: Dr John Urquhart, 975 Hamilton Avenue, Palo Alto, CA 94301, USA (fax: +1 650 324 9739; e-mail: firstname.lastname@example.org).
Abstract. Urquhart J (Department of Epidemiology, Pharmaco-epidemiology Group, Maastricht University, Maastricht, the Netherlands). Controlled drug delivery: therapeutic and pharmacological aspects (Internal Medicine in the 21st Century). J Intern Med 2000; 248: 357–376.
Concerted work to develop human pharmaceuticals based on rate-controlled drug delivery systems began in 1970. Today there are over three dozen such products, plus a few for veterinary use. In addition, osmotic minipumps have been extensively used since 1978, resulting in over 6000 publications in the pharmacological, endocrinological and physiological literature. Rate-controlled delivery provides for drug entry into the bloodstream continuously at either a constant or a modulated rate. By this means, one avoids the usual peak and trough pattern of drug concentrations in plasma, with its echoing peak and trough pattern of drug actions, during the interval between successive doses. In contrast to the happenstance release kinetics of rapid-release dosage forms, rate-controlled delivery systems can be designed to provide specific temporal patterns of drug concentration in plasma, for the purpose of optimizing the selectivity of drug action, the interval between successive administerings of drug and the likelihood that the next administering will occur at the proper time.
Drug delivery systems unshackle drug actions from the limitations imposed by pharmacokinetics when drugs are administered in conventional, rapid-release dosage forms. Because delivery systems provide control over the rate of drug delivery, they are, in principle, able to achieve whatever temporal pattern of drug concentration in plasma is best to optimize three key parameters of pharmaco-therapeutics. These are the:
1selectivity of drug action;
2interval between successive administerings; and
3likelihood that the next administering will occur at the proper time.
One of the main purposes of this paper is to review progress towards the realization of these principles as pharmaceuticals for systemic human or veterinary use, approved for marketing in major pharmaceutical markets. Thirty years ago they were a visionary's dreams, supported by strong inference from pharmacological theory and a few experimental findings [1, 2].
Early commercial successes with this approach triggered an explosion of work in the field of pharmaceutical science on various technical approaches to achieving rate-controlled drug delivery. One can explore that body of work by searching via MEDLINE on ‘drug delivery systems’. It has greatly expanded the conception of drug formulation, from its traditional role as the servant of medicinal chemistry to its present status as a source of pharmaceutical innovation. A few of these elaborations on the basic approaches to controlling drug delivery may become the basis for future products, but they are not a focus of this review, which is to summarize salient ‘lessons’ for pharmacology and therapeutics taught by registered delivery systems-based pharmaceuticals. Up until now, the pharmacological and therapeutic ‘lessons’ taught by delivery systems-based human and veterinary pharmaceuticals seem not to have had much influence on pharmacology and therapeutics, as a survey of today's leading textbooks will reveal.
The familiar ‘peak and trough’ pattern of between-dose concentrations of drug in plasma, with its echoing peak and trough pattern of drug actions, is an artefact of the happenstance kinetics of drug release from conventional, rapid-release dosage forms. Aside from intravenous infusions, it was not until the advent of practical delivery systems that pharmacology, and its sister discipline endocrinology, had practical alternatives in both research and patient care to the happenstance kinetics of conventional, rapid-release dosage forms, of which the simplest is an aqueous solution of drug. Endocrine research, of course, has always had as models the physiological patterns of glandular secretion. In effect, the glands are natural delivery systems for hormones. Glandular secretions are continuous, for the most part, but modulated, rather than unvaryingly constant; a conspicuous exception to continuous secretion is the pusaltile secretion of gonadotropin releasing hormone (GnRH), with its 5-min pulses repeated hourly .
When osmotic minipumps became available in the latter 1970s, endocrinologists were most numerous amongst the early adopters of this method of providing continuous, rate-controlled administration. That early adoption was reinforced by growing interest in peptide hormones, given the practical problems of administering peptides, many of which have such extremely short plasma half-lives that there is no alternative to the infusion mode of administration if one wishes to sustain the presence of such agents for more than a few minutes at a time.
Pharmacology has no such natural models, so continuity of drug administration was long confined to a few agents given only by intravenous infusion, and of course the volatile anaesthetics, which, since their discovery, have been dispensed in a continuous manner by vaporizers of increasing accuracy and versatility, building on John Snow's first design in the 1850s . These vaporizers are, in a certain sense, technological precursors to the delivery systems considered here, with the basic point of differentiation being that those considered here are designed to deliver solutes. For the volatile anaesthetics, precisely controllable vaporizers maintain a specified partial pressure of agent in inspired air, and thus in blood and tissues. No peaks and troughs clutter the pharmacology of the volatile anaesthetics, which have remarkably reproducible actions that create stable conditions for various types of surgery. They could, of course, be administered by intermittent bursts, although no one in their right mind would do so. But with drugs which must be administered as solutes, the long-running lack of practical alternatives caused the kinetic peculiarities of rapid-release dosage forms insidiously to intertwine with basic descriptions of drug actions, and thus with basic concepts of pharmacology, therapeutics and drug regulation, as if there were no alternative to the recurring slug-mode of drug administration. Now, however, after 30 years of technological advances in rate-controlled drug delivery systems, pharmacology is freed from its long history of virtually sole reliance on conventional dosage forms and their primitive kinetics.
The past 30 years of work has brought us approximately 40 new human and veterinary pharmaceuticals, and over 6000 scientific papers in the bibliography of published applications of the osmotic minipumps . Collectively, the literature and labelling of these products and the array of minipump-based studies in the pharmacological, physiological and endocrinological literatures define the present status of rate-controlled, continuous drug delivery. One may view these achievements as a lot or a little. For the past 30 years, 20–50 new human pharmaceutical products have been registered annually, so that innovations via the delivery systems route represent only a tiny fraction of the new pharmaceuticals introduced since 1970.
The new products that have come via the delivery systems route of pharmaceutical innovation are new forms of ‘old’ drugs, i.e. drugs that had some prior usage in conventional dosage forms. The new products that came as new chemical entities, by screening the fruits of medicinal chemistry, have virtually all been identified, tested and developed in conventional dosage forms, in the repeated-slug mode of drug entry into the body, with the usual peak and trough patterns of concentrations and actions. The intriguing question is whether we have missed valuable therapeutic agents by virtually sole reliance on this one, rate-uncontrolled approach to putting drugs into the body.
A note on terminology
The term controlled drug delivery is shorthand for dosage forms that utilize membrane technology to control the rate at which, after administration, they release drug to the surrounding biological milieu. In contrast, the release of drug from conventional dosage forms relies on dissolution to release drug, usually with rapid release during a small fraction of the interval between doses. So-called ‘sustained release’ or ‘slow release’ formulations constitute a rather ill-defined intermediate group, in which the formulator's art is used to lessen somewhat otherwise initially high rates of drug release, and slow somewhat their subsequent decline. Marketing claims and other product-specific considerations have created a great deal of ad hoc terminology, so that there is no widely recognized set of terms or standards that clearly distinguish rate-controlled drug delivery systems from other formulations.
The individual who can justifiably be designated the leading pioneer of this field, Alejandro Zaffaroni, coined the term ‘therapeutic system’ to designate pharmaceuticals that delivered their drug at a specified rate, in vivo, for a specified period of time [1, 2]. Indeed, several of the pioneering products in the field, developed by ALZA Corporation, the firm he founded in 1968, entered the US market with the therapeutic system designation.
