug development and have galvanized much research activity.

The clinical and public health concerns for the potential of noncardiac drugs to cause QTc interval prolongation and potentially fatal TdP have been eloquently summarized in an editorial [73]. Concerns have legitimately been expressed that:

‘Almost every week a new agent is added to the list of drugs associated with acquired long QT syndrome (LQTS) and torsades de pointes (TdP). Despite this impressive number of reports, the awareness of this subject is still limited among medical professionals and … It is likely that prevention of drug-induced TdP will never be fully successful, because it is a moving target. A patient may not be at risk when therapy is initiated, and may become at risk 5 days later because … It is intuitive that when two or more agents sharing potassium-channel-blocking activity are simultaneously administered, the risk of excessive prolongation of repolarization is substantially increased … The exclusion of potassium-channel-blocking properties might be considered in the future as a requirement before new molecules are approved for marketing, and more strict warnings in the package insert of drugs with known repolarization prolonging activity could be enforced.’

QTc interval and preclinical studies

In vivo preclinical studies have not always been predictive of the risk. In part, this may be because of the nonstandardized and technically challenging approaches to cardiovascular safety studies [74, 75]. However there are also the regulatory issues arising from the difficulties of recording high quality ECG traces from mobile animals, while the use of anaesthetics to restrain them introduces a confounding factor [76]. Furthermore, for ethical reasons, the number of studies that can be performed using live animals is limited, thus leaving a number of issues unresolved [77].

In contrast, in vitro electrophysiological studies offer many advantages. All the known noncardiac drugs that prolong the QTc interval and induce TdP have been shown to do so by blockade of the delayed rectifier potassium channel. The two hallmarks of a drug that truly prolongs the QTc interval are its concentration-related effect and the negative use-dependency of this effect.

At a molecular level, the channel for I_Kr current is expressed by proteins encoded by the HERG gene on chromosome 7 [78–80]. The α-subunit forming the HERG K⁺ channel carries the I_Kr current. Using a number of unicellular cardiac preparations (e.g. Purkinje fibres, papillary muscles or preferably, the ventricular myocytes known as M cells found in the mid-myocardial layer) from appropriate animals or expressed cloned channels (e.g. HERG), it is now possible to characterize the effect of drugs not only on the cardiac action potential but also to identify the ion channels likely to be affected by the drug [81, 82]. These systems also provide an opportunity of investigating risk factors for TdP that cannot adequately be studied in vivo in humans. In vitro electrophysiological studies make it possible to investigate not only the negative use-dependency (simulation of bradycardia), the effect of combinations of drugs and the roles of individual enantiomers and drug metabolites but also other risk factors such as hypokalaemia, myocardial ischaemia and concurrent administration of other drugs known to prolong the QT interval. In these investigations, comparison should be made to marketed compounds known to affect the APD and QTc interval, using (where possible) compounds having structural similarities to the new drug and/or having similar therapeutic indications. Since prolongation of the QTc interval is a function of the blockade principally of potassium channels, a preclinical investigation of the effect of the drug on this channel should provide valuable information.

Recently, Gintant et al. [83] determined whether the arrhythmogenic potential of noncardiac drugs can be assessed in vitro by investigating the effects of 12 drugs on the APD of cardiac Purkinje fibres, and compared results with clinical observations. APD changes in canine and porcine fibres were evaluated under physiological conditions. Six of the seven drugs associated with QTc prolongation or TdP in man (cisapride, erythromycin, grepafloxacin, moxifloxacin, sertindole and sotalol) produced concentration-dependent prolongation of the APD in canine fibres during slow stimulation (0.5 Hz), attaining >15% prolongation at high concentrations (≥10-fold clinically encountered plasma concentrations). Each of the five drugs not linked clinically to QTc prolongation and TdP (azithromycin, enalaprilat, fluoxetine, indomethacin and pinacidil) failed to attain 15% prolongation, with fluoxetine, indomethacin and pinacidil actually shortening the APD. Drugs eliciting the greatest prolongation also demonstrated prominent reverse rate-dependent effects. On the basis of concentration-dependent APD prolongation and reverse rate-dependent effects, this Purkinje fibre model detected six of seven drugs linked clinically to acquired long QTc syndrome and TdP, and cleared each of five drugs not associated with repolarization abnormalities (overall 92% accuracy). Terfenadine, linked clinically to dose-dependent QTc prolongation and TdP, only minimally prolonged the APD in canine and porcine fibres and exerted no effect on mid-myocardial fibres from the left ventricular free wall at supratherapeutic concentrations.

