gators. As we have shown, it is possible for a pharmaceutical manufacturer to develop a clinical research unit in a medical center, staffed by full-time employees. The physician responsible for the initial trial must be experienced in clinical research, and he must have an adequate staff and laboratories so that the first administration can be made with minimal risk. He would serve in the outpatient clinics or on the regular wards when not engaged in the evaluation of a new compound. This maintains physician-patient relationship and gives the investigator continuing experience in clinical medicine. It also provides an opportunity to become acquainted with a large number of people who might be willing to volunteer for participation in a clinical research project. Our own experience has proved this to be a practical and efficient procedure for the initial clinical studies of many new therapeutic agents. Another advantage to industry is that the physician-investigator employed by the pharmaceutical manufacturer may be able to evaluate several similar compounds before one is picked for further study.

The selection of a dose for the first administration of a new drug to a human subject is difficult. There is no formula for extrapolation from animal dose that which is safe for human subjects. A cardinal principle is to select the minimum amount expected to produce some therapeutic effect, on a weight basis, from results in dogs and/or monkeys, and then give only a portion of this amount as the first dose.

Sometimes normal subjects are employed as volunteers for the initial trial. They should be under constant observation by a trained investigator until all possible effects of a single dose have disappeared. In some instances the first use of the drug must be in actual treatment of disease. For example, the first use of a new antibiotic would be in the treatment of an infection which is not

At the Lilly Laboratory for Clinical Research, Marion County General Hospital.

severe or life threatening and one caused by organisms that are known to be susceptible to the new drug. Treatment with the new drug might also be undertaken when resistance to existing antibiotics was present.

After experience with a single dose has been gained from several subjects, the amount given is increased by small increments, one patient at a time. This, a slow and tedious process, must be continued until the therapeutic result occurs, a toxic symptom is detected, or the maximum practical dose (an amount which might be placed in a capsule or tablet-usually not more than 1 Gm.) has been given and neither therapeutic nor toxic manifestations occur. As has already been described, in many instances the preliminary clinical investigation can be made very satisfactorily in a clinical research unit maintained by the manufacturer of the drug.

Second phase of the clinical program. Having demonstrated a therapeutic effect in a small number of patients, a representative of the clinical research section of the pharmaceutical manufacturer, preferably the same individual who conducted the initial trial, will present the results to several experienced investigators. It is also important that these individuals be expert clinicians who are recognized for their knowledge of the disease to be studied. They must be familiar with the natural history of the disorder that will be investigated. Their studies will be intended to confirm and extend the observations made during the initial trial.

At this stage it is often possible to conduct clinical-pharmacologic tests. An example of this type of procedure is our experience with an antitussive agent. When evidence of its therapeutic effect had been obtained in a few subjects and there was some indication of the amount needed as single dose, Dr. Hylan Bickerman was asked to test the drug for its ability to control cough induced by inhalation of citric acid. He had developed a device for recording cough response in normal subjects and

alterations in the cough pattern produced by therapeutic agents." He was able to compare the new drug with codeine. This comparison would have been difficult, if not impossible, if it had been based on observations of the therapeutic effect in patients with a cough associated with pulmonary disease. From the results of this study it will be possible to plan further clinical trials.

The need to avoid bias in making comparisons of therapeutic effectiveness of a new drug with an existing substance in general use has increased the interest in the "double blind technique"-a procedure in which neither the subject nor the observer knows whether a placebo, a control preparation, or a new drug is being given. Unfortunately, as Modell and Houde have noted, it has come to be looked upon as a "cure all" for all problems in clinical research. As these authors have emphasized, however, the use of a placebo does not validate an otherwise poorly designed experiment.

In the attempt to compare a new drug with a blank and a known preparation, consideration must be given to the order of administration. The effect of a compound may continue after its administration has

stopped; the course of the disease may change during the period of study; the intensity of pain may change. These are but a few examples of the reasons why it is important to randomize the order of treatment when more than two drugs are involved in a therapeutic trial. One method for accomplishing a "planned randomization" is the use of the Latin square. This mathematical device was first described by Leonhard Euler in the eighteenth century. It is called a Latin square after Euler's practice of labeling each cell in the square with a letter. It was first employed in problems relating to agriculture, but has come to be a useful tool in clinical research. In Fig. 1 is an example of a 4 by 4 Latin square. It is so constructed that each letter appears once in every horizontal line and in each vertical column. This design provides an order for administration of four medicaments to four subjects or four groups of patients. A new drug “A," a blank “B," an active control (a known drug) “C,” and a larger dose of A, would be given for four periods. This might be one day to one year. At the end of the study every drug would have preceded and followed every other drug one time. The Latin square is particularly valuable when the criterion for a therapeutic trial is a subjective response.