Whilst much has been made, and written, of constant-rate drug delivery systems, that ideal is only realized with nonvolatile drugs when they are given via infusion pumps. The actuality with tablet/capsule-sized delivery systems used in ambulatory care is invariably a time-sequence of rates, starting and ending at zero, with nearly constant or, in some instances, gradually declining rates in between. It is generally accepted usage in technical circles that a delivery system is described as ‘constant-rate’ when a substantial majority of its contained drug is delivered at an approximately constant rate, but the potential for conflicting views of how best to characterize time-varying rates is evident.
Constancy of drug delivery rate is not the optimum pattern for all drugs, as some, e.g. nitroglycerin, manifest tolerance, in which drug actions fade despite the continuing maintenance of drug concentrations in plasma that initially had been associated with therapeutic action. Avoiding or minimizing the effects of tolerance calls for drug delivery according to a drug-specific temporal pattern of varying rates, designed to minimize the receptor down-regulation that appears to underlie most, if not all, instances of tolerance. Research on this approach is hardly even an agenda item in contemporary pharmacodynamics, but was undertaken for nitroglycerin, as discussed later. The field of controlled drug delivery began, however, with the assumption that constant-rate delivery would be the desired pattern of delivery for the vast majority of drugs. With the exception of transdermal nitroglycerin, most of the delivery systems products developed so far approximate continuous, constant-rate drug delivery.
Specifying strength and duration
Transdermal and implant products consistently use in vivo drug delivery rate to specify strength. Typically they leave as much as half of their original drug content undelivered at the time their specified duration is reached. The undelivered drug provides the physical–chemical driving force for ongoing drug delivery, and the limit of duration is reached when depletion of drug brings the physical–chemical driving force down to within some reasonable margin of the point at which rate, and thus strength, would reach a minimum effective level.
In contrast, the strengths of rate-controlled oral products are specified in the conventional way, by the quantity of drug contained in the dosage form, with their points of differentiation with conventional oral forms residing in dosing frequency, the label's description of the nature of the formulation and any special therapeutic claims. Rate-controlled solid oral dosage forms must achieve usual standards of reasonably complete and reproducible bioavailability. Thus, they are designed to deliver their contained drug within the limits of normal gastrointestinal (GI) transit times. These limitations could change if promising approaches to gastroretentive systems  are successful, but no such products are registered yet. The present limit on duration of rate-controlled oral products is currently set by the transit of solid dosage forms through the gastrointestinal tract: the longest of the oral delivery system products has a delivery time of 18 h .
There are natural limits on the duration of drug delivery from a single delivery system. One is size, which is determined by the rate at which the drug must be delivered to achieve requisite concentrations in plasma or other fluids, and actions. Drugs with very high molar potencies can, of course, exert therapeutic actions when administered at rates in the µg h−1 range, or lower, but the vast majority of agents in the present pharmacopoeia require rates of administration in the mg/hr range, with a few being even higher. If one contemplates delivery of a drug from, for example, an implant with an average duration of 3 years, with an hourly delivery requirement, e.g. 10 mg h−1, the aggregate quantity of drug delivered throughout the 3-year period would be 3 years × 365 days year−1× 24 h day−1× 10 mg h−1= 262800 mg, i.e. approximately 263 gm, which, for a drug of specific gravity 1, occupies a volume about the size of a coffee cup. To incorporate a drug into a working delivery system entails the addition, with one notable exception discussed below, of typically three or more times the drug's mass and volume in the form of membranes, excipients and vehicle(s), thus inflating the coffee cup to a volume of about 1 L – possibly an implant for elephants, but not for humans. Thus, delivery systems bring a new incentive to develop drugs of exceptionally high molar potency. If, in the above example, the required delivery rate were 10 µg h−1, instead of 10 mg h−1, the system volume could be approximately 1 mL, instead of approximately 1 L. This, of course, is the point at which both pharmacokinetics and pharmacodynamics impact drug delivery system design: the product of the pharmacokinetic parameter, plasma clearance and delivery rate establish the concentration of drug in plasma. The higher the clearance, the higher the delivery rate needed to achieve a certain concentration in plasma. But pharmacodynamics establishes the concentration range within which the drug's actions are exerted, which may show drug-specific values that vary by several orders of magnitude within a drug class.
Transdermal delivery also requires drugs of high molar potency – in the range of 1–2 mg day−1 or less – because of the low permeability of skin to solutes. Another factor is the reactivity of the skin to the drug, some of which is dose or concentration dependent, so that the smaller the amount of drug that has to pass a given area of skin, the less the potential for dose- or concentration-dependent irritation or pigmentation. These considerations limit the value of various maneuvers that can increase the permeability of skin to drugs.
The limit of duration of transdermal delivery systems is adhesion to the skin. The physiological process of cornification of skin cells creates a constant turnover of stratum corneum, so that the biological anchor to which transdermal systems adhere persists for more than 7 days in most patients, but not all, although a high enough percentage to make practical the week-long transdermal form of clonidine .
Exceptionally long-duration delivery systems have been developed for veterinary uses. The principle of a gastroretentive delivery system is easily realized in ruminants, by weighting the delivery system sufficiently to assure that it sinks to the bottom of the reticulum, the second of the four chambers of the rumen. Once at the bottom of the reticulum, an adequately weighted rumenal delivery system will remain in place with a high degree of certainty throughout the animal's life. Thus, it has been possible to develop and register, e.g. a 3-month duration, rate-controlled rumenal delivery system for the potent parasitocide ivermectin . Rumenal delivery systems must, of course, remain within the size limits imposed by their need to pass down the oesophagus and into the rumen without difficulty. So even these veterinary applications emphasize the value of high molar potency drugs, for the ultimate limits on their functional lifetime is the amount of drug they can contain, and the rate at which the drug must be delivered to achieve and maintain therapeutic efficacy.
Some technical basics
Naturally, with its rate of drug release reliably predetermined, a drug delivery system will release its defined quantity of contained drug during a defined period of time. The result can therefore be continuous delivery of drug, at an essentially constant rate or a predetermined sequence of rates, throughout a time-period of reproducible duration. This last point is a key factor in delivery system products for use in food-producing animals, which must have a well-defined drug-free interval to meet regulatory requirements for minimum residues of drug prior to slaughter.
Two basic technological approaches are used to achieve rate-controlled drug delivery. Both involve drugs that are surrounded by a membrane. One approach uses osmosis to control the rate at which water from the surrounding biological milieu penetrates the enclosing membrane to gain access to the drug, thence to pump a solution or suspension of the drug into the surrounding milieu. The other approach uses the enclosing membrane to control the rate at which the drug exits to the surrounding biological milieu, by either of two processes: solution diffusion through a solid membrane; or diffusion through the pores of a microporous membrane. Each of these means for achieving rate-controlled delivery have been taken multiple times through the full cycle of product development, clinical testing and registration in the US and other markets; most are prescription-only products, but a few are available over-the-counter. Today, as noted above, there are about three dozen such products in the human pharmaceuticals marketplace , plus several veterinary products [9, 11], plus a variety of osmotic minipumps in the scientific products marketplace .
From the medical perspective, of course, the ultimate arbiter is not the kinetic niceties or technical wizardry of drug delivery systems, but the therapeutic claims that can be made for resulting pharmaceutical products, based on data from properly controlled clinical trials. On the research side, the ultimate arbiters are the scientific opportunities created by rate-controlled drug or hormone administration and the scientific value of contrasting drug actions elicited by different temporal patterns of drug entry into the body. The latter point has its own terminology within the field of cancer research, where it is known as ‘schedule dependency’ of drug actions.