While prolongation of APD and appearance of EADs in unicellular preparations and blockade of HERG current constitute valuable indicators of potential clinical risk, it is important to bear in mind that in extrapolating these in vitro data to the clinical setting in vivo, three vital parameters should not be overlooked. These are the lipophilicity of the drug (or its cardiotoxic metabolite), its distribution ratio between plasma and the myocardial tissue and any other ancillary pharmacological activities of the compound (e.g. sodium or calcium channel or α- or β-adrenoceptor blocking activities) that may further modify the clinical risk. With this information, the clinical development programme can be modified to characterize the magnitude of risk. An important parameter in assessing the risk during routine clinical use of the drug might be the ratio of the concentration producing a 50% block at I_Kr compared with the concentration required for modulating the receptor targeted for efficacy. This would help to identify the risk for many drugs that block I_Kr at only high concentrations.

QTc interval and clinical studies

One clinical trial design that merits serious consideration is that used by the sponsors of ziprasidone. This study, referred to as study 054, was requested by the US FDA in order to compare the potential of ziprasadone for prolongation of the QTc interval relative to that of five other neuroleptics following their oral administration. This was an open-label parallel groups study in patients with schizophrenia. The sample size was based on within and between-patient variance from historical data derived from healthy male patients treated with placebo. Twenty-five patients per group allowed a 95% confidence interval to be constructed having a width of ±7 ms. There were five treatment periods; (1) tapering of existing medications, (2) washout and baseline period, (3) dose escalation without a metabolic inhibitor and dosed to steady state, (4) addition of an inhibitor and (5) exit and follow-up period. Three ECGs per day were taken during each of the last 3 days of periods 2, 3 and 4. ECGs were also monitored during dose escalation in period 3. ECGs and blood samples for pharmacokinetic analysis were obtained at baseline, during dose escalation and at steady state, in the absence and in the presence of a metabolic inhibitor. The times of ECG measurements corresponded to the mean t_max±30–60 min for each study drug. ECGs were read manually and QT interval was corrected by Bazett's and Fridericia's formulae and also by a study-specific formula (QTc = QT/RR^0.35 referred to as ‘baseline’ formula). Interval durations were defined precisely using a manual system that employed a digipad system and highly trained analysts. It is important to appreciate that the sample size used in this study did not have the power to provide estimates of the frequency of outliers with various categorical responses (that is, QTc interval >500 ms or ΔQTc >60 ms) to the drugs used.

In the context of clinical trials, it is important to randomize a population most representative of the target population in terms of comorbidity and to investigate interactions with comedications likely to be used by this population. The roles of pharmacogenetic traits and stereoselective metabolism should be addressed carefully. Such traits include mutations of drug metabolizing enzymes, which significantly modify concentrations of drug/metabolites or of potassium channels that decrease the basal repolarization reserve. The ECG should be recorded at peak plasma concentration of the parent drug and/or its active metabolite(s). If there is evidence of, or any reason to suspect, significant prolongation of the QT interval, increasing the size of the clinical safety database may be advisable [84] in order to fully ascertain the risk of TdP. QT prolongation by an NCE should be suspected, for example, when it is associated with proarrhythmic markers such as a particular core chemical structure, relevant pharmacodynamic properties or data from animal studies, if it belongs to a particular pharmacological class or unexpected alerts/signals emerge during clinical trials.

