Testing Drugs in People

by Ken Flieger
FDA Consumer special report

Most of us understand that drugs intended to treat people have to be tested in people. These tests, called clinical trials, determine if a drug is safe and effective, at what doses it works best, and what side effects it causes--information that guides health professionals and, for nonprescription drugs, consumers in the proper use of medicines.

Clinical testing isn't the only way to discover what effects drugs have on people. Unplanned but alert observation and careful scrutiny of experience can often suggest drug effects and lead to more formal study. But such observations are usually not reliable enough to serve as the basis for important, scientifically valid conclusions. Controlled clinical trials, in which results observed in patients getting the drug are compared to the results in similar patients receiving a different treatment, are the best way science has come up with to determine what a new drug really does. That's why controlled clinical trials are the only legal basis for FDA to conclude that a new drug has shown "substantial evidence of effectiveness."

Does It Work?

It's important to test drugs in the kind of people they're meant to help. It's also important to design clinical studies that ask, and answer, the right questions about investigational drugs. And that's no easy task.

The process starts with a drug sponsor, usually a pharmaceutical company, seeking to develop a new drug it hopes will find a useful and profitable place in the market. Before clinical testing begins, researchers analyze the drug's main physical and chemical properties in the laboratory and study its pharmacologic and toxic effects in laboratory animals. If the laboratory and animal study results show promise, the sponsor can apply to FDA to begin testing in people.

Once FDA has seen the sponsor's plans and a local institutional review board--a panel of scientists, ethicists, and non-scientists that oversees clinical research at medical centers throughout the country--approves the protocol for clinical trials, experienced clinical investigators give the drug to a small number of healthy volunteers or patients. These phase 1 studies assess the most common acute adverse effects and examine the size of doses that patients can take safely without a high incidence of side effects. Initial clinical studies also begin to clarify what happens to a drug in the human body--whether it's changed (metabolized), how much of it (or a metabolite) gets into the blood and various organs, how long it stays in the body, and how the body gets rid of the drug and its effects.

If phase 1 studies don't reveal major problems, such as unacceptable toxicity, the next step is to conduct a clinical study in which the drug is given to patients who have the condition it's intended to treat. Researchers then assess whether the drug has a favorable effect on the condition.

Three Phases of Testing in Humans

Usually, No Miracles

Again, the process appears straightforward--simply recruit groups of patients to participate in a clinical trial, administer the drug to those who agree to take part, and see if it helps them. Sounds easy enough, and sometimes it is. In what may be medicine's most celebrated clinical trial, Louis Pasteur treated patients exposed to rabies with an experimental anti-rabies vaccine. All the treated patients survived. Since scientists knew that untreated rabies was 100 percent fatal, it wasn't hard to conclude that Pasteur's treatment was effective.

But that was a highly unusual case. Drugs do not usually miraculously reverse fatal illness. More often they reduce the risk of death, but don't entirely eliminate it. They usually accomplish this by relieving the symptoms of the illness, such as pain, anxiety, heart failure, or angina. Or a drug may alter a clinical measurement--reduce blood pressure or lower the cholesterol level, for example--in a way that physicians hope will be valuable. Drug effects like these can be a good deal harder to detect and evaluate than a result as dramatic as Pasteur's rabies cure.

This is mainly because diseases don't follow a predictable path. Many acute illnesses or conditions--viral ailments like colds or the flu, minor injuries, insomnia--can usually be counted on to go away spontaneously without treatment. Some chronic conditions like arthritis, multiple sclerosis, depression, or asthma often follow a varying course--better for a time, then worse, then better again, usually for no apparent reason. And heart attacks and strokes, for example, have widely variable death rates depending on treatment, age, and other factors, so that the "expected" mortality for an individual patient can be hard to predict.

A further difficulty in gauging the effectiveness of an investigational drug is that in some cases measurements of disease are subjective, relying in part on what is essentially a matter of interpretation by the physician or patient. Such measurements can be imprecise, influenced by a patient's or physician's expectations or hopes. In those circumstances, it's difficult to tell whether treatment is having a favorable effect, no effect, or even an adverse effect. The way to answer this critical question about an investigational drug is to subject it to a controlled clinical trial.

New Drug Development Timeline

The phases of new drug development, from preclinical testing, research, and development through postmarketing surveillance are illustrated in a 6K PDF chart.

Understanding Controls

In a controlled trial, patients in one group receive the investigational drug. Those in a comparable group--the controls--get either no treatment at all, a placebo (an inactive substance that looks like the investigational drug), a drug known to be effective, or a different dose of the drug under study.

Usually the test and control groups are studied at the same time. In fact, usually the same group of patients is divided in two with each subgroup getting a different treatment. That is the best way to be sure the groups are similar.

In some special cases, a study uses a "historical control," in which patients given the investigational drug are compared with similar patients treated with the control drug at a different time and place. "Historical control" can also refer to a comparison of groups of patients treated at about the same time but at different institutions.

Sometimes patients are followed for a time after treatment with an investigational drug, and investigators compare their status before and after treatment. Here, too, the comparison is historical. It is based on an estimate of what would have happened without treatment. The historical control design is particularly useful when the disease being treated has high and predictable death or illness rates. Then investigators can be reasonably sure what would have happened without treatment.

It's important that treatment and control groups be as similar as possible in characteristics that can affect treatment outcome. For instance, all patients in specific groups must have the disease the drug is meant to treat or same stage of the disease. In a clinical trial of a drug to treat angina (chest pain associated with cardiovascular disease), for example, if one group of patients being studied actually had sore ribs rather than angina, their differing response to the drug could not be assumed to be due to its effectiveness or lack thereof.