Elementary osmotic pump – prototypical rate-controlled drug delivery system
The elementary osmotic pump (EOP) is an illustrative example of a versatile delivery system [12, 13]. In its most widely used form, as a rate-controlled oral dosage form, the EOP consists of a tablet-sized core of solid drug, overcoated with a thin, semipermeable membrane, through which is placed a single, small hole, which serves as a vent from the core to the outside milieu. The core of solid drug exerts a measurable osmotic pressure, proportional to the drug's water solubility. When the EOP is placed in an aqueous medium, e.g. the lumen of the GI tract, water commences to move across the semipermeable membrane, driven by the difference in osmotic pressure between the core of the EOP and the lumenal fluid, the latter having an osmolarity of 310 mosm L−1, and an osmotic pressure of approximately 7.75 atmospheres (atm) . In contrast, the osmotic pressure of a saturated solution of a suitable candidate drug is in the range of dozens to several hundred atmospheres, e.g. the osmotic pressure of a saturated solution of potassium chloride is 245 atm. The net difference in osmotic pressure drives the influx of water, e.g. 245–8 = 237 atm in the case of KCl. The term ‘semipermeable’ means that the membrane is permeable to water but not to the drug, so water enters via the membrane at a constant rate and drug in solution exits via the vent.
The semipermeable membrane, typically composed of cellulose esters, undergoes little or no expansion in volume as water moves into the core. The hydrostatic pressure needed to drive the saturated solution of drug out through the vent is of the order of 1 atm or less – negligible in relation to the osmotic pressure difference across the membrane. As long as solid drug persists within the core of the EOP, the osmotic pressure of the core is constant, thus maintaining a constant rate of water influx across the membrane. Thus, the time course of drug delivery rate from the EOP begins at zero, rises within a few minutes to a steady level, which is maintained until the last of the solid drug in the core has been solubilized. Thereafter the rate of drug delivery declines, as each quantum of incoming water displaces an equal quantum of drug solution through the vent, whilst simultaneously diluting the solution of drug in the core. So, for two reasons the rate of drug delivery spirals downwards, once the last of the solid drug in the core has gone into solution. The first reason is that incoming water progressively dilutes the drug solution within the core, reducing the osmotic pressure difference across the membrane, thus progressively reducing the rate of water flow into the core. The second reason is that the rate of flow of drug solution through the vent declines concomitantly with the decline in influx of water, and the concentration of drug in the fluid issuing from the vent also declines. The pumping rate of drug, which is the product of the rate of solution efflux and the concentration of drug in the effluent solution, falls, as both terms in the product decline together. The decline continues until the osmotic pressure equilibrates on either side of the membrane, and pumping ceases.
The water solubility of the drug is a critical parameter for determining the percentage of contained drug that an EOP will deliver at constant rate. For simplicity, let us assume that the volume of the core is 1 mL, that the drug has a specific gravity of 1.0, and that the concentration of drug at saturation is 0.5 gm mL−1, i.e. a 50% solution. As long as solid drug remains in the core, the osmotic pressure of the core is constant, and so are the rates of influx of water and of efflux of saturated solution of drug. Incoming water has two places to go – most goes out of the vent as the aqueous component of a saturated solution of drug; the other destination of incoming water is to remain in the core, occupying the space previously occupied by solid drug that has been solubilized and pumped out through the vent. At the moment when the last of the solid drug in the core goes into solution, the 1 mL core volume consists of 1 mL of a 50% solution of drug, i.e. 500 mg of drug in solution. Thus half of the original 1 gm of drug has been delivered at constant rate (neglecting the brief start-up transient). Thereafter, drug delivery rate declines, as described above. If the core were composed of an agent of lower solubility, e.g. saturation at 20% solution, the amount of drug remaining in the system when the last of the solid drug has gone into solution would be 200 mg, with 80% having been delivered at constant rate. Of course, the lower solubility means a smaller osmotic pressure, and a lower limit on drug delivery rate. By changing the composition of the membrane to increase its hydraulic conductivity, the effect of a lower osmotic pressure can be offset, within certain limits. Advances in osmotic pump and membrane technology over the past quarter-century have made it possible to extend the capabilities of elementary osmotic pumps so that they can reliably deliver agents of relatively low solubility at essentially constant rates.
Implicit in the foregoing is a strong theoretical basis for quantitatively predicting the time course of drug delivery from an EOP placed in fluids of physiological osmolarity. Water movement by osmosis has a temperature coefficient of approximately 3% per degree centigrade . As both the osmolarity and temperature of body fluids are homeostatically closely regulated, the impact of variations of either on the rates of delivery of drug from EOP tablets is small, although predictable in relation to other variables that often influence the disposition of drug in the body. Special advantages of the EOP formulation are that, unlike many conventional formulations, its delivery rate profile is independent of pH and of the extent of stirring of the surrounding fluid [12, 13].
A notable feature of the EOP is its uniquely high volume-efficiency. Basically the drug that the EOP is meant to deliver comprises almost its entire mass and volume. In its simplest embodiment: the only nondrug component of the EOP is the membrane, which is a thin film that coats the drug-containing core. The membrane typically constitutes 10% or less of the total volume of the EOP. Thus, upwards of 90%, or more, of the volume of an EOP can be drug. Of course, if the drug requires excipients to facilitate its tabletting and/or minimize its friability during the coating process that applies the semipermeable membrane, then some space will have to be given over to those agents. Moreover, if the drug in its simplest chemical form is at either extreme of water solubility, it may be necessary to use a salt or ester form of the drug to attain the requisite intermediate range of solubility needed for optimum osmotic pressure and to maximize the proportion of drug delivered at constant rate. The added mass and volume created by the derivatizing moiety is another element that reduces the volume efficiency of the delivery system. These various modifications may bring the volume efficiency down from approximately 90% to the range of 40–80%. Other approaches to achieving rate-controlled drug delivery have volume efficiencies of less than 30%, and in some instances less than 5%. The concept of volume efficiency applies to all dosage forms, rapid-release as well as rate-controlled delivery systems.
In considering variability in the performance of drug delivery systems, one should put the physical–chemical sources of variability in rank order with other sources of variability in drug administration and disposition in body fluids. An often overlooked, large source of variability arises from the fact that drug dosage is rarely adjusted in relation to body size, or any index thereof (e.g. lean body mass), in adult medicine, despite an interpatient range in body size of about 3-fold in adults. Oncological drugs are an exception, of course, but the general ignoring of interpatient variations in adult body size amounts to a de facto acceptance by prescribers of a roughly 3-fold variation in the relation between dose and average concentration of drug in body fluids. Paediatricians, of course, routinely adjust prescribed doses by indices of body size. The upshot is that prescribers in adult medicine have a restricted license to object to minor sources of variance in dosage form or delivery system release kinetics.
The EOP is but the simplest of an extensive series of osmotically powered, rate-controlled delivery systems. The more complex systems, which are also amenable to development as oral dosage forms, use specially compartmentalized excipients, rather than the drug itself, to bring water into the system. In such elaborations on the basic theme, drug is pumped out of the delivery system by displacement, with the drug in suspension, thereby avoiding concerns that extremely concentrated solutions of drug may have some potential for irritating tissues in the vicinity of the delivery system. Osmotic pumps that work by displacement are not as volume efficient as the EOP, because of the space occupied by additional membranes and excipients. These elaborations of osmotic pumping technology are capable of delivering virtually all of their contained drug at a constant, prespecified rate in suspensions or dilute solutions .
Osmotically powered drug delivery systems are extremely versatile in biomedical applications. Although originally conceived for use as oral dosage forms for humans, they have served as both the starting point for further innovations in osmotically powered rate-controlled delivery, and have been adapted to a variety of other uses, including implantable pumps [5, 9], rumenal delivery systems of multimonth duration in food-producing ruminants , and multipurpose drug delivery systems for use in experimental animals . The versatility of the approach is illustrated by the wide range of sizes, pumping rates and durations of pumping that are available for the miniature osmotic pump implants used in experimental animals ( Table 1)
Table 1. Dimensions, pumping rate and duration of delivery from miniature osmotic pump implants *
Reservoir volume (µL)
Pumping rate (µL h−1)
For use in experimental animals only; data courtesy of Durect Corp.
