The Economics of Psychotropic Drug Development

Joseph A. DiMasi and Louis Lasagna

INTRODUCTION

The benefits that drugs, often in conjunction with psychotherapy or other adjunctive therapy, have brought since the 1950s to the treatment of many psychiatric disorders are substantial. It has been estimated that two to four million people in the United States are severely mentally ill (1), with nearly two-thirds of them suffering from schizophrenia (31). In the late 1950s, two-thirds of schizophrenics spent most of their lives in mental hospitals; 95% of schizophrenic patients, however, were treated on an outpatient basis in the late 1980s (2). In the United States the inpatient schizophrenic population fell by more than 400,000 from the mid 1950s, when chlorpromazine was introduced, to the late 1980s (2). While not the only factor, the development of chlorpromazine and other antipsychotics surely played a major role in the deinstitutionalization of severely mentally ill people.

Similarly, lithium has transformed the treatment of bipolar disorder, so that now the majority of manic-depressive patients can effectively cope with their condition. The introduction of tricyclic antidepressants, monoamine oxidase (MAO) inhibitors, and, more recently, the selective serotonin reuptake inhibitors has provided effective means for relieving the symptoms of many patients suffering from major depression or dysthymia.

While substantial progress has been made over the last four decades in developing new psychotropic drugs with improved side effect profiles and, to some extent, in finding expanded uses for already available drugs, relatively few breakthroughs have been achieved. Notable exceptions are clomipramine for obsessive-compulsive disorder and clozapine and risperidone for schizophrenia. This has led some to argue for increased government/industry collaboration in this area, the creation of an expert panel that would assess the potential contribution to public health of therapeutic agents that are used in other countries, or additional economic incentives to develop novel, but seemingly unprofitable, psychotropic drugs (16,17,30,43).

Clearly, some of the obstacles to the development of new drugs that would be useful to patients who are refractory to current treatments are economic. In general, new drug development is a costly, lengthy, and risky endeavor. Psychotropic drug development is no exception to the rule. The evidence on changing development costs is striking. DiMasi et al. (4) examined the research and development (R&D) costs of a group of 12 U.S. pharmaceutical firms for their new chemical entities (NCEs) that began clinical testing during 1970 to 1982. The investigators found the fully allocated cost per drug approved in the United States to be $312 million (in 1997 dollars). A comparable study by Hansen (21) covering an earlier time period (NCEs entering clinical testing during 1963 to 1975) found the average R&D cost for NCEs to be $135 million (in 1997 dollars).

Marketing approval in the United States for NCEs of the same vintage as those in the Hansen (21) sample generally occurred in the 1970s, while NCEs of the same vintage as the DiMasi et al. (4) sample generally achieved marketing approval in the 1980s and early 1990s. Thus, R&D costs more than doubled in inflation-adjusted dollars in approximately one decade. This remarkable rate of increase in the resources devoted to new drug development cannot continue indefinitely. There is, however, no evidence as yet that these costs have declined. Costs of this magnitude present imposing barriers that developers of new drugs must surmount. High R&D costs and uncertainties in the drug development process can be particularly problematic for biotechnology firms as they often depend on frequent infusions of large amounts of venture capital.

The length of the entire drug development process, from discovery of a compound to its approval for marketing, can also serve as a disincentive to innovation. Lengthier development times are generally associated with higher development costs (both out-of-pocket and in terms of lost investment income from alternative uses of the funds). In addition, the longer it takes to test a drug and get it approved, the shorter will be the period of patent exclusivity and, therefore, the less time a firm will have to recoup the large fixed costs associated with modern new drug development. DiMasi et al. (5,6) have shown how protracted the process has become for drugs approved in the United States. The mean time from synthesis to U.S. marketing approval rose from approximately 8 years for drugs approved in the mid to late 1960s to approximately 15 years for drugs approved in the 1990s (6).

The high risks faced by developers of new drugs can also inhibit innovation. The number of compounds synthesized for every one that proves to be marketable has been estimated to be in the thousands (13,20,23,46,47). Although the adoption of "rational drug design" techniques may reduce the number of compounds that are investigated, the risks are likely to remain high. Even for drugs that are judged promising enough to warrant testing in humans, substantial uncertainty remains. Only about one in five new drugs that reach the clinical testing stage in the United States will eventually be approved for marketing (4-8,42) (see also Early-Onset Mood Disorder, this volume).

