A parallel group design is a complete randomized design in which each patient receives one and only one treatment in a random fashion. Basically there are two types of parallel group design for comparative clinical trials, namely, group comparison (or parallel-group) designs and matched pairs parallel designs. The simplest group comparison parallel group design is the two-group parallel design which compares two treatments (e.g., a treatment group vs. a control group). Each treatment group usually, but not necessarily, contains approximately the same number of patients. The ICH E9 guideline "Statistical Principles for Clinical Trials" indicates that the parallel group design is the most common trial design for confirmatory trials (ICH E9,1998). An example of a three-group parallel design with a test treatment and two controls (e.g., an active control A and a placebo control B) is illustrated in Figure 5.2.1. This is a typical example of multiple controls (ICH E10, 1999). The use of this three-group parallel design with an active control and a placebo control can distinguish an ineffective drug from an ineffective design. An effective design can be determined based on the evaluation of assay sensitivity by showing the superiority of the active control over placebo. As indicated by ICH E10 guideline, this design is particularly useful when the test drug and the placebo provide similar results. This is because it provides
evidence that the test drug has little or no efficacy as compared to the placebo. On the other hand, if neither the test drug nor the active control can be distinguished from the placebo in terms of efficacy, this clinical trial is said to be lack of assay sensitivity, and hence, it does not provide evidence to conclude that the test drug is effective. Some of major advantages of a parallel group design are summarized below:
1. It is simple and easy to implement.
2. It is universally accepted.
3. It is applicable to acute conditions (e.g., infection or myocardial infarction).
4. Analysis is less complicated, and interpretation of the results is straightforward.
In addition, for ethical consideration with the control (e.g., the placebo), we can allocate patients unequally to treatment groups (in a random fashion) to allow more patients to receive the treatment (e.g., in a 2 to 1 or 3 to 1 ratio). A parallel group design is probably the most commonly used design in phases II and III of clinical trials. However, it usually requires more patients than other comparative designs.
The matched pairs parallel group design is a randomized complete block design with a block size of 2 in which each patient is matched with another of similar prognostic characteristics (e.g., obesity) for the disease under investigation. One patient in each pair is assigned the treatment, and the other receives the control. As compared to parallel group designs, matched pairs parallel group designs can reduce variability from treatment comparison. In addition a matched pairs parallel group design requires a smaller patient population. Therefore it is considered a more suitable design for progressive diseases such as cancer. However, matched pairs group designs suffer the disadvantages such that (1) the prognostic characteristics are not easily defined and (2) patient recruitment is usually slow. In practice, the selection bias for matched pairs designs is usually a concern in patient recruitment, which often limits its applications in clinical trials. Note that a matched pairs design is in fact an extreme case of stratification which is often considered to achieve balance in covariates or prognostic factors. When the number of covariates is large, the matched pairs design is difficult to implement. Hence the matched pairs design is not of practical interest in this case. Although at the planning stage it is almost impossible to identify all of the covariates that may have an impact on the disease, an unbiased estimate of the treatment effect can still be obtained by adjusting these covariates in analysis regardless of whether they are used for stratification or matching in order to achieve the balance in covariates.
In clinical trials, for a given clinical endpoint, basically there are two kinds of variability associated with the response. These two kinds of variability are known as the interpatient and intrapatient variabilities. Statistically the smaller these variabilities are, the more accurate and reliable the clinical results will be. For a parallel group design, however, these variabilities cannot be identified because each patient receives the same treatment during the entire course of the study. In other words, the observed variability for any comparisons between groups contains both interpatient and intrapatient variabilities that cannot be separated and estimated due to the nature of the parallel group design. As a result a parallel design does not provide independent estimates of the interpatient and intrapatient variabilities. In practice, a parallel group design is an appropriate design for comparative clinical trials if the interpatient variability is relatively small compared to the intrapatient variability. This is because a valid and efficient comparison between treatment is often assessed based on the intrapatient variability. Therefore, if the interpatient variability is relatively small compared to the intrapatient variability, the observed variability will be close to the intrapatient variability. In this case the parallel group design will provide a more accurate and precise assessment of the treatment difference. Other considerations for the use of parallel group designs include patient characteristics (e.g., acute or chronic and very ill or life threatening) and the nature of study medicine (e.g., potential toxicity and long elimination halflife). In some cases financial consideration may be a key factor for selecting parallel designs.