Applications of drug delivery systems that have broken new ground in clinical or basic pharmacology
To understand what contributions delivery systems have thus far made, it is useful to consider the attributes of some of the delivery system-based products that have broken new ground in clinical pharmacology by improving selectivity of action of the delivered drug. Then we shall turn to examine future prospects.
Transdermal scopolamine for motion sickness
The indication for transdermal scopolamine is prevention of motion sickness. This development was based on recognition, especially in military medicine, that scopolamine, given by rapid-dissolution capsule or injection, is the most effective single agent against motion sickness, although both oral and injected scopolamine create a mixture of unpleasant, occasionally debilitating side-effects – drowsiness, dry mouth, cycloplegia, constipation, tachycardia, anhydrosis, difficulty urinating and a variety of central nervous system effects that, depending on dose, range from restlessness to hallucinations and/or amnesia. All are reversible after dosing stops. This mixture of side-effects made the drug basically impractical as a single agent, despite its effectiveness against motion sickness. For some years prior to the introduction of the scopolamine ‘patch’, the US Air Force dispensed a combination of scopolamine and ametamine to prevent airsickness, the main role of ametamine being to attenuate or mask the side-effects of scopolamine [14, 15].
The transdermal form of scopolamine was designed to deliver drug at the rate effective against motion sickness, which fortuitously is lower than the rates of scopolamine delivery needed to elicit most of the other effects of the drug. Placebo-controlled clinical trials of transdermal scopolamine showed a 75% reduction in the incidence of motion sickness, with about two-thirds of patients having mostly a mild degree of dry mouth, and with about 15% of patients becoming drowsy . It is unclear whether all instances of drowsiness are side-effects of the drug or early manifestations of the motion sickness syndrome, which includes drowsiness. A rarely reported postmarketing finding has been hallucinations . A mild degree of cycloplegia, which is optically correctable, begins to become evident after several days of transdermal scopolamine use . The net result is that scopolamine, delivered at a low, essentially constant rate, protects a large majority of patients against motion sickness, with only minor, reversible side-effects.
A useful construct for interpreting the contribution made by the rate-controlled delivery system is to rank-order scopolamine's many actions by the minimum delivery rate (MDR) needed to elicit each, starting with the lowest rate and proceeding to the highest. Scopolamine's action to inhibit motion sickness appears to be second on the rank-ordered MDR list. The first on the list is the drug's weak vagolytic action that induces a few beats per minute increase in heart rate; the third is its inhibition of gastric acid secretion ; fourth is inhibition of salivary secretion . It is, of course, fortuitous that the effect of scopolamine on motion sickness is elicited at such a low rate of delivery. If the result had been otherwise, so that a much higher rate were needed, then rate-controlled delivery would have provided no advantage over conventional forms of the drug, or could even have created a more troublesome time-course and mix of side-effects than rapid-release administration.
The transdermal scopolamine product is widely used by people prone to motion sickness, especially sea-sickness. It has never been a high-revenue product because most people who benefit from its use have only infrequent, brief need for it. The product's importance lies in the fact that it rewrote the pharmacology of a century-old agent, whose therapeutic utility was virtually nil because of its multiple, mostly unpleasant actions. Perhaps its relatively selective inhibition of gastric acid secretion might have had practical therapeutic value if the product had been developed earlier, during the era of anticholinergic approaches to the inhibition of gastric acid secretion.
The pharmacological principle this product teaches is the utility of examining drug actions under conditions of various constant rates of administration, to understand how each of a drug's actions rank-order with respect to the minimum rate of delivery needed to elicit the desired response , and thus whether rate-controlled drug delivery might enhance the drug's selectivity of action.
Constant-rate vs. intermittent, rapid-release administration of bleomycin
A corollary of the rank-order MDR approach is to undertake, when technically feasible, comparative dose–response testing, with drug given in conventional, rapid-release form on the one hand and given by constant-rate administration on the other hand [18, 19]. Some drugs have adverse effects that are only triggered when drug concentration in plasma reaches a high level. If so, it may be advantageous to avoid the repeated peak concentrations produced by rapid-release dosage forms. On the other hand, if adverse effects depend upon relentless presence of drug, then the usual peak and trough pattern may minimize adverse effects, if the troughs are low enough and long-lasting enough. If trough concentrations are associated with loss of therapeutic action, then constant-rate administration will be advantageous.
Historically the first example of these contrasting possibilities was an animal model study of the dose-dependent, antineoplastic effects and toxic effects of bleomycin . The study was performed two decades ago, but it remains a lesson for those who would perseverate with exclusive use of conventional dosage forms. The study compared dose–response relations elicited by the usual periodic injection mode of rapid-release dosing with those of constant-rate delivery from implanted osmotic minipumps. The dose–response relation for antineoplastic efficacy was shifted substantially to the left with constant-rate delivery compared with intermittent, rapid-release dosing, thus indicating that constant-rate delivery requires a smaller total amount of drug to achieve a given level of antineoplastic activity. In contrast, the dose–response relation for pulmonary fibrosis was shifted to the right with constant-rate delivery compared with intermittent, rapid-release dosing, thus indicating that constant-rate delivery requires a larger total amount of drug to achieve a given level of toxicity.
The foregoing study was published in 1978. Its striking results gradually influenced experimental designs in the screening and testing of new anticancer drugs. A decade later, Collins and his colleagues reviewed developments, noting the then rapidly growing number of clinical studies that evaluated effects of continuous exposure, due in part to technical advances which made the use of continuous delivery feasible . Today, there is a bibliography of over 600 papers in which osmotic minipumps have been used in studies that compare the effects of continuous, constant-rate infusion versus intermittent, rapid-release dosing of a wide array of drugs .
The conclusion to be drawn from all this work is not that constant-rate delivery is some kind of panacea, but rather that comparative evaluation of drug actions elicited by constant-rate delivery versus by intermittent dosing with rapid-release dosage forms illuminates the drug's pharmacodynamics, which in turn informs the choice of formulation and regimen. The effects of delivery kinetics on the relative positions of dose–response relations are drug-specific: some agents have the shifts seen with bleomycin, others have opposite shifts, and still others show no change [22–24].
Nifedipine-GITS for vasospastic angina, chronic stable angina and hypertension
Nifedipine was originally formulated in a conventional capsule, providing rapid release, rapid absorption, and the usual peak-and-trough pattern of drug concentration in plasma. Its actions are closely associated with the drug's concentration in plasma . Nifedipine has a plasma half-life of approximately 2 h, so in rapid-release capsule form thrice daily dosing is needed for continuous action. Although the drug has a demonstrable antihypertensive effect, its actions on blood pressure were deemed too erratic, when administered by rapid-release capsule, to permit its use for hypertension, so it was indicated for use only in vasospastic angina and chronic stable angina . A frequent but episodically occurring cluster of side-effects of the rapid-release capsule form is called ‘reflex tachycardia’– a mixture of light-headedness, facial flushing, heat sensation, dizziness, giddiness and palpitations. Various mixtures of these side-effects occur episodically in about 25% of patients versus an incidence in placebo recipients of approximately half that level . Nifedipine in quick-release form has a so-called ‘first-dose effect’, namely that an initial dose titration is indicated to minimize hypotension-related side-effects .