In this chapter, we investigate some of the factors that affect the economics of psychotropic drug development and review what is known about the benefits and costs of treating mental illness. Specifically, in the section entitled "Discovery to Marketing: A lengthy process" we compare results on development and regulatory review times for marketed drugs in the major psychotropic drug subclasses to one another and to results for all new drugs. Data on U.S. investigational drugs are used in the section entitled "Trends in Psychotropic Drug Innovation" to quantify trends in psychotropic drug innovation and success rates for such drugs. The section entitled "The Cost of New Drug Development" presents some results on the cost of developing new psychotropic drugs. In the section entitled "The Availability of Psychotropic Drugs in the United States" we review the literature on the availability of psychotropic drugs in the United States in comparison to other countries. We discuss some of the practical problems and ethical concerns in conducting clinical research on psychotropic drugs in the section entitled "Operational and Ethical Impediments to Psychotropic Drug Development." Studies on the cost of mental illness are critically assessed in the section entitled "The Cost of Mental Illness." Finally, we summarize our findings and discuss the potential societal value of and the challenges facing psychotropic drug development in the section entitled "Conclusions."

DISCOVERY TO MARKETING: A LENGTHY PROCESS

The length of the development process for drugs as a whole has been well documented (4-6,9,27). The literature, though, lacks specifics about psychotropic drugs. The Tufts Center for the Study of Drug Development (CSDD) maintains a database of new chemical entities (NCEs) that have received U.S. marketing approval since 1963. The database contains both publicly available and proprietary information obtained from surveys of pharmaceutical firms. We use information in this database to determine average development and regulatory review times for psychotropic drugs and compare them to similar data for all new drugs.

For the purposes of this chapter, an NCE is defined as a new molecular compound not previously tested in humans. Excluded are new salts and esters of existing compounds, diagnostics, and biologics. We confined our analysis of psychotropic drugs to the NCEs in the antianxiety, antipsychotic, sedative/hypnotic, and miscellaneous psychotherapeutic categories in Drug Facts and Comparisons (1997). Information on milestones in the development process were obtained from the CSDD database for 51 psychotropic NCEs approved in the United States from 1963 through 1997. Table 1 lists those drugs by subclass. For each drug it also shows the date on which a new drug application (NDA) was submitted to the Food and Drug Administration (FDA) for marketing approval, the date of NDA approval, and the review time for the drug, defined as the time from NDA submission to NDA approval.

The period of analysis (1963-1997) is long and studies have found increases in the length of the new drug development process over this time frame (5,6,9). Proportionately more of the psychotropics have been approved in the first 17 years of the 35-year study period than is the case for all NCEs (47% versus 37%). As a class, psychotropics have had development time trends that are qualitatively similar to those of NCEs in general. Under the assumption then that the trends reflect meaningful changes over time in the development process, comparison of psychotropic drug development times to those of all NCEs over the study period may understate the extent to which psychotropic development times exceed the overall averages for recent development.

The psychotropic subclasses differ, however, in terms of how their approvals are distributed over time. The antianxiety and antipsychotic approvals are concentrated in the first half of the study period, while the antidepressants and sedative/hypnotics have been approved predominantly in the second half of the period. The number of approvals for the subclasses in the two periods, though, tend to be small, thus limiting the ability to do meaningful analysis by subperiod. We do, however, report results in the text for more recent approvals when the results differ substantially from those for the whole period. In addition, a regression analysis was conducted that allows comparisons of development and review times for psychotropics and NCEs in other therapeutic classes, while controlling for the year of approval.

Clinical development and regulatory review

Mean clinical phase times for psychotropic subclasses and for non-psychotropic NCEs in the CSDD database are shown in Figure 1. We report results only for those NCEs for which we have complete phase data. Phase length is defined as the time from the start of testing in the phase to the start of the next phase. These phase lengths may overestimate or underestimate actual phase testing times. Testing in a phase may end before the next phase begins or it may overlap with the next phase. DiMasi et al. (4) found small differences, on average, between phase lengths defined in this way and actual phase lengths for Phases I and II. However, Phase III extended, on average, six months into the NDA review period. Thus, the Phase III times in Figure 1 may underestimate the actual amount of time spent in Phase III clinical testing.

Mean Phase I time for psychotropics for the whole period (13.7 months) is somewhat higher (2.0 months) than the average for all other NCEs. Average Phase I time for the last 18 years of the study period, however, is 4.2 months longer for psychotropics. Among the subclasses, only the antidepressant category has longer than average Phase I times for both halves of the study period. The psychotropics tended to spend much longer in testing than do other NCEs in Phases II and III (27.5 and 37.8 months versus 19.6 and 29.6 months, respectively), where efficacy is explored and established. For the most part, however, this is a recent phenomenon. Phase II and Phase III mean times for psychotropic drugs are 0.8 and 2.6 months longer than for all other NCEs over the first 17 years, respectively. For the last 18 years, however, mean Phase II time is 11.9 months longer than for all other NCEs and mean Phase III time is 11.0 months longer.