Before patients enter a clinical trial, a run-in (or lead-in) period of placebo, no active treatment, dietary control, or active maintenance therapy (e.g., diuretic and/or digoxin in heart failure studies) is usually employed prior to randomization. The inclusion of a run-in period prior to the active treatment has the following advantages:
1. It acts as a washout period to remove effects of previous therapy.
2. It can be used to obtain baseline data and to evaluate if patient fulfills study entry criteria.
3. It can be used as a training period for patients, investigators, and their staff.
4. It helps in identifying placebo responders.
5. It provides useful information regarding patient compliance.
6. It can be used to estimate and compare the magnitude of possible placebo effects between groups.
In clinical trials it is desired to have a washout period prior to an active treatment period to wear off effects of previous therapy for an unbiased and valid assessment of the study medicine. A run-in period, however, may not be suitable for patients whose conditions are acute requiring immediate treatment. It is acceptable if patients can remain without active therapy for a short period of time. In many clinical trials, it is not uncommon to observe the placebo effect for many drug products. For example, for antidepressant agents, an intensive care period may significantly improve the patients' depression without any treatment. At the end of active treatment period, it is important to determine whether the observed significant effect is due to the placebo or treatment. To eliminate the possible placebo effect, it is suggested that a run-in period be included to establish patient comparability between treatment groups at baseline, and this helps to remove placebo effect from comparison at the endpoint evaluation. In clinical trials, patients' cooperation and/or their compliance to study medicine is always a concern. A run-in period can be used as a training period for patients, investigators, and their staff. For example, if the trial requires patients to complete diary cards, a run-in period provides a training period for the patients to be familiar with the diary cards. In addition it may help in identifying uncooperative patients at an early stage and provide the necessary counseling. This information is useful in improving a patient's compliance when the study moves to the active treatment period.
Note that a run-in period is usually employed based on a single-blind fashion. In other words, the participated patients are not aware of receiving a placebo. Although the inclusion of a run-in period in clinical trials has many advantages, it increases the length of a study; consequently it often requires extra study visits. This has a direct impact on the increase of cost and potentially a decrease in enthusiasm by patients and investigators.
Examples of Parallel Group Design in Clinical Trials
During clinical development of a drug product, parallel group designs are often considered to evaluate the efficacy and safety of a monotherapy or combination therapy with other agents of the drug product. In addition a parallel group design may be used to study the dose response of a drug product. For example, consider the clinical development of Glucophage (metformin hydrochloride). Glucophage is an oral agent for the treatment of type II noninsulin-dependent diabetes mellitus (NIDDM). Although Glucophage has been on the European market for more than 20 years, it was not available for the U.S. market until it was approved by the FDA in the late December 1994. Over the past few years a number of clinical trials were conducted to further investigate the clinical pharmacology and other uses of the drug product.
To illustrate the application of parallel group designs in clinical trials, consider the clinical development of Glucophage. Table 5.2.1 lists three studies of Glucophage regarding evaluation by monotherapy, combination therapy with insulin, and dose response. For the first study (Dornan et al., 1991), the objective was to test the efficacy and tolerability of Glucophage. This study was an eight-month double-blind placebo-controlled trial of Glu-cophage monotherapy in 60 obese patients with NIDDM. This study had a typical parallel group design with a run-in period. After a dietetic review and a one-month run-in period, patients were stratified according to the levels of glycosylated hemoglobin (HbA1C) concentration and randomized to receive either Glucophage or an identical dose of placebo. The starting dose was one tablet (500 mg) daily increased at weekly intervals to three tablets daily after one month. Thereafter the dose was increased by one tablet daily at weekly intervals to a maximum of two tablets three times daily, aiming for lowering the level of fasting blood glucose less than 7mM (7mmol per liter or 126 mg per deciliter, mg/dL). Patients were fasted at the beginning and end of the run-in period, and after 1, 3, 5, and 8 months of treatment they were weighted and their blood pressure was measured. In addition, blood was taken for fasting glucose, total cholesterol, triglycerides, HbA1C, and serum insulin. The results indicated that Glucophage reduced HbA1C levels from 11.7% to 10.3%, whereas the placebo treatment resulted in a rise from 11.8% to 13.3%. The mean percent reduction in HbA1C of Glucophage is 23% lower than the placebo without weight gain. In addition the final mean fasting blood glucose level was 5.1 mM (92 mg/dL) lower on Glucophage than on the placebo. The fasting glucose level fell from 13.5 (243 mg/dL) to 10.2mM (184 mg/dL) (about 24%) on Glucophage and rose from 12.7 (229 mg/dL) to 15.3 mM (275 mg/dL) (about 17%) on the diet plus placebo. No changes or differences between groups were observed in body weight, blood pressure, C peptide,
Was this article helpful?