Evaluating Safety and Efficacy

To approve a drug for marketed use in the United States, the FDA requires conclusive evidence that a drug is safe and effective for its intended use. As a general rule, data from at least two pivotal Phase III trials are required for regulatory decisions, although in some circumstances, for example in the case of a life-threatening disease such as cancer, only one trial might be required (when it would be considered unethical to withhold an effective treatment from patients and conduct a second trial). The concern is that results of one trial may represent a chance occurrence and should thus be confirmed by a second trial. Earlier trials, that is, Phase I and II, cannot provide data on which to base regulatory or clinical decisions. The FDA notes that each of these pivotal trials must be “adequate and well-controlled” and describes common characteristics of study designs that are considered adequate and well-controlled (Table 90-2).⁸⁵ Specific comments about trial design, especially Phase III trials, and these characteristics are made later in this chapter.

Trial I Error Rate Control.

A type I error is the error that occurs when one rejects the null hypothesis when it is true (i.e., rejecting the null hypothesis that there is no effect of the treatment or drug in question and concluding that there is an effect). The probability of a type I error is the level of significance of the statistical test of the hypothesis and is denoted by alpha. This error contrasts to a type II error, when one rejects the alternative hypothesis (fails to reject the null hypothesis) when the alternative hypothesis is actually true. A clinical trial should be adequately designed so that the alpha level is set at p ≤ 0.05.

Clear Statement of the Objectives, Proposed and Actual.

The objectives of the trial, primary and secondary, must be clearly defined within the protocol at the start of the study. The stated objectives should be adequately considered to address the study hypotheses. Exploratory objectives may also be included in the trial protocol, but these objectives are typically considered as hypothesis-generating rather than hypothesis-testing. Analyses not predefined and generated during the course of the study or assessed at the end of the study are typically denoted as post hoc analyses and are usually not appropriate for regulatory evaluation or decision.

Methods of Analysis in the Protocol, Statistical Plan, and Reports.

The statistical methods planned for the analysis of trial data should be prespecified in the trial protocol, perhaps expanded in a statistical analysis plan, ideally before the start of the trial but at least prior to the locking of study data. In this fashion, all methods are predefined and not subject to any bias as a result of the trial outcomes. Additionally any manuscripts or reports arising from the trial should adequately describe the analytical methods so that readers can appropriately consider and interpret the results of the trial.

Methods of Adequate Assurance of Patient Selections.

The protocol should include methods to enter patients appropriately so that the study population is representative of the broader population in general, that the study outcomes can be generalized to a broader population of similar patients. Multicenter trials (versus single-center or a limited number of centers) increase the heterogeneity of the study population and make the study results more generalizable. There must be adequate procedures in place to ensure that patients are randomly included and not specifically selected potentially to skew study results.

Patient Assignments That Minimize Bias, Group Comparability.

Patients should be randomly assigned to treatment groups so as to minimize the chances for any systematic differences between groups. Treatment groups should be as comparable as possible; in this manner, any differences in study outcomes between groups can confidently be ascribed to the study treatment. Treatment assignments should be sequential and generated via appropriate randomization schemes (usually computer-generated) to eliminate any potential for bias in treatment selection by investigators or sponsors. Should there be particular characteristics of interest (e.g., smoking status for periodontal trials), treatment assignments can be stratified to assure an equal distribution of those characteristics among all treatment groups.

A comment is needed about controls. Dental studies often use teeth randomized to different treatment assignments within the same patient (i.e., split-mouth designs), highlighting that patients serve as their own controls. This type of assignment should be discouraged, since the units of statistical analysis are no longer independent. With this intrapatient design, there are considerable concerns that treatment for one (or more) teeth (periodontal pockets) within one treatment group could affect the other treatment sites within the same patient or that treatments of all treatment groups could be affected by host factors, either equally or differentially. The parallel-group design is preferable so that the statistical unit is the patient, assuring that sites within treatment groups are completely independent and not subject to possible confounding by the other treatment group(s).

Methods to Minimize Bias for All Parties: Patients, Investigators, and Data Analysts.

Pivotal trials are typically double-blinded (or triple-blinded) so that no study participants (e.g., patients, investigators or statistical analysts) are aware of treatment assignments. In this fashion no patients, investigators, study staff, or those who analyze the data can either consciously or inadvertently influence the study outcome.

Endpoints Well Defined and Addressing the Primary Hypothesis.

The study protocol should clearly describe all of the endpoints that will be assessed and the methods of assessment. Pivotal trials may include many dozens of study sites; there is considerable risk for systematic differences among different investigators. Protocols must standardize study methods as much as possible among study sites so that all treatments and study procedures are performed in the same manner at all study centers and that all endpoints are assessed in the same manner as well. The endpoints chosen should be relevant to the study objectives.

A comment is needed about endpoints in periodontal trials. Current designs of clinical trials to assess periodontal outcomes are generally limited to the evaluation of surrogate endpoints (e.g., probing depth, attachment level) rather than direct and more clinically relevant endpoints such as tooth survival. An endpoint of tooth survival, although arguably more ideal, is not practical for a clinical trial that is to be conducted within a reasonable length of time or with a reasonable number of patients. Surrogate variables, however, are considered reasonable endpoints in periodontal clinical trials and relevant to tooth retention,28,34 although the inherent weaknesses of surrogate outcome variables have been noted.³³ Additionally, there may be discordance among surrogate periodontal outcome endpoints,⁵³ further highlighting the need for long-term data from clinical trials with tooth mortality as the primary objective.

Analysis of Results Allows for the Interpretability of the Effects of the Study Drug.

