Abuse Liability of Drugs: Testing in Humans

views updated


There is probably no drug used to treat illness that does not also pose certain risks. One such risk, generally limited to drugs that have actions on the central nervous system, is that the drug will be misused or abused because of these effects. Drugs such as these are said to have abuse potential or abuse liability. If the drugs have important therapeutic use, they may still be made available, but they will be subject to certain legal controls under various federal and state laws (see Controlled Substances Act). Over the past fifty years, a number of methods have been developed to test new drugs to determine their abuse liability, so that both the public and the medical profession can be warned about the need for appropriate caution when using certain drugs. These methods involve both testing in animals (preclinical) and testing in humans (clinical).

Several important reasons exist for why testing with humans is useful and necessary in the development of safer and more effective pharmacological agents. The research on laboratory animals demonstrating greater or lesser degrees of the abuse liability of drugs must be validated with humans; this reduces the likelihood of error in assessing potential risks. Moreover, certain self-reported changes associated with the subjective effects of medicinal drugs can be more readily evaluated in the humans for whom they were developed. Human clinical studies are also important in determining appropriate dose levels and dosage forms to ensure safety and efficacy while minimizing unwanted side effects. Finally, comprehensive and effective testing with humans helps to reduce the availability of abusable drugs to those who are likely to misuse them and to provide for the legitimate medical and scientific needs for such pharmacological agents.


One of the most important factors in drug-abuse-liability testing with humans is the way the volunteer subjects are chosen to participate in the assessment procedures. In most studies, the human volunteer subjects are experienced drug users, but wide variations exist in the nature and extent of their drug use and abuse. Some studies, for example, use students and other volunteers whose misuse and abuse of drugs has been mostly "recreational"; other studies involve people with histories of more intensive drug use and abuse over extended periods. Also, the settings in which the tests are conducted vary widely, from residential laboratory environments, where the subjects live for several days or weeks at a time, to laboratories, where the subjects do not remain in residence but continue their daily routine after drug ingestion. Variations also occur in the age of the subjects tested and the time of day that the drug is administered. Often subjects have been selected for certain human drug-abuse-liability tests on the basis of some special features (e.g., anxiety levels, level of alcohol consumption) to determine the extent to which such factors influence the outcome of the tests.

Convincing evidence now exists that many of these factorsparticularly the prior experience of the subject with respect to drug and alcohol use and misuseplay an important role in the assessment of abuse liability. The obvious value in using such subjects lies in the fact that these individuals are similar to those most likely to abuse drugs with abuse liabilityfor example, drug abusers who help determine whether a new drug has a greater or lesser chance for abuse than the one they already know. It is also important to carry out abuse-liability testing with people who, for example, do not usually abuse drugs but are light social drinkersto assess the likelihood of abuse of certain generally available medications, such as sleeping pills or appetite suppressants.


The prediction of a drug's abuse liability, based on a wide variety of testing procedures with humans, is further complicated by the fact that drugs of abuse are often used in combination and simultaneously with other pharmacological agents. This creates some very difficult problems for the testing of abuse liability, because of the large number of possible drug combinations that need to be tested and their unknown, potentially toxic, effects. While it has long been known that drugs such as Cocaine and Heroin or Marijuana and Alcohol are used simultaneously by drug abusers, few testing procedures have been developed for assessing their interactions. Even more puzzling is that some drugs with opposite effects (e.g., stimulants like the Amphetamines and depressants like the Barbiturates) are known to be used simultaneously by Polydrug Abusers, suggesting that unique subjective-effect changes may be important factors in such abuse patterns.


Based on extensive research undertaken over the past several decades, some important general principles governing abuse-liability testing with humans have been established. In the first instance, for example, a meaningful assessment requires that the test drug be compared with a drug of known abuse liability to provide a standard for evaluation. Second, the assessment procedure must involve the indicated comparison over a range of doses of both the test drug and the standard drug of abuse. This permits both a quantitative and a qualitative comparison of the drugs, while guarding against the possibility of overlooking some unique high- or low-dose effects. Third, the testing procedures should include measures of drug effects in addition to those like drug taking in the lab, which directly predict the likelihood of abuse. With these additional measures, it is often possible to obtain reliable estimates of abuse liability by comparing test drugs with a drug of abuse across a range of effects as a standard for evaluation. Fourth, confidence in conclusions regarding the abuse liability of a drug compound can be enhanced by utilizing a multiplicity of measures and experimental procedures. This is the case because our present level of knowledge in this area does not permit a firm determination of the best or most valid predictor of the likelihood of abuse. Finally, a population of test subjects with histories of drug use appears to be the most appropriate selection for predicting the likelihood of abuse of a new test drug, since this is the population who might use such a drug in that way.


The origins of assessing drug-abuse liability with humans can be found in some of the earliest writings of civilization, describing the subjective effects of naturally occurring substances, such as wine. Since the mid-nineteenth century, literary accounts of the use and misuse of opium, marijuana, and cocaine, among other substances, have emphasized their mood-altering effects and their potential for abuse. Only in recent years, however, have systematic methods for measuring such subjective effects been refined through the use of standardized questionnaires. Volunteers who are experienced drug users complete the questionnaires after they have taken a drug; their answers to the subjective-effects questionshow they feel, their likes and dislikesreadily distinguish between the various drugs and doses, as well as between drug presence or absence (i.e., placebo).

