J Dent Res 83(Spec Iss C):C6-C14, 2004
© 2004 International and American Associations for Dental Research
The Classic Caries Clinical Trial: Constraints and Opportunities
J.W. Stamm
School of Dentistry, #7450, The University of North Carolina at Chapel Hill, Chapel Hill, NC 27599-7450, USA; john_stamm{at}dentistry.unc.edu
KEY WORDS: clinical trial • caries prevention
 |
INTRODUCTION
|
|---|
The history of clinical trials would include events in 1747 on board the Salisbury, a British Navy vessel at sea with 12 seamen critically ill with scurvy. Involving these 12 sailors in a study, an officer on board by the name of Lind evaluated six potential treatments for scurvy, and rapidly reached the conclusion that daily consumption of citrus fruits returned the men fit for duty in approximately six days (Bull, 1959). The concept of experimental randomization was first developed by Sir R.A. Fisher (1925, 1926), and the method was introduced to medical research via a study of tuberculosis treatment by Amberson and co-workers (1931), who randomized 24 TB patients into two groups, one to receive the experimental therapy, the other serving as the control. Amberson et al. also incorporated the concept of blinding into their study. Sir Austin Bradford Hill codified and built on the principles of scientific experimentation developed by Fisher, and introduced the use of random numbers in the allocation of patients in the British Medical Research Council (1948) study of the effect of streptomycin in the treatment of tuberculosis (Daniels and Hill, 1952; Hill, 1952). The first applications of clinical trial methodology for testing interventions on dental, oral, and maxillofacial diseases and conditions are more difficult to determine. For dental caries prevention, however, Chilton and Fertig (1958) and Slack and Martin (1964) were certainly among the early caries clinical trial pioneers.
As clinical trials have come into the mainstream of clinical research in medicine and dentistry, a great deal of developmental work has focused on their methodological enhancement. The most successful of these efforts have come from fruitful, ongoing collaborations among clinician investigators, biostatisticians, data management specialists, biomedical ethicists, and others with an academic interest in clinical trial design and utilization. During the past 25 years, the emergence of systematic reviews and the evidence-based medicine (EBM) movement have also contributed significantly to the increasing reliance on randomized clinical trial outcomes for the advancement of better clinical practice (Richards et al., 1997; Straus and Sackett, 1998; www.cochrane.org/cochrane/ccbroch.htm#BDL, 2002).
|
CLINICAL TRIAL DEFINITION
|
|---|
A clinical trial is defined as a formally planned, prospective study in human beings that compares the effect and value of an intervention(s) with that achieved by a control treatment (Friedman et al., 1996). The study must be prospective, moving forward in time from baseline measurements established just prior to the application of the intervention to be evaluated. The clinical trial must evaluate one or more interventions that may be diagnostic, prophylactic, or therapeutic, comprised of agents, devices, procedures, or regimens. The trial must incorporate a control group that at the outset of the study is sufficiently similar to the treatment group(s) so that subsequent changes in health status may be reasonably ascribed to the intervention being tested.
|
TYPES OF CLINICAL TRIALS
|
|---|
It is common to categorize and refer to clinical trials as Phase IBIV studies. Phase I and Phase II studies rely on small convenience samples, sometimes fewer than 20 persons, and seek to establish, respectively, the safety and the dose-response of the (usually pharmacological and therapeutic) intervention. Phase III clinical trials represent the common, large-scale, randomized, double-blind clinical trials well-known to most clinical investigators. Occasionallym a larger Phase II trial achieving dramatically beneficial results may be extended and eventually accepted in lieu of a Phase III study. Phase IV studies have a shorter history, and represent large surveillance studies directed at monitoring primarily the safety of a therapy in its post-marketing period. Phase IV studies generally do not have a control group, although statistical control methods facilitate important sub-group comparisons.
Phase III clinical trials may be referred to as randomized clinical trials (RCTs) or as controlled clinical trials (CCT). In the context of the International Consensus Workshop on Caries Clinical Trials, it is implied that the term "caries trial" or "caries clinical trial" refers to a Phase III trial, one that incorporates randomization and blinding, as well as all the best features that make Phase III trials such a powerful method for evaluating the effectiveness of clinical interventions on human populations. The term pivotal trial or pivotal study usually refers to a Phase III clinical trial.