The statistical analysis should be conducted so that the study results can be interpreted confidently as caused by the study drug or treatment under investigation and not influenced by any systematic differences between treatment groups or biases of any of those who participated in any aspects of the study trial.

The basic objective of any clinical trial is to answer a particular research question. The worst result possible is to complete a trial but not to be able to interpret the outcome of the trial with confidence, essentially a nonresult that would require the repeat of the trial. The many elements of study design described above all contribute to the degree of confidence with which one can consider the outcome of the trial and the interpretation of the result.

The pivotal trials for the chlorhexidine chip³⁶ and the minocycline microspheres⁹¹ provide examples of Phase III designs for the study of adjunctive, locally delivered, controlled-release antimicrobials for the treatment of periodontitis. Representative populations of patients with chronic adult periodontitis were enrolled into multicentered, double-blinded clinical trials and were randomly assigned to treatment groups. Target pocket sites received SRP at baseline with or without adjunctive drug as per the randomization. At 3 and 6 months, sites randomized to drug received additional drug only if probing depth remained 5 mm or more. Final study results showed a reduction of probing depth from baseline at 9 months as expected in all groups, but the treatment group (SRP + adjunctive therapy) showed a significantly greater reduction from baseline versus control (SRP alone or SRP + placebo). If the adjunctive therapy had no effect, probing depth at these sites would have trended back toward baseline as in the SRP alone (or SRP + placebo) sites; the observation that probing depth remained reduced clearly demonstrated that the drug was efficacious. It should be noted for the chlorhexidine chip trials³⁶ that baseline measurements were obtained following a supragingival scaling. Since a supragingival scaling might result in a mean probing depth reduction of approximately 0.3 to 0.4 mm,²² interpretation of the absolute changes in probing depth as a result of treatment should include a consideration of this potential prebaseline change.

Also with respect to these trials, it has been hypothesized that trial results might have been different had control groups also received repeated treatment (i.e., repeated SRP).⁵ If the control group had received repeated SRP, however, to maintain design balance the adjunctively treated group also would have had to receive repeated instrumentation. It cannot be assumed that results might have been different only for the control group. Additionally, locally delivered, controlled-release antimicrobials may be effective in the presence of calculus³⁸ or with reduced amounts of instrumentation.⁹⁰ Additional instrumentation also would have made data interpretation more difficult; multiple treatments at multiple times make it more difficult to separate treatment effects meaningfully. Further in support of the lack of an impact of additional instrumentation on the outcome, a similarly significant difference was reported when both treatment arms received SRP at baseline and a supragingival prophylaxis at 3 months.⁶⁰

A final comment regarding locally delivered, controlled-release antimicrobials is needed regarding the consideration of adjunctive therapy or monotherapy. As suggested in Chapter 89, the general consensus of the professional community is that the best use for these products may be in the adjunctive setting. There are considerable data, however, that support efficacy in the monotherapy setting (for example, see Garrett et al²⁵), and Garrett et al²⁴ have suggested that monotherapy may be equally as effective as mechanical instrumentation for supportive periodontal maintenance. The authors cautioned, however, that further studies are needed to validate this treatment regimen and emphasized the importance of compliance including frequent supportive maintenance professional visits.

Surgical Therapy

Pocket sites that do not seem to respond adequately to nonsurgical therapy and evidence residual probing depth with inflammation are often treated with follow-up surgical therapy. An interesting question is whether locally delivered, controlled-release antimicrobials could augment surgical results. This hypothesis was tested in a pilot study by Hellström and coworkers³² with minocycline microspheres at all pocket sites 5 mm or more. Sixty patients with at least one nonmolar pocket site with probing depth 6 mm or more in each of two quadrants were treated with either (1) microspheres at baseline, then immediately following each of two surgical procedures at weeks 2 and 3 and again at week 5; or (2) standard surgical therapy. The mean probing depth was reduced in both treatment groups, but the reduction was significantly greater in the test group versus control at 25 weeks (2.51 ± 0.10 mm vs. 2.18 ± 0.10 mm, respectively; p = 0.027).

A specialized periodontal surgical procedure, regenerative surgery, has become a standard procedure within periodontal practice. In this regard it had been suggested verbally, as early as 1993 (Killoy WJ, personal communication, 2002), and in print in 1998,40,45 that adjunctive locally delivered, controlled-release antimicrobials may improve outcomes following regenerative periodontal surgery. In a pilot clinical trial, the adjunctive use of the chlorhexidine chip with regenerative surgery resulted in more than a 100% greater mean improvement from baseline in bone height and mass 9 months after surgical treatment compared with SRP alone and surgery.⁷¹ It is worth noting that both groups had also received prophylactic systemic antimicrobial treatment as well prior to surgery, but the addition of a locally administered antimicrobial still resulted in a significant benefit compared with control.

Other reports support the lack of efficacy of systemic antimicrobials,⁵⁴ and the benefit of locally delivered antimicrobials,⁹² as adjunctive treatments in the regenerative setting. A microbiological rationale for the use of locally delivered antimicrobials in the regenerative setting was also supported by Sbordone et al⁷⁸ with regard to the tetracycline fiber. Aichelmann-Reidy and coworkers³ have suggested that regenerative surgical procedures should include an adjunctive locally delivered, controlled-release antimicrobial agent in order to provide a more consistent clinical benefit.

These preliminary reports have suggested that the adjunctive use of locally delivered, controlled-release antimicrobials may improve clinical outcomes following periodontal surgery both in the regenerative and nonregenerative settings. These hypotheses remain to be tested in prospective, adequate, and well-controlled trials; none of the three available agents are currently approved t/>