This basic subjective-effects methodology has been further refined in recent years by using a training procedure to ensure that the human volunteer can differentiate a given drug (e.g., morphine) from a placebo (i.e., nondrug). Then new drugs are tested to evaluate their similarity to the trained reference drug of abuse. This behavioral drug discrimination method permits the volunteer to compare a wide range of subjective and objective effects of abused drugs with those of new drugs. These highly reliable procedures have proven very useful in identifying drugs that may have abuse liability.

Among the most important factors in assessing abuse liability is the determination of whether humans will take the drug when it is offered to them and whether such drug taking is injurious to the individual or society. These cardinal signs of drug abuse have provided an important focus for laboratory animal self-injection experiments, but systematic studies in which humans self-administer drugs of abuse have been less common. Methods have been developed with humans, however, for comparing the behavioral and physiological changes produced by self-administration of a known drug of abuse with the changes produced by other self-administered drugs.

The measure that has proven most useful in this approach to human drug-abuse-liability assessment is the ability of a drug to reinforce and maintain self-administration behaviors much like the behaviors used to obtain food and water. Such reinforcing effects of drugs are an important determinant influencing the likelihood that a particular drug will be abused. Laboratory studies with volunteers who are experienced drug users, for example, have shown that they will perform bicycle-riding exercises to obtain doses of abused drugs (e.g., pentobarbital). There is a systematic relationship between the amount of exercise performed and the amount of drug available (i.e., higher doses and shorter intervals between doses produce more exercise behavior than lower doses and longer interdose intervals). When a placebo or a drug that is not abused (e.g., Thorazine) is made available for bicycle riding, on the other hand, the rate of self-administration declines to near zero.

Differences between drugs in abuse liability can also be assessed by determining whether humans prefer one drug of abuse to another. During a training period, for example, experienced drug users sample coded capsules containing different drugs or different doses of a drug. Then, during subsequent test sessions, they are presented with the coded capsules and allowed to choose the one containing the drug or drug dose they prefer. This "blind" procedure (i.e., the volunteers are not told what drugs the capsules contain) prevents biases that might be introduced by using the drug names. When neither the volunteer subject nor the person conducting the test knows what drug the capsules contain, the procedure is referred to as "double blind."

Not surprisingly, it has also been shown that the preference for one drug over another or one drug dose over another agrees well with ratings of "drug liking" made independently of the choice tests just described. In self-administration studies in which volunteers show a preference for one drug of abuse over another, subjective ratings of "liking" and positive mood changes were clearly more frequent for the preferred drug than for the drug chosen less often. Such self-reports have inherent limitations, however, because of variations in individual verbal skills, which make it necessary to confirm such findings with other measures.

In addition to the self-administration and subjective-effects measures of obvious value in testing the abuse liability of drugs in humans, other quantitative drug-effect measurements have proven useful. When, for example, observer ratings (e.g., nurses watching the patients) and performance tests (e.g., speed of movement) are measured after different drugs are administered to volunteers, the results can be compared to determine whether the behavioral changes produced by a test drug are the same as or different from those of a known drug of abuse. When a number of different performance tests (e.g., arithmetic calculations, memory for numbers and letters, speed of reaction) are given following such drug administration, it is possible to construct a behavioral profile showing the performance effects of different drugs. Comparisons between different drugs and test drugs with regard to the similarity of such profiles across their respective dose ranges increase confidence in assessments made of the abuse liability of unknown drugs.


Because of the availability of procedures for abuse-liability testing in humans, it seems reasonable to ask how well they work. That is, has it been possible to predict from the results of these tests whether a new drug will be abused when it becomes generally available? The two major sources available for checking the effectiveness of human abuse-liability testing procedures are case reports by clinicians of patient drug abuse, and Epidemiological surveys of large numbers of individuals as well as of specific target sites (e.g., hospital emergency rooms). Both of these approaches have many shortcomings, since they lack the precision and focus that human laboratory testing can provide. But despite the draw-backs, they can detect abuse-liability problems in both specific groups of individuals and the population at large, in a manner that has generally validated the results of human laboratory-testing procedures.


A number of codes and regulations agreed on by scientists and the lay public provide norms for the conduct of research and testing with human volunteers. In general, they require a clear statement, understandable to the volunteer, of the risks and benefits of the testing procedure, as well as an explicit consent document in written form. After it is clear that the participant thoroughly appreciates all that is involved and the potential consequences of participation, the volunteer signs the consent form in the presence of a witness who is not associated with the research. These required procedures ensure both the autonomy and the protection of volunteers for drug-abuse-liability testing.

(See also: Abuse Liability of Drugs: Testing in Animals ; Research: Animal Model )


Brady, J.V., & Lukas, S. E. (1984). Testing drugs for physical dependence potential and abuse liability. NIDA Research Monograph 52. Washington, DC: U.S. Government Printing Office.

Fischman, M. W., & Mello, N. K. (1989). Testing for abuse liability of drugs in humans. NIDA Research Monograph, 92, No. 89-1613. Washington, DC: U. S. Government Printing Office.

Joseph V. Brady