We have more data on investigational new drug application (IND) and NDA filing dates than we do on phase starting dates. Clinical development time can also be measured as the time from IND filing to NDA submission. We define this period to be the IND phase. Similarly, we measure the regulatory review period as the time from NDA submission to NDA approval (NDA phase). Mean IND phases for the psychotropic subclasses and for all NCEs are shown in Figure 2, while mean NDA phases are shown in Figure 3. Medians, ranges, and sample sizes are shown in Table 2.

Clinical development periods (Figure 2) tend to be longer for psychotropics than for other NCEs (a mean of 74.1 months compared to 65.5 months for 1963 to 1997). This is especially true for the more recent approvals. The mean IND phase for psychotropics approved during 1980 to 1997 is 95.8 months, compared to 73.7 months for all other NCEs. Over the whole study period clinical development times for the psychotropic subclasses tend to be longer than those for NCEs in general. For the last half of the study period, all psychotropic subclasses have mean clinical development times that are much longer than for all NCEs.

The regulatory review process for new drugs in the United States has undergone a substantial transformation in recent years. The Prescription Drug User Fee Act of 1992 (User Fee Act) authorized the FDA to collect fees from sponsors when marketing applications are submitted for to the agency for review (28,39). The fees have been earmarked to increase FDA reviewing staff. The additional resources and performance goals for the FDA that were linked to the Act have been associated with a substantial shortening of regulatory review phases in the United States (10,28). The legislation was enacted in late 1992 and implemented in 1993. There are certain exceptions to the user fee program and the fees and goals applied to new submissions. Consequently, the annual share of new drug approvals that have been part of the user fee program has grown steadily from 1993 to the present. For example, only 22% of new molecular entity approvals granted by the FDA’s Center for Drug Evaluation and Research in 1993 were subject to the user fee program. However, 67% of the 1994 approvals, 85% of the 1996 approvals, 91% of the 1996 approvals, and 93% of the 1997 approvals were subject to the user fee program. Thus, when analyzing regulatory review times it is informative to consider the period 1994 to 1997 separately from earlier periods. The Act had a five-year horizon, but it has since been reauthorized along with new legislation enacted in late 1997 with the intent to, in part, further improve the efficiency of the regulatory review process and to also help shorten development times (Food and Drug Administration Modernization Act of 1997).

For the 1963 to 1993 period, each of the psychotropic subclasses had average regulatory review times that were longer than for all other NCEs (Figure 3). The Use Fee Act has had the effect of shortening approval times generally and compressing differences across therapeutic classes. The average approval time for the four psychotropic drugs that have been approved during 1994 to 1997 is 20.6 months, essentially the same as the average for all other NCEs approved during this period.

Preclinical and total development time

While psychotropics appear to have, on average, longer clinical testing periods than do other NCEs, the evidence does not suggest the same for preclinical testing. Our measure of the length of preclinical testing is the time from the first tests for pharmacologic activity to the first human tests. Figure 4 shows average preclinical time, measured in this way, for psychotropic subclasses and for all other NCEs (medians and ranges are given in Table 3). Although mean preclinical time is slightly longer for the antipsychotics than for all NCEs, the other subclasses, and psychotropics as a whole, tend to move from first pharmacologic testing to human testing quicker than do other NCEs. Mean preclinical time for psychotropics (30.0 months) is 9.5 months shorter for all other NCEs. The differences in average preclinical time between psychotropics and all NCEs are nearly identical for both halves of the study period.

Development costs, and so the economic viability of new drug development, depend on the entire period from discovery of a compound to its marketing approval. We show the mean time from synthesis of a new drug to its approval for marketing in the United States for psychotropic subclasses and for all other NCEs in Figure 5 (medians and ranges are given in Table 3). For the whole study period, mean synthesis-to-approval time is moderately longer (5.6 months) for psychotropics than for all other NCEs. However, for the last half of the study period, the average time from synthesis to approval was more than two years longer for psychotropic drugs than for all NCEs (16.4 years versus 14.2 years). For the first half of the study period, the psychotropic subclasses had below average synthesis-to-approval times. Thus, psychotropic total development time has increased markedly in relation to other NCEs in more recent years.

Regression analysis

As noted above, the new drug development process has tended to lengthen over time. For a thorough analysis of development and regulatory review times of a class of drugs, we should, therefore, allow for trends in the data. There may also be other characteristics of new drugs that should be accounted for in isolating the length of the process for a group of drugs. Regression analysis allows us to control for multiple factors that may affect the length of the development or regulatory review periods. We use regression analysis to evaluate the major components of the process - the NDA phase, the IND phase, and the time from synthesis to marketing approval.