A small pharmacodynamic study, using programmed intravenous infusions, examined the consequences of different temporal patterns of nifedipine concentration in plasma . The results suggested that reflex tachycardia is triggered by a high rate of increase in the concentration of nifedipine in plasma. When high rates of increase in concentration were avoided by constant-rate infusion, producing a smooth, asymptotic approach to a constant plasma level, there was consistently no triggering of reflex tachycardia. Instead, the infusion produced a gradual, asymptotic decline in diastolic blood pressure, closely and consistently related to the asymptotic rise of the drug's concentration in plasma . These findings suggested that an oral, rate-controlled form of nifedipine might have substantial advantages over the conventional, rapid-release form – avoiding reflex tachycardia, producing a consistent, clinically useful antihypertensive action, and having a once-daily dosing regimen. Those expectations were borne out by the resulting nifedipine-GITS product , the acronym ‘GITS’ standing for ‘gastro-intestinal therapeutic system’, based on the above-mentioned use of the term ‘therapeutic system’ for rate-controlled, duration-specified pharmaceuticals. This product has been very widely used and, until 1995, was the largest-selling cardiovascular product in the history of the US pharmaceutical market. As such, it is the biggest commercial success amongst all of the delivery systems-based products .
Looking back, nifedipine was the first of its class to reach the market – initially as a nonspecific peripheral vasodilator in the German market, and then later entering the US market under the rubric of the calcium channel blockade mechanism. A later entry was diltiazem in conventional form, which is basically free of the reflex tachycardia side-effect that vexed nifedipine. It seems reasonable to believe that, had diltiazem been the first calcium channel blocker to enter the market, nifedipine would never have been able to compete as a late-entry product – a fate that befell nicardipine. As the lead-off product in its class, nifedipine had a considerable success in the marketplace, although diltiazem, after its market entry, began taking most of the growth in the market. When nifedipine-GITS appeared it was masterfully launched in the US market by Pfizer, first capturing most of the conventional nifedipine business, then taking most of the growth in the calcium antagonist market. At its peak in the early 1990s, nifedipine-GITS had more than doubled the sales that the conventional form of nifedipine had had during its earlier heyday, and became, as noted earlier, the largest-ever cardiovascular product in the US market until 1995. Later, diltiazem was re-formulated as a delivery system-based product, which reduced the dosing frequency but made no claims for changes in the clinical profile of the drug in conventional form.
In retrospect, one could say that nifedipine in conventional formulation had clinical attributes good enough to reach the market and command sizeable sales, but with problems big enough that when solved they resulted in a clearly superior product, with a major new claim: improved clinical profile and simplified dosing regimen. It is probably an unusual scenario, from both the pharmacodynamic and commercial perspectives. In any case, the resulting rate-controlled product is not only clearly superior to rapid-release nifedipine, but also strongly competitive with other products in the calcium antagonist class.
Like the transdermal scopolamine delivery system, transdermal fentanyl has a 3-day duration of action from a single application, but instead of a single rate of drug delivery for all patients the fentanyl product is provided at four different rates: 25, 50, 75 and 100 µg h−1, which are nominal in two senses:
1the delivery rate changes somewhat during the 3-day functional lifetime of each system;
2the dispersion of in vivo delivery rates is fairly wide .
Provision of systems at four different nominal rates reflects a diversity in both the pharmacokinetics and the pharmacodynamics of fentanyl. The relation between rate and concentration is variable, as is the relation between drug concentration and analgesic effect, and as is the rate at which individual patients develop tolerance, requiring higher delivery rates for effective analgesia . The rather complex instructions for using transdermal fentanyl are aimed at exploiting the MDR approach, as discussed above with scopolamine, with the important difference being that the respective MDRs for analgesia and for respiratory depression are only separated by a factor of approximately 2 . Thus, the product labelling has a number of strongly emphasized warnings – against use under the age of 12, against use in those under 18 who weigh less than 50 kg, for extreme caution when used together with many other CNS-active agents, and in situations in which the surface temperature in the vicinity of the ‘patch’ might be elevated, thus increasing the rate of fentanyl delivery . In short, this product has a narrow margin for error, requiring exceptional care amongst those who prescribe and use it, with the need for individualization of delivery rate being paramount. These many complicating factors combine to contraindicate the product's use in situations in which pain is rapidly changing, e.g. postoperative or post-traumatic pain, or with pain that can be controlled by non-narcotic analgesics.
All this complexity notwithstanding, however, the product's salient advantage is its provision of round-the-clock continuous analgesic effect in patients suffering from chronic pain. Its value reflects the truth of a now widely accepted lesson first taught by the late Dame Cicely Saunders, from her pioneering work with hospice care for dying patients. The lesson is that poor-quality analgesia and needless suffering are the frequent results of the conventional ‘prn pain’ analgesic order, which in effect requires the recurrence of pain to initiate remedication. Saunders also taught that it was essential when using short-duration analgesics – the only forms available in her time – to remedicate whilst the prior dose was still acting, because the recurrence of pain not only demoralizes the patient but also makes it much more difficult to restore analgesia. It included awakening sleeping patients to remedicate, rather than allowing them to sleep on only to awaken later in pain. Tolerance occurs, of course, during long-term use of fentanyl or other opiates, requiring use of higher rates of delivery, but it does not obviate the risk of respiratory depression, which is always present, but fortunately develops tolerance concurrently with analgesic tolerance. The ratio of MDRs for the two effects would appear to remain unchanged as tolerance develops .
This product, indicated for the management of attention deficit disorder (ADD), was approved by the Food and Drug Administration (FDA) just as this review was being written, so only the premarket data are available. The product utilizes an adaptation of osmotic pump technology to emulate, with a single morning dose, the time-course of drug concentration in plasma produced by a morning dose, followed by a mid-day dose, of rapid-release drug. The new formulation eliminates the mid-day dose, which is usually given at school, with its attendant logistical problems. The rate-controlled formulation has a sequence of programmed rates: first, a loading dose, followed by a two-step sequence of rates . On face value, the product overcomes one of the barriers to treatment of ADD, but, with fewer than 500 patients involved in premarket studies, there is little more that can be said until use-experience grows.
The original nitroglycerin patch products, introduced about 20 years ago, were designed for continuous wearing, providing essentially constant-rate delivery around the clock, with 24-hourly patch replacement. They were registered on bioequivalence data, showing that they could maintain concentrations of nitroglycerin in plasma comparable to those maintained by 8-hourly re-administered ointment forms of the drug. Once in the market, however, the impact of the ‘patch’ products on exercise tolerance was studied by a number of academic researchers, who found that initially beneficial effects faded away within a few hours, not to return as long as nitroglycerin administration continued by successive daily replacements of patches. Only after much debate and discussion was a cooperative study arranged, involving the FDA and the several manufacturers of nitroglycerin patch products. The upshot was confirmation of the development of tolerance, and the recommendation that the nitroglycerin patch products be removed at bedtime, with a fresh system re-applied upon awakening. All the products were re-labelled with that recommendation [cf. 29, 30].
Through the rise of interventional cardiology and development of coronary bypass surgery, the clinical management of angina and other manifestations of coronary insufficiency has undergone immense changes during the two decades since these products were registered. At the time, the nitroglycerin patch products were widely used and commercially very successful. The postmarketing problems with their efficacy served to drive home the point to all involved in drug delivery systems and their development that constant-rate drug delivery is not the optimum for all drugs or clinical situations. Fortunately, there were no evident adverse consequences of the continuous delivery product during the time required to develop the evidence on which the regimen was revised to restore useful drug action [29, 30].
OSMOSIN: snake in the garden?