There are other schemes by which to categorize clinical trials (Jadad, 1998). Schwartz and Lellouch (1967) questioned whether the randomized clinical trial as usually conceived represented a sufficient strategy to permit the optimal selection of therapeutic interventions for normal medical practice. The thesis advanced by these authors was that the rigid system of controls, the very strength of the RCT, worked against producing the types of data necessary for deciding whether to implement an alternative therapeutic regimen. Schwartz and Lellouch argued that the requirements for internal validity by the RCT altered the definition of treatment and changed the manner of disease assessment from that used in conventional practice. Further, they suggested that participant selection, statistical management of withdrawal, and non-compliance, as well as the methods used for effecting study group comparisons, did not meet the needs for real-world decision-making. As part of their critique, they suggested that clinical trials have two purposes: one that is strictly scientific or explanatory, while the other would be decision-oriented or pragmatic. This theme was picked up and expanded for the dental clinical trials literature by O’Mullane (1976).
|
ETHICAL ISSUES
|
|---|
Ethical concerns properly arise when a randomized clinical trial is contemplated, because the safety and dignity of the human participants must be paramount. The issues to be considered are many and varied but inevitably include concerns about the process for gaining people’s participation, completeness and understanding of informed consent, adequacy of study design, potentially harmful effects of the agent being tested, and the possible withholding of beneficial treatment (Zeisel, 1970; Levine, 1986). Comprehensive international guidelines on ethics in biomedical research are readily accessible (CIOMS/WHO, 1993). The requirements of the US Department of Health and Human Services are also excellent, and are constantly being updated (USDHHS, 1991). The oral sciences literature contains several excellent discussions of ethical issues in dental clinical trials (Heifetz, 1983; Levine and Dennison, 1997). Since the very essence of randomized clinical trials is to provide the dental profession, the dental industry, and public health agencies with better therapeutic knowledge to apply in patient care, the public has the most to gain by continuing to give caries clinical trials its informed and full support.
Over 50 years of experience indicates that the risks to study participants enrolling in caries clinical trials are extremely low. Yet it is essential that all the conventional and currently accepted standards for the conduct of clinical trials be incorporated into caries studies. Thorough review of the clinical trial protocol by a properly constituted Institutional Review Board is the best assurance for this purpose. A subtle ethical issue that needs constant vigilance is potential abuse of administrative availability of study participants. Caries trials usually need large numbers of study participants, a need that typically has implications for both recruitment and informed consent procedures. Recruitment procedures must eschew the inappropriate reliance on the administrative availability of varying institutional populations. Every effort must be made to ensure that potential study participants and/or their guardians freely agree, as individuals and without prejudice, to enroll in caries clinical trials. Moreover, every participant must be aware of his/her complete right of withdrawal from a trial, without giving reason and without prejudice.
The potential role of placebo controls in caries clinical trials may also arise as an ethical issue. The consensus appears to be that where fluoride preventives are at issue, standard fluoride-containing positive controls should displace placebo controls. With respect to non-fluoride-containing preventives, placebo controls may be considered only after very careful deliberation. Included in such deliberation should be due consideration of "clinical equipoise", which may provide the rationale for launching a placebo-controlled study while expecting, a priori, that the active therapeutic agent may be effective in preventing dental caries (Freedman, 1987; Levine and Dennison, 1997; Shapiro and Glass, 2000; Weijer et al., 2000). Some have suggested crossover placebo-controlled and split-mouth designs as potential protocols for ethically incorporating placebo therapies into a trial, but no consensus has been reached on this point of view (Louis et al., 1984).
The increasing internationalization of caries clinical trials may introduce significant additional complexities, including unique social, cultural, and political issues. For example, recent experiences have demonstrated that caries clinical trials designed and sponsored by scientists from country A for a study to take place in country B may encounter a variety of concerns and objections on both ethical and technical grounds, e.g., language of informed consent or the use of radiographs. While resolution of these concerns may seem straightforward initially, in fact what seems like an appropriate resolution locally may compromise the sponsor’s longer-term desire to use the study results to support a caries prevention claim back in country A, or in yet other countries. Additional examples of the problems faced include: (1) excessive economic incentives to obtain permission to conduct the study in defined international settings, (2) inappropriate economic incentives for gaining individual informed consent, (3) objections to wording in informed consent forms with specific reference to human rights, and (4) potential inability of the investigators to provide or refer for care when dental disease is discovered as part of the study’s clinical examinations. These and other concerns and misunderstandings are examples of issues that may arise in internationally placed caries clinical trials, problems that invariably need knowledgeable and diplomatic resolution prior to study launch.
|
RANDOMIZATION IN CLINICAL TRIALS
|
|---|
A distinguishing and unique feature of the RCT is that it includes, for prospective investigation, one or more experimental group(s) and a concurrent control group to which subjects who enter the study are allocated by a formal randomization procedure. The purpose of randomization is to ensure that, at the outset of the study, the experimental and control groups are as alike as possible. During the actual conduct of the RCT, it is the role of the study protocol to maintain the comparability of the two groups on all factors other than the intervention being investigated.