One characteristic that may partially explain development and review times is the therapeutic rating that the drug receives from the Food and Drug Administration (FDA). From 1976 to 1991 the FDA gave NCEs thought to represent a significant gain over existing therapy a 1A rating, a modest gain over existing therapy a 1B rating, and little or no gain over existing therapy a 1C rating. The FDA retroactively rated NCEs approved during 1963 to 1975 in accordance with this rating scheme. Since 1992 the FDA has simplified its rating scheme - new drugs receive either a priority (1P) or a standard (1S) rating. For the regression analysis, we created a dummy variable (RATING) that takes on the value one if the NCE has a 1A, 1B or 1P rating and zero otherwise.

The regression format allows us to compare psychotropic development and review times to those for other therapeutic classes. We formed dummy variables for major therapeutic categories. Specifically, we use the variables ANINF for antiinfectives, ANALG/ANEST for analgesic-anesthetics, ANTINEO for antineoplastic NCEs, CARDIO for cardiovascular NCEs, ENDO for endocrine NCEs, GI for gastrointestinal NCEs, OCNS for central nervous system NCEs that are not psychotropics, RESP for respiratory NCEs, and MISC for a miscellaneous category of NCEs. To capture a trend in the data independent of therapeutic rating and therapeutic category, the IND and synthesis to approval regressions also include the explanatory variable YEAR - the year in which the NCE was approved. Since the User Fee Act substantially altered the regulatory review process in the United States by providing increased resources and establishing performance goals for the FDA, the NDA phase regression contains a period dummy variable. The variable Y9497 takes on the value one if the NDA was approved during 1994 to 1997 and zero otherwise.

Table 4 shows ordinary least squares regression estimates for the NDA phase, the IND phase, and synthesis-to-approval time for the period 1963 to 1997. The regressions support the hypothesis of increasing trends in development times, independent of shifts over time in therapeutic classes or therapeutic significance. The year of approval is positively and significantly related to the length of each of the phases. The coefficients of the YEAR variable imply that, other things being equal, there was an upward trend of 1.7 months per year for the IND phase 3.0 months per year for the synthesis-to-approval phase. The coefficient of Y9497 in the NDA phase regression implies that, other things being equal, regulatory review times declined by slightly more than one year from 1994 onward.

The coefficients of the therapeutic rating variable are statistically significant. As has been suggested elsewhere (9,10,26), the results indicate that NDA review times are shorter and that development times are longer for priority-rated NCEs. The shorter review times support the notion that the FDA has been successful in evaluating more expeditiously drugs that it believes represent significant gains over existing therapy. The longer development times could be explained, in part, if priority-rated NCEs are more often singular in their therapeutic effects or pharmacologic mechanisms of action than are standard-rated NCEs.

The psychotropic class is the omitted therapeutic category in the regressions in Table 4. Thus, the coefficients of the therapeutic class variables can be interpreted as the differences in review or development times between the class in question and the psychotropic class. With a small number of exceptions, the therapeutic class coefficients are negative and the majority are statistically significant. None of the positive coefficients are statistically significant. Thus the results indicate that psychotropic drug development and regulatory review are lengthier than for many other classes, with no evidence that they are shorter than those of any of the other classes. The data are too thin to use the regression format to analyze differences with and among psychotropic subclasses.

Separate regressions using only data from the last half of the study period support the observations made above about review and development times for psychotropic drugs in relation to other NCEs for this subperiod. The therapeutic class coefficients for these regressions are much larger in absolute value than they are in Table 4. For example, the results in Table 4 imply that for the whole study period psychotropics took, other things being equal, 16 months longer than antiinfectives to be reviewed, 36 months longer than antiinfectives in clinical testing up to NDA submission, and 38 months longer than antiinfectives from synthesis to marketing approval. The regressions for the last half of the study period, however, indicate that the differences between psychotropics and antiinfectives for 1980 to 1997 approvals are 21 months for review, 47 months for the IND phase, and 55 months from synthesis to approval.

TRENDS IN PSYCHOTROPIC DRUG INNOVATION

The final output of pharmaceutical firm new drug development can be measured by the number of NCE approvals. If, however, we examine investigational NCEs, then we can also glimpse where firms had invested their hopes and we can form expectations about how much innovation is likely in the near future. The CSDD maintains a database of investigational NCEs. We examined this database for trends in the number and kinds of psychotropic NCEs entering clinical testing in the United States.