When only two drugs are being compared, or when the investigator wishes to demonstrate the difference between a blank and the new drug, a "cross-over" design should be used. In this type of trial the subject serves as his own control.

Allocation of patients is also very important. When one group will serve as controls and another will receive the new drug, it is essential to randomize the distribution. This can be done by a table of random numbers or by chance determined by flipping a coin.

Occasionally the double blind technique is employed when objective responses are being measured. For example, in the investigation of the effect of beta sitosterol on serum cholesterol by Shipley," an attempt was made to eliminate any possibility

that merely taking medicine might motivate the patient to observe the prescribed diet, by giving a placebo pre- and posttreatment.

Another important purpose of the second stage of clinical evaluation of a new drug is to obtain evidence of its safety. A battery of laboratory tests is performed prior to the first dose of the drug, and they are repeated at regular intervals over a period of months. Examinations for liver and kidney function are essential, and a careful check on hematopoiesis is also conducted. It is important that the tests be performed by competent technicians. An error in these examinations could cause toxicity to be missed, or an unwarranted stigma might be placed against the drug. The latter could cause cessation of clinical studies and the loss of a valuable therapeutic agent.

When results of laboratory tests suggest some toxicity, further development of the drug will depend upon the nature of the disease for which the drug was intended. Some toxicity is allowable if the drug offers treatment for a heretofore fatal illness. If it only provides a substitute for a safe, existing therapeutic agent, however, the new compound must pass rigid tests for both safety and effectiveness.

Third phase, broad clinical trial. The need and extent of this part of the clinical trial will vary according to the nature of the drug. Although it will be less well controlled than the preceding studies, a great Ideal of valuable information can be obtained if it is properly planned and executed. It should not be considered as a "promotional gimmick." The incidence of side effects and the occurrence of reactions due to hypersensitivity cannot be determined until a larger number of patients have received a drug. Our recent experience with Carbutamide, a hypoglycemic agent for oral use in the treatment of diabetes, is an example of the value of a broad clinical trial. In this instance the results of carefully controlled studies in the management of 600 patients had been re

ported before an extensive trial was initiated. No serious untoward symptoms had been observed. After the administration of several thousand doses of Carbutamide, however, the occurrence of undesirable effects became apparent. The clinical trial was halted and further development was discontinued.

When a broad trial is undertaken it is important that the physicians be supplied with all of the pertinent information about the drug. When this stage has been reached, a definite dosage schedule has been established. It is essential that the physician be given instructions regarding the selection of suitable patients. Usually it must be emphasized that adherence to the dosage recommendations is imperative. The physician should also be given a questionnaire so that he will know in advance the information that is desired. Placebo and the double blind technique are not often necessary in this phase of clinical trial.

Reducing bias in clinical research

The term bias is derived from the word biais, which means a slope or slant. In a clinical trial program it is of utmost importance that all chances for bias, either favorable or unfavorable, be eliminated or at least reduced as much as possible. Bias may occur in several forms. First, the investigator himself is generally recognized as the source of bias. His enthusiasm or antagonism may have a positive or negative effect. As Shapiro10 has pointed out, the attitude and actions of the physician may influence the response of the patient to a new drug.

Recently Kast and Loesch conducted an experiment which illustrated the interaction between patient and medical environment. In this study consistent and deliberate variation in the physician's attitude was introduced. Twenty patients suffering from an anxiety syndrome with primary gastrointestinal complaints were observed during a period in which the expected benefit of a combination of tranquilizer and anticholinergic drugs was stressed. Two physicians

saw the patients at weekly intervals. They were careful to make favorable comments regarding progress in the presence of each patient. They were always on time; the patients were seen promptly, and everything was done to make each visit a pleasant one. During the second period the physicians continued to be solicitous and kind to the patient, but a placebo was substituted for the active drug. In the third period the patients received the same drugs, but in a capsule rather than in tablet form. They were told it was different medication. The physicians were changed and this time only one saw the patient. He was deliberately antagonistic. He was purposely late and never offered a favorable comment. During the first period 18 of 20 patients reported marked improvement. When the placebo was given, only 3 patients maintained their status. In all others complaints recurred. During the third period, of the group that had benefited during the first stage, 3 did not experience relief from anxiety and gastrointestinal complaints became worse in 5. This provides a very definite example of how the physician's behavior toward the patient can influence and alter results of a clinical trial.