In 1983, an EOP form of indometacin was introduced by Merck, Sharp & Dohme in several European markets, with the trademark OSMOSIN. It had a once-daily dosing regimen and was promoted on the basis of placebo-controlled studies in patients with osteo-arthritis , in which the incidence of familiar side-effects of indometacin were not statistically different in the group assigned to OSMOSIN from that amongst those who received placebo. The incidence of the familiar nonselective NSAID-associated side-effects were far from zero in the placebo group, but the claim of placebo-identical side-effects probably prompted many to infer that the new form of indometacin was free of side-effects – in effect, ‘indometacin without tears’.
When the product entered the UK market, it had the steepest growth in new prescriptions in the history of that market up to that time. After some months, however, reports began to appear via the UK's spontaneous reporting system (‘Yellow cards’) of GI bleeding and perforations of the gut at various points from the oesophagus to the colon, some of which were fatal. Despite the British origin of the data, these adverse reactions became a major topic in the German news media during August and September of 1983. Rapidly escalating controversy led the German regulatory authorities to suspend the product's registration in September 1983, which action was promptly followed by regulatory authorities in the other half-dozen countries in which the product had been registered.
A widely held view at the time was that the stream of concentrated drug, which had been coformulated with potassium bicarbonate, exerted a focal irritating effect on the mucosa, leading in the extreme to bleeding and/or perforation. The German news media dubbed this hypothesis the ‘blowtorch effect’.
Later, a different view of events emerged. It was based on the fact that the promotional claims made for the product encouraged its use in patients who had had prior gastro-intestinal difficulties and who were therefore at high risk of recurrent problems irrespective of the medicines they took. The occurrence of this phenomenon, which came to be known as ‘channeling’, was suggested by an unusually high incidence of GI problems amongst later recipients of OSMOSIN, prior to the first prescription for the product. This finding emerged from a prescription event-monitoring study carried out in the wake of the withdrawal of OSMOSIN from the UK market .
It is noteworthy that, a few years later, SmithKline introduced into the Dutch market a delivery system form of ketoprofen, carrying the tradename, OSCOREL, with a marketing approach that closely resembled the ill-fated one that MSD had used for OSMOSIN. As described in , a similar sequence of unexpectedly high reporting of adverse events ensued for the new product. A key difference, however, was that the scenario had been predicted beforehand, and that the channeling bias amongst recipients of OSCOREL was quickly established, so that regulatory review could proceed unperturbed by overheated public controversy. The regulatory review concluded with the judgement that the product posed no unusual hazards .
The OSMOSIN crisis catalysed not only the recognition of the channeling phenomenon but also the development of a novel method for assessing in human volunteers the local irritation potential of various drug formulations . The results of those local irritation studies indicated the high potential for local irritation of highly soluble salt forms of various drugs, in comparison to their relatively insoluble native forms . That finding catalysed extensions of the osmotic delivery technologies towards mechanisms that could deliver drug in suspension rather than in concentrated solutions.
The OSMOSIN crisis also highlighted the special difficulties that a new form of an already established drug faces when entering the contemporary pharmaceutical marketplace. Techniques for postmarketing surveillance and pharmaceutical risk assessment that now are standard hardly existed prior to 1980, so that earlier products, e.g. the conventional dosage form of indometacin, gained wide clinical usage with little or no systematic evaluation of the hazards associated with their use. The absence of such ‘baseline’ information helped to create the impression that the adverse events reported for the new form of the drug represented greater hazards from which the long and widely accepted earlier form of the drug had been, somehow, exempt. In the end, there is no way to account definitively for this unfortunate event, which at the time threatened to halt further development of osmotically rate-controlled oral products. The episode served to catalyse a number of developments: better postmarketing surveillance, recognition of the channeling phenomenon and of the need to assure symmetrical comparisons between new and old forms of long-used drugs, and new extensions of osmotic delivery technology to avoid delivery of concentrated solutions of ionic drug.
The 5-year levonorgestrel implant
This product was developed by researchers at the Population Council, an American nonprofit foundation focused on improving methods of family planning. The high molar potency of the contraceptive steroids and their solubility in silicone rubber allowed application of the solution diffusion approach to rate-controlled, multiyear delivery of the progestational steroid, levonorgestrel, from a cluster of subcutaneous implants.
The desire to avoid an overtly surgical procedure for implant placement forces implant designs toward cylinders small enough to fit into the lumen of the largest diameter object that could plausibly be called a ‘needle’. The 5-year implant form of levonorgestrel, trademarked NORPLANT, illustrates the impact of size limitations on product design.
Levonorgestrel is the active isomer, so its use in place of the racemic mixture halves the volume of drug. The delivery rates achieved by the final product start out at about 85 µg day−1, declining gradually to about 35 µg day−1. Approximately 110 mg of drug is delivered during the 5-year period recommended between insertion and removal of the implant. Drug is contained in six silicone rubber (dimethylsiloxane/methylvinylsiloxane copolymer) tubes, each having a diameter of 2.4 mm and a length of 34 mm, and each containing 36 mg of levonorgestrel , for a total drug load of 36 × 6 = 216 mg. The total volume of these six tubes is about 920 µL, of which about 24% is occupied by drug.
The mechanism of drug release is solution diffusion: drug dissolves in the wall of the tube, diffusing towards the outside, where processes of drug absorption into capillaries and lymphatics in surrounding tissues maintain a nearly zero concentration of drug at the outer surface of the tube. The concentration gradient across the tube wall is constant as long as solid drug in the lumen contacts the tube's inner wall. The decline in delivery rate occurs because ongoing delivery of drug, amounting cumulatively to 110 mg by the end of year 5, is continually subtracted from the original content of 216 mg. In the process, the area of membrane in contact with drug falls steadily. The residual amount of undelivered drug provides not only the motive force for continuing drug release through the end of the 5th year of use, but also a certain margin of safety against two kinds of variability:
1delayed institution of follow-on contraception;
2tubes that deplete themselves of drug soonest, from a combination of two statistical extremes – lowest drug loading and highest release rate.
To support a claim that all systems provide 5 years of effective rates of drug release, the average system has to be designed to provide the delivery of effective rates for about 40% longer than the labelled duration, i.e. approximately 7 years for the 5 years claim. This difference is not unique to NORPLANT, but applies to any type of implanted drug delivery system.
The economic temptation is obvious to play the odds and hope for several extra years of satisfactory contraception beyond the 5th year. This is an area where judicious use of therapeutic drug monitoring could inform judgements about when a particular patient's implant was approaching the end of its useful lifetime. In fact, the US labelling for NORPLANT uses ambiguous language to describe the duration of action of the product: the labelling states that it is ‘...a long-term (up to 5 years) reversible contraceptive system. The capsules should be removed by the end of the 5th year. New capsules may be inserted at that time if contraceptive protection is desired’. The ordinary meaning of the phrase ‘up to 5 years’ is that some patients may, and some may not, get 5 years of contraceptive protection. The other sentences within the above quote from the labelling do not explicitly claim that contraceptive protection continues uniformly to the end of the 5th year. Patients should, however, have an unequivocal statement of minimum duration of effective action that reaches out to include distant outliers. Furthermore, a reasonable policy would be that, since the product is capable of maintaining an annual conception rate of 0.05% per year , then the claimed duration of product efficacy should cover 99.95% of the treated population. The process of drug release should be carefully measured, and estimates made of the variance in both release rate and drug loading, from which one can estimate the effective duration of 99.95% of the implants. Retrieval of units after various periods of time since implantation is essential to be able to give adequate definition of the product's reliable lifetime. It is not clear from the information provided in the labelling that this kind of careful work was done and properly analysed. The statement that contraceptive effectiveness is provided for ‘up to 5 years’ does not inspire confidence that either the developers or the regulators understood the need for a reliable definition of the product's minimum duration of effective action for ‘all’ patients, as ‘all’ is statistically defined.