The CSDD database contains information on commercial IND filings for NCEs by 32 U.S. pharmaceutical firms during the period 1963 to 1994 and by additional firms for portions of this period. We examined trends in the extent of psychotropic drug innovation for the 32 firms with complete data over the first three decades of this period . In the first decade these firms filed INDs on 560 NCEs. The number of IND filings declined 19% to 452 in the second decade, and then increased 13% to 515 for the last decade. As can be seen in Figure 6, the pattern was different for psychotropics. While there also was a decline for psychotropics from the first decade to the second, it was proportionately greater (28%) than was the case for all NCEs. Additionally, the number of psychotropic NCEs tested continued to decline for the last decade (9%). One factor that may have affected the incentive to pursue psychotropic drug development is the lengthening of the development process over time in relative as well as absolute terms, as noted above.

The IND filing trends, though, differ for psychotropic drug subclasses. The largest group of psychotropics investigated is the antidepressant category. The extent of the decline in activity from the first decade to the third for antidepressant development is similar to that for psychotropics as a whole. However, activity on anxiolytics and antipsychotics increased in the last decade analyzed and equaled or exceeded activity in the first decade, while activity on sedative/hypnotics dropped dramatically over time.

The disparate patterns for the subclasses may reflect, in part, differing scientific and economic opportunities. In particular, the developmental and commercial success of clozapine and a potentially large market for drugs to treat schizophrenia may have spurred development of antipsychotics. For example, the world market for drugs to treat schizophrenia is forecast to be valued at $2 billion (1992 dollars) in 2002 (37). The use of neuroleptics for psychoses is expected to increase 2.8% per year from 1992 to 2002, with most of the growth fueled by the introduction of clozapine in markets where it has not been launched and the approval of new drugs (dopamine antagonists, serotonin antagonists, and dual action serotonin-dopamine antagonists).

Success rates for psychotropic clinical development

Output from the new drug development process depends not only on the number of drugs investigated, but also on the probability that an investigational drug will be approved. We used the CSDD database on investigational NCEs to examine clinical success rates for psychotropics and for all NCEs. A clinical success rate is defined as the percentage of NCEs with INDs filed that are given U.S. marketing approval.

Figure 7 shows clinical success rates (as of October 31, 1998) for self-originated psychotropics and for all other self-originated NCEs for five IND filing periods. Success rates for psychotropics have declined over time, both in absolute terms and relative to other NCEs. Psychotropic success rates are clearly higher than those for all other NCEs for the 1960s filings. None of the psychotropics from this period are still in active testing. Success rates for the two groups for the early 1970s are nearly identical. None of the psychotropics with INDs filed during the 1960s and 1970s are still in active testing, so the success rates shown for the first three filing intervals are final. If all of the investigational psychotropic compounds that are still in active testing eventually are approved for marketing, then the success rate for 1980 to 1984 filings would be 15.4% and the success rate for 1985 to 1989 filings would be 20.0%. Many of the other NCEs with filings in these periods are still active and thus the final success rates for other NCEs could also be higher than shown in Figure 7.

The clinical success rates for late 1970s and early 1980s psychotropic filings are well below average. For the late 1980s filings, the pyschotropics success rate currently also falls short of the success rate for all NCEs, but with many compounds from this period still active and with a tendency for psychotropic drugs to have longer than average development times conclusive results for the late 1980s filings cannot be obtained at this time.

It is not clear why psychotropic clinical success rates declined over time. However, what is certain is that, other things being equal, a lower clinical success rate does imply a higher development cost per approved drug. The reason is that a lower success rate means that the costs of research failures are greater for every drug that does get approved.

THE COST OF NEW DRUG DEVELOPMENT

As noted above, DiMasi et al. (4) estimated average R&D costs for sample of NCEs that were first tested in humans during 1970 to 1982. The costs of research failures and income foregone from investing in development for a period before any returns are earned (time costs) were included. Analyses in DiMasi et al. (7) of clinical phase costs for neuropharmacologic and all NCEs using data from this sample are shown in Figure 8. The neuropharmacologic class includes more than psychotropics. Small sample sizes precluded further decomposition of costs by therapeutic category.

Phase III costs for neuropharmacologics are about average. Costs for the earlier phases and animal testing done during the clinical period, though, are lower for neuropharmacologics. Each phase cost includes the costs of every drug that entered the phase, whether the drug was eventually abandoned or approved. The fully allocated cost per approved drug, though, depends critically on success rates, phase attrition rates, and the time spent in the various testing phases.

The clinical period out-of-pocket, time, and total costs per approved NCE for neuropharmacologics and for all NCEs in the sample are shown in Figure 9. Lower than average success rates and longer development and regulatory review times for neuropharmacologics account for the fact that total cost per approved neuropharmacologic is 11% above average. The regressions presented above suggest that clinical testing and regulatory review have been longer for psychotropics than for other central nervous system drugs. Thus, the average cost of developing psychotropic drugs may be even higher than the average cost for all neuropharmacologics.