Another source of bias may be the patient. Wolf13 has called attention to the hazard of using persons who refuse treatment as control subjects. In addition, normal persons who volunteer for clinical trial may represent a distinct personality pattern. Patients, in their desire to please their physician, may introduce bias. An example of cooperation by the patients and "salesmanship" on the part of the investigator is illustrated by the experiment conducted by Gruber. Eleven patients on a research ward were told that the purpose of an experiment was to change their sleep pattern. On alternate nights they were given a red capsule and were told it would make them sleep. On the next night they were given a white tablet and told it prevented sleep. Both preparations were placebos. This procedure was continued for 10 consecutive nights. Each night the patients were re

minded of the result to be expected. Next morning they were interviewed. Three subjects were very cooperative. They slept or stayed awake just as they were told. Three patients who may have consciously or subconsciously resented being "guinea pigs" responded in exactly the opposite manner from that requested.

Dr. Gruber has also demonstrated that merely increasing the dose of a placebo can influence the occurrence of side effects. A group of patients were each given one placebo capsule. Several reacted by describing side effects. And when those who had not experienced any untoward symptoms were given two capsules of the same placebo, several complained of headache, dizziness, etc. Dr. Gruber has devised a method for avoiding this source of bias when the dosage of a new drug is being determined. He uses three placebo tablets as an initial dose. This permits three increments in the amount given at one time, without the subject being aware of a change. The same result could be obtained by using capsules of the same size and appearance, but containing varying amounts of the drug to be tested.

Another source of bias can be the attitude of paramedical personnel who may talk with the patient. Enthusiasm or anlagonism of a nurse toward an experiment can influence the attitude of the patient. In fact, the inflection in the voice of the observer or the way the question is worded can be a factor in studies involving subjective response.

Dr. Gruber has employed a pegboard when evaluating an analgesic. Questions are printed on the board and beside each are holes, one for the affirmative and another for a negative reply. The nurse hands the board to the patient who is asked to place pegs in the hole after each question. It is our opinion that this simple procedure reinforces the double blind technique. Although neither the nurse nor the patient knows when an active drug or placebo is being given, nevertheless, inadvertentlymerely by the phrasing of the question-a

source of bias could be introduced which could influence the response.

Another problem in any clinical trial arises when some of the patients do not complete the series of test drug, control, and placebo. No completely satisfactory adjustment can be made for such missing data. If these subjects are excluded and not reported, the frequency of side effects as well as the number of failures or even successes may be altered. The usual reason for a person's stopping treatment is occurrence of side effects or failure to obtain benefit. Thus it is likely that the greatest number of poor responses would be found in the group that did not complete the study. The investigator should make note of this in his published report. At least, mention should be made that there were subjects who did not complete the study.

In our effort to avoid bias, we must guard against making the study so complex that sensitivity of the method is lost. Modell and Houde have pointed out that a negative conclusion is meaningless unless a control is incorporated which demonstrates a positive result.

Development in areas

other than testing

The preparation of new drugs for clinical study occupies an important place in pharmaceutical manufacturing. Sometimes it is a very simple task of merely filling capsules or making tablets. Here, as in clinical trial, unless attention is paid to details, factors can be introduced that will negate all care in execution of a clinical study. For example, stability of the drug must be determined, and it must be established for the form in which the drug will be given. Disintegration of the tablet may be an important factor. The pharmacist working in the pilot plants must be kept informed of any clinical experience so that he may work to improve the pharmaceutical form.

It is very unsatisfactory to depend on patients for information as to preference for taste of a liquid preparation. It has been found that a taste panel of volunteers, if

properly selected and using double blind technique, can give very satisfactory information as to taste preference.

As the clinical trial progresses, there are many important tasks that must be completed prior to submission of the new drug to the Food and Drug Administration. Methods and tests for control of composition of ingredients as well as standards for the final product must be developed. The package must be designed, and in some instances, as with aerosols, this can be a very important part of the clinical trial. All of these efforts must be coordinated, and this requires communication between all groups.

The selection of a suitable name may present many problems. It is no longer an easy task to find one that is not already in use.

Recording and publishing results

As in other types of experimentation, attention must be paid to the recording of results in the course of a therapeutic trial. The design of the forms or questionnaires to be used by the observers is an important part of the planning of a study for the evaluation of a new drug. When the results can be measured objectively, keeping the data is relatively easy, but when a subjective response is the basis for the assessment of the therapeutic effect, the problem of recording data can be difficult. This must be done without introducing bias.

Usually the investigator interviews the patient and records his or her description of the effects of the drug. This is often done by grading the response from 1 to 4. Since the individual who is being treated outside of the hospital may be seen only at weekly intervals, there is a chance the patient may fail to recall the events that followed the ingestion of the drug. Some investigators have found that the use of a daily report card by the patient is helpful in getting information about the subjective response. Instances have been known in which the subject completed all reports for a week while waiting to see the physician.

We have used a daily telephone call3 as a practical method of securing information.