Presenting clinical data on QTc interval

From a regulatory perspective, a number of factors make it difficult to evaluate the proarrhythmic significance of mean increases in QTc interval observed following administration of a new drug. Apart from the potency of the drug in prolonging cardiac repolarization relative to its potency in modulating the primary pharmacological target, the proarrhythmic risk from prolongation of the QT interval is modified by other ancillary properties of the drug. Differences in the design of studies investigating the effect of a drug on ECG make it especially difficult to evaluate only mean data on QTc interval. These differences include the dose used, duration of the study, sample size and types of subjects enrolled. Difficulties in evaluation are further compounded by diverse practices in reporting the results.

Regarding the presentation of data from clinical trials, historically the only data presented were the mean baseline values and mean changes from baseline. While some studies have reported mean change from baseline to maximum QTc interval (‘peak effect’), others have reported mean change in QTc interval averaged across the dosing interval. Since drug metabolites may also mediate blockade of potassium channels, the peak effect of interest is the maximum effect on QTc interval and not necessarily the effect at peak concentration of the parent drug. For terfenadine 60 mg dose, the mean peak increase is 18 ms while the average mean increase over the dosing interval is 6 ms! Yet others have reported mean change from baseline to end of study, regardless of the peak plasma drug concentrations or effect.

Review of the data on some of the most torsadogenic drugs is highly instructive and revealing. Mean increases in peak QTc interval were 9 and 22 ms following single oral doses of 10 mg and 50 mg thioridazine, respectively [85] and 23, 19 and 0 ms following single oral doses of 200 mg racemic terodiline, 100 mg R-terodiline and 100 mg S-terodiline, respectively [40]. Single oral doses of moxifloxacin increased the mean peak QTc interval by 15 ms at 400 mg dose and by 17 ms at 800 mg, both relative to placebo [86]. This compares with sparfloxacin-induced increases in the mean peak QTc interval of 15 ms following a single oral dose of 200 mg and of 14 ms following a single oral dose of 400 mg, both relative to placebo [87]. In one study, single oral doses of 6 mg pimozide increased mean peak QTc interval by 13.3 ms [88], while another study reported that mean (±s.d.) daily doses of 10.68 (±7.22) mg pimozide for 9 weeks increased the mean QTc interval by 24 ms, there being no relationship to dose or age of the patients [89]. Mean peak increases of 10, 16, 29, 51 and 60 ms in QTc interval were reported following single oral doses of 200 mg, 400 mg, 800 mg, 1200 mg and 1600 mg of sparfloxacin to healthy volunteers [90, 91]. These compare with a steady state increase in mean QTc interval of 11 ms following 200 mg sparfloxacin in 813 Phase III patients [92]. Cisapride 20 mg twice a day for 7 days increased the mean peak QTc interval by 6.8 ms at 1.5 h and 10.9 ms at 3 h after dose [AstraZeneca, personal communication]. The mean increase in QTc interval (peak or average not specified) was 6 ms following administration of 10 mg astemizole for 2 weeks. The mean increase in peak QTc interval at steady state was reported to be 21 ms following 12–24 mg and 31 ms following the highest dose of 24 mg of sertindole. The difficulties in interpreting such heterogeneous data on mean changes from baseline when comparing or evaluating drugs are immediately apparent.

A more standard approach to trial design to include positive and negative controls and to allow for high plasma drug concentrations in some individuals has much to recommend itself. For example, at the FDA Psychopharmacological Advisory Committee meeting on 19 July 2000, the following data were made available for public disclosure by the sponsor of a large comparative study of the effect of 6 antipsychotic agents on QTc interval (study 054 referred to earlier). Mean increases in peak QTc interval (%) reported at steady state were as follows: 20.3 ms (5.2%) after 160 mg ziprasidone, 11.6 ms (2.9%) after 16 mg risperidone, 6.8 ms (1.8%) after 20 mg olanzapine, 14.5 ms (3.7%) after 750 mg quetiapine, 35.6 ms (9.1%) after 300 mg thioridazine and 4.7 ms (1.2%) after 15 mg haloperidol. Data from such studies might give a more reliable comparison of the QT prolonging potencies of different drugs at their therapeutic doses. Based on these and other data on nontorsadogenic drugs, the likely prognostic significance of the placebo-corrected mean peak effects on QTc interval, computed by the author, is shown in Table 3.