THE AVAILABILITY OF PSYCHOTROPIC DRUGS IN THE UNITED STATES

Psychopharmacological innovations are not disseminated to all countries at even roughly the same time. In some instances, the lag between the availability of a psychotropic drug in two countries is substantial. A number of studies have specifically examined the availability of new drugs in the United States in relation to other industrialized countries. Wardell (44) documented a U.S. lag with respect to the United Kingdom in the availability of new drugs introduced in either market during 1962 to 1971. He found that, of 28 psychotropic NCEs approved in either country, 13 were exclusively available in the United Kingdom, while only four were exclusively available in the United States.

Later studies updated the Wardell analysis for more recent periods. Wardell (45) examined NCEs approved in the United States and the United Kingdom during 1972 to 1976. Of the 14 psychotropic NCEs made available during this period, eight were exclusively available in the United Kingdom and three were exclusively available in the United States. The most recent update in this series (25) found that, of 29 psychotropic NCEs made available during 1977 to 1987, 12 were exclusively available in the United Kingdom and five were exclusively available in the United States. Of the 12 psychotropic NCEs that were mutually available during this period, nine had been introduced in the United Kingdom first.

More recently, Kessler et al. (29) of the FDA compared the availability in four countries (United States, Germany, Japan, United Kingdom) of drugs introduced onto the world market from 1990 to 1994. The therapeutic value of these drugs was assessed based on the FDA’s therapeutic rating system and the expert judgment of the study’s authors. Overall, the availability of drugs that entered the United States market and the markets of at least one of the other three countries was obtained more quickly in the United States than in Germany and Japan. The speed with which drugs entered the U.S. and U.K. markets was similar.

Moreover, in two-country comparisons of drugs that were available in one country but not the other, the authors found that the United States had more exclusively available important drugs when compared to each of the other countries. However, when we examined the drugs in the Kessler et al. (29) sample that were available in the United States and not in one of the other three countries (as of April 1995) we found few psychotropic drugs. Of the 18 drugs that were available in the United States but not the United Kingdom, none are psychotropics. Additionally, of the 31 drugs that were available in the United States but not in Germany, three are psychotropics (nefazodone, sertraline, and venlafaxine). Finally, of the 62 drugs that were available in the United States but not in Japan, five are psychotropics (nefazodone, paroxetine, risperidone, sertraline, and venlafaxine). Among the psychotropics that were available in the United States and not in at least one of the other three countries, the United States was the first market worldwide for only one (venlafaxine).

The availability of psychotropic drugs was the focus of two recent studies. Vinar et al. (43) surveyed clinical pharmacologists about the usefulness of 50 psychotropic drugs that were either marketed or in late stage clinical trials in Europe, but not available in the United States, for therapy or as a research tool. Twenty of these drugs were judged to be of particular interest after experts rated the drugs on the basis of novelty of mechanism of action and probable safety in comparison with drugs available in the United States. Some of the drugs have been available in Europe for decades. As of the end of October 1998, none of these drugs had been approved in the United States.

Many of the drugs in the Vinar et al. (43) list have lost, or are close to losing, patent protection. They therefore likely have limited economic potential. Vinar et al. argue for the creation of a board of pharmacological experts that would make recommendations to the FDA about foreign drugs that should be considered for accelerated regulatory review and, like compounds now given orphan drug designation, be granted a period of marketing exclusivity.

Glick et al. (17) examined the controlled clinical trial literature and surveyed 61 expert clinicians (from six West European and two East European countries), European regulatory authorities, and representatives of the pharmaceutical industry to identify psychotropic drugs with apparent advantages over conventional treatments that are available in Europe but not in the United States. The authors identified 12 compounds, eight of which had been approved in three or more of six countries whose approvals were examined. As of this writing, none of the 12 drugs have been approved for marketing in the United States. As in Vinar et al. (43) the authors recommend that a board of experts be established in the United States that would be empowered to extend marketing exclusivity for promising compounds (if approved by the FDA) that are available abroad. Additionally, for compounds that have lost patent protection they suggest government support of clinical trials in academic centers.