That criticism notwithstanding, the most remarkable feature of the levonorgestrel implant is its label claims for contraceptive efficacy: this product has the lowest conception rate, i.e. the highest degree of contraceptive efficacy, of any of the steroidal contraceptives. In contrast, the once-daily oral forms of norgestrel or other progestin-only contraceptives have the highest rates of conception, i.e. the lowest degree of contraceptive efficacy, of any of the steroidal contraceptive products . This difference reflects the degradation of efficacy created by variable patient compliance with the once-daily oral regimen, which of course is completely obviated by the multiyear implant. It is a stunning object lesson in the improvement of efficacy that can be achieved by converting from patients to implants as the agent of drug administration. This finding is consistent with the estimated differences in contraceptive efficacy of the combined oestrogen/progestagen oral contraceptives, which have a conception rate of 1 per thousand per year under conditions of strictly correct daily dosing versus a conception rate of 50 per thousand per year in unselected patients who presumably reflect the average range of poor, partial, good and strictly punctual compliance with the once-daily dosing regimen .
The combination of its long action without the need for daily re-medication, and its remarkable contraceptive efficacy, attracted many women to the implant when it first entered the US pharmaceutical market. Progestin-only contraception is vexed, however, by mid-cycle bleeding, which appears to have been no less frequent or profuse amongst users of the implant than amongst users of the progestin-only oral products. This factor, plus the usual reasons for early discontinuation of contraception of any kind, resulted in numerous women seeking to have the implants removed soon after they had been implanted. Many explantation procedures were difficult and lengthy. The resulting bad publicity terminated demand for the product and sales rapidly declined to low levels . One must therefore judge the product a commercial failure, but it is important to recognize one powerful lesson NORPLANT has provided: stunning efficacy.
It is difficult to escape the conclusion that, in another time, with another drug of high molar potency, and a different medical situation, the implant approach could make a major difference in the quality of ambulatory pharmacotherapy. That consideration brings us to two final topics:
1the future of implants; and
2the role that drug delivery systems can play in minimizing therapeutic, clinical and economic problems created by variable patient compliance with prescribed drug regimens.
Recent developments in the implant field: future prospects
A noteworthy recent development has been an adaptation to humans of the 23-year experience with the osmotic minipump implants for experimental animals.
These are designed as research tools, for use in experimental animals. Minipumps have three sizes: the largest is about the size of the distal two joints of the fifth finger, with a 2-mL reservoir; the intermediate is about the size of one of the larger antibiotic capsules, with a 200-µL reservoir; the smallest is about the size of the pasta called orzo, with a 100-µL reservoir. As these are products designed for researchers, who have many different needs, they are supplied with empty reservoirs, to be filled by the user with solutions or suspensions of drug or hormone as the user's research needs dictate. Minipumps are fabricated with membranes of different hydraulic conductivities, to provide an array of choices for the minimum time-period of constant-rate delivery needed to pump the contents of the reservoir: 1, 3, 7, 14, 28 or 42 days (see Table 1). Technical details and the products' extensive bibliography of over 6000 published studies based on their use can be found at http://www.alzet.com That large body of published studies spans many different areas of biomedical investigation and, in many ways, represents the vanguard of the applications of drug delivery systems throughout pharmacology, endocrinology and physiology.
Implantable osmotic pumps for human therapeutic use
The human adaptation of the osmotic minipumps has involved reconfiguration of the geometry so as to make a longer, smaller-diameter unit, suitable for implantation via a large-diameter needle. The units are designed to be prefilled with a specific agent, ready for implantation. The pumping system has been redesigned, from the collapsible polymeric bag used in the minipumps, to a piston moving along a rigid cylinder, with the piston driven osmotically by influx of water from surrounding tissues. The overall external volume of the unit is approximately 0.5 mL. The first such product, with the trademark VIADUR, has recently been approved for marketing by the FDA to deliver the GnRH analogue, leuprolide acetate, at a constant rate for a minimum of 1 year, for the purpose of inhibiting pituitary gonadotropin secretion in the management of hormonally dependent tumours, e.g. carcinoma of the prostate. The implant maintains therapeutically effective action for a minimum of 1 year without human involvement, with an agent whose plasma half-life is a fraction of a day. Further technical details on this novel product, which is illustrated in Fig. 1, can be found at http://www.viadur.com
A number of other implant products are presently in development, and it is still too soon to know the acceptance and use of the leuprolide implant. A new firm, Durect Corporation, was recently founded, dedicated solely to the development of multiple applications for implantable osmotic pumps in human therapeutics; its progress can be followed at http://www.durect.com
Whilst it is not usual practice to discuss firms in scholarly medical reviews, readers should know that delivery systems products did not come from either academia or from the major pharmaceutical firms, but rather were, with the exception of NORPLANT, pioneered by small, entrepreneurial firms that focused exclusively on delivery systems, to the exclusion of other types of therapeutic products. ALZA, Elan and Key Pharmaceuticals (later acquired by a major pharmaceutical firm) were responsible for most of the innovations in the field, although most of the resulting products are marketed by the large firms. Special mention should be made of the firm formerly known as CIBA-GEIGY which, in partnership with ALZA, entered into research and development in the delivery systems field in 1978, much earlier than any of the other large firms.
In the case of implants, which pose many, very specialized problems, three always essential ingredients for success are persistence, focus and relentless search for sound therapeutic applications. These ingredients are most likely to be maximized in a firm wholly dedicated to translating the past successes of implants in basic research to making implants an integral part of human therapeutics.
Delivery systems and patient compliance with prescribed drug regimens
The clear message from the NORPLANT story is unparalleled efficacy when drug exposure is automatically assured. That demonstration defines a clear prospect for future implant products. We cannot, however, look to long-term implants as the panacea for compliance problems, because of the special requirements for high molar potency drugs and other technical considerations that will foreseeably limit the use of implants for the indefinite future.
One can usefully look back to the seminal studies in the late 1950s by Wood, Feinstein and coworkers on antibiotic prevention of recurrent streptococcal infections and acute rheumatic fever, to see the consistent superiority of depot penicillin over many different approaches to oral regimens, as reviewed in . That early work, with monthly depot penicillin as a crude precursor of modern delivery systems, falls outside the reach of modern literature searches, but it teaches a number of important lessons about the efficacy-degrading impact of variable compliance with regimens that require daily or more frequent involvement of the patient in remedication.
A great deal has been learned about patient compliance with prescribed drug regimens in the past decade, since the advent of electronic monitoring methods for capturing dosing histories of ambulatory patients [39–41]. As noted in , the development of electronic monitoring methods is itself a spin-off from the drug delivery systems movement, for the simple reason that ‘better compliance’ has always been an inferential claim for delivery systems products that reduce dosing frequency. Sound claims, however, warrant sound data, based on sound methods of measurement, no less than any other important variable in human therapeutics.
A recurring theme in drug development during the past several decades has been the push for once-daily drug regimens, on the supposition that ‘once-a-day is best’ for patient compliance. The field of hypertension has been particularly strongly engaged in this thrust, with the result that a once-daily regimen is virtually a sine qua non for any new antihypertensive. It is, however, a triumph of propaganda over data, for the objectively measured average differences between compliance with once-daily and twice-daily regimens are so marginal, and the range of individual patients' compliance data so wide, that there is little to be said beyond the fact that about a third of patients with either regimen comply quite poorly, probably too poorly to benefit from treatment with most drugs [39–41, 43]. Moreover, the ‘drug holiday’ pattern of three or more consecutive days without any drug intake has the potential to create harm with drugs subject to rebound effects (when dosing halts abruptly) and with drugs subject to first-dose effects that may recur if dosing resumes at full strength as a holiday ends .