Table 3.

Likely prognostic significance of 1x clinical dose mean maximum or peak placebo-corrected effects on QTc interval.

Mean maximum or peak placebo-corrected increase in QTc interval	Likely potential torsadogenic risk
≤5 ms	None
6–10 ms	Unlikely
11–15 ms	Possible
16–20 ms	Probable
21–25 ms	Almost definite
≥26 ms	Definite

While mean changes in peak effect from baseline may provide a helpful signal, it is the outliers with categorical responses (QTc interval >500 ms and/or ΔQTc >60 ms) [71] who provide the most valuable information on the potential of a drug to prolong the QT interval and induce TdP. Apparently small mean changes may easily conceal large changes in individuals of specific regulatory interest. Small, apparently insignificant, increases in the QT interval also occur in many patients taking an offending drug but in only a few susceptible patients (generally the ones excluded from clinical trials) are these changes marked enough to lead to induction of ventricular tachycardias. Although the mean change in QTc interval produced by sparfloxacin in 813 patients amounted to 11 ms (+2.9%), it had exceeded 500 ms in 10 of these patients [92]. Evaluation is greatly facilitated by this ‘outliers’ analysis, which the regulatory authorities expect to be provided. However, consideration of changes in individual subjects raises another dilemma – given that there are spontaneous diurnal changes in QT interval [93], how can one be certain that an observed effect is drug-induced?

Isolated prolongation of the QT interval may be a common finding in early dose-escalation studies. QT interval may be influenced by a number of factors and shows spontaneous diurnal variation. A relationship to the drug is likely if similar changes are not observed in the placebo group and increases in the QT interval occur frequently, are dose-related and have a time course consistent with drug effect. Accurate centralized QTc measurements that eliminate site-to-site variability in techniques and analysis are critical if a signal is to be detected.

In a double-blind, four-period crossover, dose escalation study, which involved 28 normal healthy volunteers and 28 patients with stable cardiovascular disease [94], the mean QTc interval at baseline was 407 ms in normal subjects and 417 ms in patients with cardiovascular disease (P<0.01). The largest increase in mean QTc on terfenadine was 24 ms in a normal subject and 28 ms in a patient with cardiovascular disease. The longest average QTc observed was 449 ms and 501 ms in any normal subject and patient with cardiovascular disease, respectively. Compared with baseline, terfenadine 60 mg twice daily was associated with a mean QTc increase of 6 ms in normal subjects and a 12 ms increase in patients with cardiovascular disease (P<0.01 vs baseline; P>0.05 when the two populations were compared). Although the QTc increase from baseline was statistically significant, the magnitude of the spontaneous variability in QTc in the same patients was found to be much greater. Because 40 ECGs were obtained while taking placebo in each participant, the investigators also determined spontaneous variability in QTc interval with placebo. From the observed placebo variability, it was calculated that an increase in (Bazett-corrected) QTc of 35 ms or more while receiving drug therapy is likely to represent a drug effect at the 95% confidence level. It was also calculated that the probability of a 50 ms increase being of chance origin was 0.0003 over 1 day and 0.002 over 6 days. Since there are no comparable data on Fridericia-corrected QTc interval, one can only speculate on the magnitude of change in the Fridericia-corrected QTc interval which is likely to represent a drug effect. These data were dependent on QT measurements being made in a precise manner to three decimal places using a manual digipad technique. Data derived from automated computer generated measurements are generally acknowledged to be unreliable.