Many of the psychotropic drugs that do make it to the U.S. market have had lengthy periods of prior foreign marketing. The CSDD database on NCEs approved in the United States contains information on the country in which the NCE was first marketed and the date of first marketing. Figure 10 shows how psychotropic NCEs approved in the United States are distributed according to the number of years of prior foreign marketing. The United States was the first market for only 29% of the drugs. The proportions of drugs approved first in the United States or within one year of approval in the United States are similar for all NCEs as for psychotropic drugs (33% approved in the United States first and 15% approved within one year of first foreign marketing for all NCEs). Proportionately many more of the psychotropic drugs that have been approved in the United States, however, have been available in a foreign market for a very long time. Only 10% of all NCEs have six to ten years of prior marketing, and 9% of all NCEs have been marketed in another country for more than ten years prior to U.S. approval. The lengthiest periods of prior foreign marketing for the psychotropics were for some of the most innovative drugs in this class. Clozapine was available in a foreign market (Switzerland) 17 years prior to U.S. approval, and clomipramine was first available 20 years before U.S. approval (Finland).

OPERATIONAL AND ETHICAL IMPEDIMENTS TO PSYCHOTROPIC DRUG DEVELOPMENT

As shown above, clinical development of psychotropic drugs is a lengthy and costly process. Significant improvements in the process can therefore pay off handsomely for society as well as for developers. There are a number of impediments to efficient and effective clinical development of psychotropic drugs that are worth noting. The problems include the allocation and adequacy of resources, special patient population characteristics, proper trial design, legal considerations, and ethical concerns.

For example, schizophrenic patients are often treated in a number of settings depending on the severity and the stage of the illness. Thus, the development of antipsychotics often requires extensive coordination and capabilities across varied facilities. A Phase II multicenter trial for an antipsychotic compound could, for example, easily require 20 sites (24).

The impaired cognitive capabilities of schizophrenic patients also present some difficulties for clinical research. The accuracy of symptoms reported by psychotic patients is questionable. Investigators, however, must often rely on these subjective reports. The accuracy of the reports may improve after treatment, but this makes difficult the establishment of a baseline. In some cases, it may be important to recruit subjects who have not been previously treated. Results from animal studies, for example, have suggested that the effects of D₁ dopamine antagonists might differ according to whether the subject had been pretreated with a D₂ dopamine antagonist (24). Finally, the inclination of some psychotic patients to cooperate with clinical investigators or their capacity to give truly informed consent can be affected by their condition.

Klein (31) has identified a number of potentially correctable problems associated with Phase II trials for many psychotropic drugs. These include selecting incorrect target populations, study periods that are too short to detect beneficial effects, recruitment difficulties, and ignoring drug withdrawal problems. Klein offers a number of suggestions to ameliorate these and other problems that would, hopefully, lead to more useful Phase III trials.

Finally, for almost two decades, certain kinds of research were greatly hampered by an FDA pronouncement that women of child-bearing potential should not be exposed to a drug candidate until some evidence of safety and efficacy was in hand from the study of other populations. The basis for this regulation was the residual anxieties that followed the thalidomide teratogenicity tragedies of the 1960s. This restriction was a serious one in the case, e.g. of antidepressants, where women of child-bearing potential represent the largest experimental population. The 1993 revised status of this population has seemingly eliminated this problem.

THE COST OF MENTAL ILLNESS

The potential for psychotropic drugs to relieve pain and suffering, and even to reduce the economic burden of mental illness on society, is substantial. While pain and suffering are difficult to assess quantitatively, we can more easily measure the economic costs to society of mental illness in general, or of specific psychiatric disorders in particular. We review here the primary methodological approaches that have been used to measure the cost of illness and discuss the results of some of the more recent studies of the cost of mental illness.

Cost of illness studies have typically attempted to measure both direct costs of treatment (payments made) and indirect costs (resources lost). Two basic methodological approaches to valuing human life have been used. These are the human capital approach, which measures an individual's lost productivity due to morbidity or mortality, and the willingness-to-pay approach, which measures what an individual would be willing to pay to avoid an increased risk of death or morbidity. Each of these approaches has conceptual problems. There have also been attempts, though, to combine the two approaches (33).

The human capital approach only counts economic loss in the marketplace. It excludes the value to individuals of reduced pain and suffering, changes in leisure time, and changes in risk per se. Typically, indirect costs of disease are measured partly on the basis of expected future lifetime earnings. This biases downward cost estimates for children, the elderly, and unpaid household workers. The approach is also problematic if there are imperfections in labor markets. The value of the output of some individuals will be underestimated if, for example, wage discrimination on the basis of race or gender exist. For these and other reasons, many analysts have rejected the human capital approach. It still may be useful, though, as long as one realizes what it does and does not measure.

The willingness-to-pay approach has a conceptual basis in economic theory. Life is valued according to what individuals would be willing to pay to reduce the probability of illness. The results of willingness-to-pay analyses, though, depend on the income distribution of the individuals who would be affected by the illness. The approach also has substantive measurement problems. Typically, individuals are asked hypothetical questions about valuing very small reductions in the probability of death. In addition, the approach is usually operationally very difficult to undertake. The relative ease with which human capital studies can be conducted has helped make it the most commonly used methodology in cost of illness studies.