The ur -text for marketing claims about the superiority of once-daily regimens is a paper by Gatley  that must number amongst the most-cited clinical papers, at least for citations in drug advertisements. Yet it was thoroughly discredited a decade ago , and of course the pill-count method by which compliance was measured in the Gatley study has also been thoroughly discredited since 1989 [46, 47]. In reality, once- and twice-daily regimens are basically equivalent. It is an almost invariable finding, however, that a slightly higher percentage of prescribed doses are taken with once-daily than with twice-daily regimens, but the differences in the means are rarely more than a few percentage points, with little differences in the very wide ranges of individual patient's values. Measures of aggregate drug intake, of course, were the only thing possible when initial ideas about patient compliance were being formed. When electronic monitoring methods allowed the focus to shift to dose-timing aspects, Kruse et al. made this summary of their comparison of once- and twice-daily regimens:
Days without any dosing events were twice as [frequent] with the QD than the BID regimen. … Episodes of 3 or more subsequent days without dosing events … were also observed more often with the QD than the BID regimen. … Doses were omitted more frequently on weekends than on any other day of the week … (P < 0.001). … Evening doses were omitted about twice as often as … morning doses [in]… patients prescribed the BID regimen (P < 0.001).' 
The implication of Kruse's finding is that one must look not only at aggregate amounts of drug taken, but at the timing of doses, and make a case by case judgement about which regimen provides more patients with continuity of drug action. Moreover, each drug has to be studied in order to determine the optimum dosing sequence with which to phase back into correct dosing, when doses have been omitted, as has been done for the oral contraceptives [cf. 49]. Some patients may prefer one regimen over the other, of course, but even that preference may have no echo in reliably measured compliance with either type of regimen.
A further, rather bitter, twist is the recent recognition of how poorly our most prevalent chronic disease, hypertension, is managed in drug-treated patients – setting aside the undiagnosed and the diagnosed but untreated. The 6th Report of the Joint National Commission on High Blood Pressure in the US noted that only 27% of treated hypertensives achieve the target blood pressure of 140/90 mmHg ; comparable surveys in Canada  and in a number of western European countries show similarly low figures . These countries, of course, have widely different policies for reimbursement of health care and prescription drug costs, and they differ in their relative emphasis on primary versus specialist care. Thus, the similarity of poor outcomes in the face of these major differences indicates that the origin of these poor results is deeper-seated than who pays and who treats. The Joint National Commission concluded that pervasively poor ‘adherence’ (an alternative term for ‘compliance’) is mainly responsible for this big shortfall .
Meanwhile, technical progress in development of antihypertensive pharmaceuticals has brought us a number of products with once-daily dosing, efficacy, and minimal problems from side-effects – certainly a vast improvement over reserpine, alpha methyl dopa, and other early antihypertensives that are little used today. So somehow we have converged on antihypertensive agents with essentially ideal pharmaceutical attributes, but we are still left with major problems due to noncompliance, as Brunner pointed out in a recent interview .
One of the most important insights to emerge from recent research on the clinical pharmacological and therapeutic impact of variable patient compliance has been to focus on consequences of the most common errors in compliance, which are to omit a single dose or to omit two sequential doses. (Patients make many other errors, of course, but those two happen to be the two most common, in that order.) This new focus raises the question of how long therapeutically useful antihypertensive drug action persists after a last-taken dose of drug. In turn, that question leads to an experimental design, originally used to answer the analogous question with oral contraceptives, which is to substitute placebo dosage forms for active drug, in a controlled, suitably blinded manner, and make the necessary measurements to document how long useful drug action persists. The first such study in hypertension, carried out by Johnson and Whelton, compared two once-daily beta blockers, atenolol and betaxolol . By about the 30th hour (6 h after the placebo substitution), atenolol's antihypertensive action was fading, whereas betaxolol's action persisted beyond the 48th hour. An obvious conclusion is that a patient could occasionally omit a single dose of betaxolol with no lapse in pressure control, whereas the same error made with atenolol would result in a period of 18 or so hours of uncontrolled blood pressure. Whether betaxolol's action would persist through two sequential dose omissions is unknown, for the study did not continue past the 48th hour; in principle, the study could be repeated with measurements continuing until therapeutic action fades. Several other groups have performed similar studies, comparing other pairs of antihypertensive agents [55–7]. These studies put the focus on the margin for error in remedication as the factor much more pertinent to therapeutic success or failure than simple percentages of doses taken.
To put this matter in quantitative terms, a useful parameter, called ‘forgiveness’, is defined as the postdose duration of therapeutically effective drug action minus the prescribed interval between doses . Thus, the forgiveness of atenolol is about 30–24 = 6 h versus at least 24 h for betaxolol, from the Johnson–Whelton data. The placebo substitution protocol seems to be the best way to determine a drug's postdose duration of effective action, when ethically possible. If not possible, then one must fall back on observational studies, looking for clinical correlates of various spontaneously occurring lapses in dosing.
With this new approach, one sees that a once-daily product with minimal forgiveness requires punctual remedication for continuity of action. Yet only about 15% of patients remedicate punctually over long periods of time , so such a product is not likely to benefit other than a small minority. In contrast, dividing the dose of an unforgiving once-daily product and giving it on a twice-daily basis would make a much more forgiving product, allowing occasional doses to be missed without loss of therapeutic action. A third possibility is to reformulate the drug in delivery system form, with the aim of lengthening postdose duration of action. Obviously, these alternative approaches need to be studied in a properly controlled fashion, but the basic point is clear that simply tallying up the percentage of doses taken, without considering the timing aspects, misses a main point about variable patient compliance and its therapeutic impact.
The new focus on timing of doses and of drug actions will undoubtedly take quite a while to gain wide acceptance, but meanwhile it may help improve the quality of decision-making about drug regimens, drug regulation and opportunities for new applications of drug delivery systems.
Drug delivery systems have been developed during the past 30 years from concept, through an awkward stage where they were exotic, costly, technological ornaments, to their present status as a small but useful path of pharmaceutical innovation and a widely used method of drug administration in experimental pharmacology, endocrinology and physiology. Today's drugs were all discovered, screened and developed in dosage forms and regimens of administration that result in the familiar peak and trough pattern of drug concentrations in plasma, and their echoing peaks and troughs in drug action. Now we can confidently say that this is not the only way to give drugs or drug-candidates when we test them for therapeutic utility. Where this new capability will lead is not clear, but there are good reasons to hope that it will provide therapeutically useful agents that otherwise would have been missed.
The viewpoints in this review owe much to many discussions during the past 40 years about the actions of drugs and hormones at the biochemical, physiological or epidemiological levels, with Ernst Báràny*, Richard N. Bergman, Douwe D. Breimer, Roy C. Cooley*, Joyce A. Cramer, Claude Crescioni, James O. Davis, Bradley Efron, John W. Fara, W. H. W. Inman, Barrie R. Jones, Erik de Klerk, Ronald L. Krall, Ernst Knobil*, Wolfgang Kruse, H. G. M. Leufkens, Gerhard Levy, C. C. Li, Herman L. Marder, Jean-Michel Métry, Curtis R. Morris, Walter S. Nimmo, Carl C. Peck, Hans Petri, Virgil A. Place, L. F. Prescott, Lotte Schenkel*, Jane E. Shaw, Lewis B. Sheiner, Jos Smits, Harry Struijker-Boudier, Felix Theeuwes, Joan Urquhart, Robert Vander Stichele, Ellen Weber*, F. Eugene Yates and Alejandro Zaffaroni. Lorri Perkins and Joan Urquhart provided invaluable bibliographical help. *Deceased.
Received 18 August 2000; revision received 28 August 2000; accepted 4 September 2000.