Makkar et al. [53] have reported on 314 cases of TdP associated with cardiovascular drugs; in 154 of the patients in whom the information was available, the mean increase in QTc interval from baseline was 130 ms in 54 males and 160 ms in 100 females. The mean QTc interval while receiving the offending drug was >510 ms in 90% of the cases. The utility of the corrected QT interval is emphasized by the fact that in contrast, only 80% of these patients had an absolute uncorrected QT interval greater than 510 ms. In one analysis of 21 cases of sotalol-induced TdP, the mean baseline QTc interval was 440 ms and this had increased to 540 ms (by +23%) at the time of TdP [95]. Analysis of QT and QTc intervals of patients who developed TdP on cardiac drugs such as prenylamine or bepridil and noncardiac drugs such as terfenadine, cisapride and terodiline support the same conclusions. Of the 85 patients who had developed TdP on terodiline (n = 25), cisapride (n = 43) and terfenadine (n = 17) and in whom this information was available, 76 (89.4%) had QT/QTc interval ≥500 ms prior to the onset of tachyarrhythmia.

The border between antiarrhythmic and proarrhythmic prolongations of QT interval is neither sharp nor well defined but there is now persuasive evidence that a prolongation of QT interval, corrected for heart rate, above 500 ms carries undue risks of TdP, particularly when associated with slow heart rates. Available data also suggest that in individual subjects, (i) an increase of 60 ms or more in peak or maximum QTc interval over baseline or (ii) a postdose QTc interval of 500 ms or more (irrespective of the magnitude of increase from baseline) is highly predictive of the potential risk. Such ‘outliers’ analysis is expected to be included in any regulatory submission.

QT interval and rate-correction

Much has been made of the limitations of Bazett's correction for heart rate. Critics of Bazett's square root formula (α = 0.50) argue that it overcorrects at low and higher heart rates and, therefore, introduces a bias against the drug when the drug increases the heart rate as well. This is true but relevant only for drugs which concurrently have a significant effect on heart rates. Nevertheless, some investigators have recently advocated a greater use of alternative formulae. There is a choice of over 30 correction formulae, but they all have their deficiencies and none is ideal [96]. A preferred alternative seems to be the Fridericia correction. Others have advocated the use of a study-specific correction formula derived from regression of QT interval on heart rate during the baseline placebo treatment. An even more rigorous method of rate correction would be to derive individual correction factors (α) for each of the subjects participating in a study [96].

It is acknowledged that study-specific correction formulae may be less susceptible to the vagaries of changes in heart rate and therefore, more appropriate on which to base regulatory decisions on risk. However, their use (which may not be practical unless large numbers of subjects are studied on placebo with frequent ECGs) may not be without problems. This is well illustrated by data comparing ziprasidone with 5 other antipsychotic agents in the study referred to earlier. The data are shown in Tables 4 and 5. Following the administration of these drugs in the absence of a metabolic inhibitor, there were changes in mean heart rates ranging from −2.9 beats min⁻¹ on haloperidol to +11.2 beats min⁻¹ on quetiapine. Although study 054, as designed, was not powered to provide estimates of the frequency of the outliers with the six drugs, various interesting points emerge. Firstly, the outlier analysis suggests that risperidone, olanzapine, quetiapine and haloperidol may all have the potential to prolong the QT interval according to Bazett's correction but not according to baseline correction yet, clinically, prolonged QT interval and TdP have been reported with haloperidol. Secondly, for mean changes in QTc on these four drugs, there is little to choose between any of the three non-Bazett formulae. Since most drugs acting on the heart have a tendency to increase the heart rate (rather than decrease it as with haloperidol) due to the presence of either α-blocking or anticholinergic activities, the standard Fridericia correction or the study-specific ‘baseline’ formula represents better alternatives since these formulae exaggerate the change at lower heart rates when the proarrhythmic risk is higher. Thirdly, despite the inadequate power of the study to estimate these ‘events’, the data on the number of outliers with categorical changes in corrected QT interval ranks the QT prolonging potential of ziprasidone immediately below that of thioridazine but above haloperidol irrespective of the correction formula but clinically, prolonged QT interval and TdP have been reported with haloperidol as well as with thioridazine and no so far with ziprasidone. Overall, however, it is questionable whether any drug has attracted an unwarranted adverse regulatory evaluation of its potential to prolong the QT interval simply because of any bias resulting from the formula used for rate correction.