Seminal work on the economic cost of mental illness was done by Fein (14), who estimated the direct and indirect costs of mental illness to be at least $2.4 billion in the mid 1950s. Since then, the trend in estimates of the cost of mental illness has been decidedly upward. The estimates actually fluctuate quite a bit, but this is undoubtedly due to differences in how and what costs were included (35).

Table 5 shows some of the more recent estimates of the cost of various illnesses. The costs of mental illness are roughly on a par with those of such major diseases as cancer and cardiovascular disease. Costs for depression and schizophrenia are similar to those for arthritis and coronary heart disease. A major portion of the costs of mental illness are attributable to indirect costs. For example, the indirect costs of depression in Greenberg et al.(19) account for 72% of the total cost of $44 billion. In fact, one of the indirect costs counted in Greenberg et al. (19) accounts for a significant portion of the difference in total cost between that study and the Stoudemire et al. (41) study. Greenberg et al. included an estimate of the reductions in productive capacity (20%) of depressed individuals while at work during episodes of depression. While quality-of-life changes are not measured in these studies, these are undoubtedly important costs of depression and other mental illnesses (see Tic Disorders, this volume).

CONCLUSIONS

New drug development is a complex and uncertain process that depends on a mix of economic factors, regulation, the progress of science, and serendipity. The economic considerations include not only the R&D costs that are paid out-of-pocket, but also the length of the development process and the returns that can be expected from successful development. The data presented here show that psychotropic drug development in the United States has been lengthier than average, particularly for clinical development and regulatory review. We also found clinical development to be riskier than average, in the sense that proportionately fewer psychotropic drugs investigated clinically make it to marketing approval than is the case for investigational drugs as a whole. Both of these factors contribute to above average clinical development costs for psychotropics. Longer than average development times also tend to reduce the period of patent protection.

While the costs and risks of psychotropic drug development are substantial, the markets for new psychotropic drugs that are demonstrably more effective are potentially lucrative. Even drugs that offer markedly improved side effect profiles can pay off well. Markets for psychotropic drugs have been growing strongly. The world market for psychotropics has expanded at a 17% compound annual rate of growth from $2 billion in 1986 to $4.4 billion in 1991 (32). This growth was led by antidepressants, with sales that grew at a 42% compound annual rate during this period. The introduction of fluoxetine and other selective serotonin re-uptake inhibitors accounted for much of this growth.

The considerable potential value to society of effective psychotropic drugs is clearly illustrated by the economic costs to society of depression, schizophrenia, and other mental illnesses that some studies have demonstrated. In an era when cost containment is a widespread objective, developers will have to make convincing cases for the kinds of reimbursement levels that are needed to make further development economically worthwhile. At least in the published literature, studies examining the cost-effectiveness of psychotropic drugs have been relatively few in number. One exception is clozapine, where recent published studies have shown it to be cost-effective for treatment-resistant schizophrenia when used for long-term maintenance therapy, despite prices that some argue are unjustifiably high (15,34). In the current economic environment, developers will also find it increasingly necessary to find ways to make the psychotropic development process more efficient and fruitful.

If the unmet needs of psychiatric patients are to be met with new drugs or new uses for already marketed drugs, industry, academic, and government researchers must explore new avenues vigilantly. The search for new psychotropic drugs has invited a number of strategies. The most successful, in fact, has been the serendipitous discovery of useful therapeutic activity by astute clinical observers administering NCEs to psychiatrically ill patients (11). A second approach has been to follow up on serendipitous discoveries by molecular modification of first generation discoveries. A third approach is to devise animal models, usually constructed so that they would have identified standard drugs. The latter two techniques have the theoretical limitation that they may propose as candidates drugs that are only marginally different from already available medications. The more recent attempt to identify drug receptors as a basis for selecting new candidates is conceptually exciting but awaits empiric testing to confirm or reject biotechnological guesses and the current optimism about this approach.

Since there is general agreement that predicting psychotropic activity with precision remains a difficult challenge, the drug developer is left with an unattractive scenario: a potentially valuable drug will be expensive to test in numerous different clinical settings, but picking "the" correct type of patient in advance is difficult if not impossible. Some have proposed that we go back to the practices of earlier days, and let experienced clinicians expose small numbers of patients with different psychiatric problems to a drug candidate in "open" clinical trials rather than conduct formal controlled clinical trials. Opponents of this view argue that false positives will be frequently encountered with the "open" proposition and lead to wasted resources. The solution to this dilemma is not apparent.

published 2000