Treatment and control groups should also be of similar age, weight, and general health status, and be similar in other characteristics that could affect the outcome of the study, such as other treatment being received at the same time.

Two principal methods have been used to achieve this all-important comparability. One is to carefully pair each person in the treatment group with a control patient who has closely matching characteristics. This method is rarely used today because even in the best of circumstances, it's difficult to match pairs of patients for the myriad factors that could have a bearing on results.

In the more common approach, called randomization, patients are randomly assigned to either the treatment or control group, rather than deliberately selected for one group or the other. When the study population is large enough and the criteria for participation are carefully defined, randomization yields treatment and control groups that are similar in important characteristics. Because assignment to one group or another is not under the control of the investigator, randomization also eliminates the possibility of "selection bias," the tendency to pick healthier patients to get the new treatment.

When It Helps to Be 'Blind'

In clinical trials, bias (a "tilt" in favor of a treatment) can operate like a self-fulfilling prophesy. The hope for a good outcome can skew patient selection so that the treatment group includes a disproportionate number of patients likely to do well whatever their treatment. The same kind of inadvertent bias can lead both patients and investigators to overrate positive results in the treatment group and negative findings among controls, and cause data analysts to make choices that favor treatment. Clinical trials that include such biases are likely to be incapable of assessing drug effect.

In conjunction with randomization, a design feature known as "blinding" helps ensure that bias doesn't distort the conduct of a study or the interpretation of its results. Single-blinding consists of keeping patients from knowing whether they are receiving the investigational drug or a placebo. In a double-blind study, neither the patients, the investigators, nor the data analysts know which patients got the investigational drug. Only when the closely guarded assignment code is broken to identify treatment and control patients do the people involved in the study know which is which.

Ethical Questions

Testing experimental drugs in people inevitably presents ethical questions. For example, is it ethical to give patients a placebo when effective treatment is available? Not all authorities agree on the answer. But the generally accepted practice in the United States--and one increasingly being adopted abroad--is that well and fully informed patients can consent to take part in a controlled-randomized-blinded clinical trial, even when effective therapy exists, so long as they are not denied therapy that could alter survival or prevent irreversible injury. They can voluntarily agree to accept temporary discomfort and other potential risks in order to help evaluate a new treatment.

In any trial in which a possible effect on survival is being assessed, it's important to monitor results as they emerge. That way, if a major effect is seen--positive or negative--the trial can be stopped. This happened in the first clinical study of the AIDS drug zidovudine (AZT), when a clear survival advantage for patients receiving zidovudine was seen well before the trial was scheduled to end. The trial was then ended early, and within a week FDA authorized a protocol allowing more than 4,000 patients to receive zidovudine before it was approved for marketing under the brand name Retrovir. This is an example of the ethical principle that if a lifesaving or life-extending treatment for a disease does exist, patients cannot be denied it.

In some cases, a new treatment can be compared with established treatment, so long as the effectiveness of the latter can readily be distinguished from placebo and the study is large enough to detect any important difference.

It is also possible to evaluate new drugs in this situation in "add-on" studies. In this kind of trial, all participants receive standard therapy approved for treating the disease, but those in the treatment group also get the investigational drug. The control group gets either no added treatment or placebo. Any difference in results between the treatment and control groups can be attributed to the investigational drug. It is common to study new anti-seizure drugs in this way, as well as new agents intended to reduce mortality after a heart attack.

Testing in Women and Children

In recent years there has been growing interest at FDA and by the public in drug testing in patient populations that have been relatively neglected in clinical trials, especially women and children. Children are generally not included in trials at all until the drug has been fully evaluated in adults, unless the drug is intended for a pediatric disease, such as acute lymphocytic leukemia. When children are not likely to use drugs frequently (for example, drugs to treat high blood pressure), they often have not been included in clinical trials at all. 

Without pediatric studies or other sources of scientific information, labeling cannot include guidance about dosage, side effects, and when a drug should or should not be used in children. In October 1992, FDA proposed changes in its regulations governing drug labeling for "pediatric use." The proposal is aimed at encouraging drug sponsors to develop pediatric information--through clinical trials in children or by extrapolation of findings in adults--that can be included in drug labeling.

Although both sexes now are generally represented in clinical trials in proportions that reflect gender patterns of disease, FDA and women's health advocates agree that less care has been taken to develop information about significant differences in the ways men and women respond to drugs.

A new FDA guideline on the study and evaluation of gender differences in clinical drug trials, issued in July 1993, encourages drug companies to include appropriate numbers of women in drug development programs and to pay particular attention to factors that can affect drug behavior, such as phases of the menstrual cycle, menopause, and the use of oral contraceptives or estrogens. Another focus is discovering gender-related differences in how a drug is absorbed, metabolized or excreted, and how it works.

The guideline also does away with an FDA policy dating from 1977 that excluded women of childbearing potential from participation in early clinical studies. The agency believes that institutional review boards, as well as clinical investigators and women themselves, can gauge whether women's participation in clinical trials is appropriate and make sure that fetuses are not unduly exposed to potentially toxic agents. Studying drugs in people will probably never be an exact science. But steady progress in the methodology and, in a way, the philosophy of clinical trials is making the process more productive, more reliable, and more beneficial for us all.

Ken Flieger is a writer in Washington, D.C.