Table 4.

Effect of rate-correction formula on mean change in QTc interval from baseline.

	Ziprasidone	Risperidone	Olanzapine	Quetiapine	Thioridazine	Haloperidol
Increase in QT interval (ms)	6.8	−12.1	−8.9	−12.2	18.7	12.5
Change in heart rate (beats min−1)	4.6	9.5	6.5	11.2	5.7	−2.9
Change in QTc (ms) interval corrected by
Bazett (α = 0.50)	20.3	11.6	6.8	14.5	35.6	4.7
FDA-proposed (α = 0.37)	16.5	4.9	2.3	6.9	30.8	6.8
Baseline (α = 0.35)	15.9	3.9	1.7	5.7	30.1	7.1
Fridericia (α = 0.33)	15.5	3.1	1.1	4.8	29.6	7.3

Table 5.

Effect of rate-correction formula on number of outliers with categorical change in QTc interval of >60 ms from baseline in the absence (and in the presence) of a metabolic inhibitor of the drug concerned.

	Ziprasidone	Risperidone	Olanzapine	Quetiapine	Thioridazine	Haloperidol
Increase in QT interval (ms)	6.8 (10)	−12.1 (1.1)	−8.9 (−1.8)	−12.2 (−15.8)	18.7 (33.3)	12.5 (22.5)
Change in heart rate (beats min−1)	4.6 (3.6)	9.5 (0.5)	6.5 (3.0)	11.2 (15.1)	5.7 (−2.1)	−2.9 (−5.7)
Number of patients
Completing the study	31 (31)	25 (20)	24 (24)	27 (27)	30 (30)	27 (20)
With categorical increase in QTc (ms) interval corrected by
Bazett (α = 0.50)	7 (3)	2 (0)	1 (0)	3 (4)	9 (6)	1 (0)
Baseline (α = 0.35)	2 (2)	0 (0)	0 (0)	0 (0)	5 (7)	0 (0)

Conclusions

What is the impact of finding a drug effect on QTc interval on drug development?

In order to ascertain this, the author solicited information on the experience of a major multinational pharmaceutical company. It emerged that over the 18 months to November 1999, 11 compounds were found to have an effect on QTc interval – representing an attrition rate of 10%. Eight compounds were dropped from further development and three projects had slowed down. These 11 compounds included cardiovascular as well as noncardiovascular drugs covering a range of pharmacotherapeutic and chemical classes, and none was intended to have an effect on ion channels. Recently, it has also become evident that when the characterization of the ‘QTc liability’ of the new drug has been deferred for too long, some projects had to be terminated at an advanced stage of clinical development following a late discovery of this liability.

The adverse effect of drugs on QTc interval and the resulting proarrhythmias have proved to be a major public health issue and regulatory authorities have the expectation that this potential risk will be adequately characterized during the development of all NCEs. Given the exorbitant costs involved in drug development, it is vital that this issue is addressed as early as possible with a view either to making a go/no-go decision or to modifying the clinical development programme accordingly. It is not inconceivable that drugs with a potential to prolong QTc interval may be taken forward provided a carefully planned clinical development programme identifies a population in whom the benefits of the drug can be shown to outweigh the small potential risk of proarrhythmias, or the drug can be shown to fulfil an unmet need. For such drugs, the prescribing information would require detailed information on the proarrhythmic risk as well as appropriate contraindications, a description of drug interactions and special precautions and monitoring requirements during clinical use.