Randomization of study participants may be attained through one of several methods, each of which may be associated with inherent advantages and disadvantages. For details concerning these techniques and indications for their use, good reference materials may be consulted (Pocock, 1983; Meinert, 1986; Friedman et al., 1996). Randomization methods may be fairly involved and labor-intensive, but are also essential to the RCT for several key reasons, as follows:
First, all proper randomization procedures seek to prevent selection bias. That is, randomized assignment of treatment to study participants is unpredicted, thus eliminating bias from the assignment of treatment once the study subject has agreed to enter the trial. In caries trials, for example, this would ensure that a potentially effective preventive is not disproportionately assigned, even subconsciously, to subjects most in need, or vice versa.
A second advantage is that randomization tends to balance the treatment groups with respect to prognostic factors or covariates that may be related to dental caries outcomes. Significantly, this advantage holds whether such prognostic factors are known or unknown in advance. For example, stratified randomization will balance study groups with respect to age, a known prognostic variable for caries increments. It is clear, however, that this could also be accomplished post hoc by the use of covariance analysis. But in distinct contrast, covariance cannot be used at the analysis stage to balance or adjust study groups post hoc for antecedent fluoride tablet use, a variable on which it is almost impossible to obtain reliable information. However, in such circumstances randomization still remains effective. The applications of stratification and covariance techniques for managing prognostic factors have been described, respectively, by Kingman (1984) and Grainger et al.(1984).
A third advantage to randomization is that it provides some of the necessary theoretical underpinnings for the statistical procedures to be used in the study analysis. In practice, this means that the investigator may apply the appropriate statistical tests to the data generated by the trial and, on the basis of the outcomes from the analyses, may assign statistical significance levels to observed differences in treatment effect.
Randomization of study participants may be grouped into two distinct categories—fixed allocation randomization procedures and adaptive randomization procedures. The first category is comprised of the most commonly known randomization procedures used in caries clinical trials, namely, (1) simple randomization, (2) blocked randomization, and (3) stratified randomization. Excellent and extensive sources may be consulted for the proper selection and implementation of these important randomization techniques. Moreover, the resultant approaches to statistical analysis are powerful and well-known. The second category, adaptive randomization, includes procedures such as (1) baseline adaptive randomization and (2) response adaptive randomization. These methods have not gained much popularity in the caries trials field, in part because the additional effort-to-yield ratio seems low, the associated statistical analysis methods are less well-known, or the slow emergence of the caries outcome limits the applicability of the method.
Much more important is that a proper randomization process be used, even if relatively uncomplicated, and that it is clearly detailed in the methods section of the trial. It is altogether too common to find that randomization procedures in caries clinical trials are incompletely described.
|
BLINDING IN CLINICAL TRIALS
|
|---|
While randomization eliminates bias at the outset of a RCT, the use of appropriate blinding or masking techniques as part of the ongoing study protocol is essential to maintaining freedom from bias in generating the data, i.e., during the trial. Common methods include the single-blind and the double-blind study designs, although triple-blind designs to cover third-party examiners and/or data analysts have also been advanced. It is clear that double- and triple-blinding is preferred wherever such techniques are feasible. The need for and the method of blinding should be carefully planned and monitored. In particular, since the double- and triple-blind procedures are complex and fragile, the investigators should ensure that the integrity of such procedures is maintained throughout the study. If not, the effort and expense devoted to using these techniques in the first place will be largely wasted.
If the proper blinding procedure is to be maintained in caries clinical trials, great attention must be paid to basic matters such as the packaging and labeling of the therapies to be evaluated. The object must be to ensure that treatment and control therapies are indistinguishable from each other to the participant and the clinician/examiner alike. The physical characteristics of the therapies, including the color, flavor, viscosity, etc., should be similarly indistinguishable. Maintaining subject and examiner blindness becomes more problematic the greater the number of active therapies being evaluated. This problem arises because it is difficult to produce a positive control therapy that conforms to the characteristic of multiple treatment groups simultaneously. This same general difficulty is encountered when caries-preventive products are tested for equivalency in caries prevention effectiveness. Friedman et al.(1986) suggest a complex but ingenious effort to cope in part with this type of clinical trial problem, one that has been successfully applied in a cardiovascular clinical trial (CAPS, 1994).
|
OVERVIEW OF DESIGNS FOR CARIES CLINICAL TRIALS
|
|---|
Various experimental plans that may be considered for designing a caries clinical trial were originally derived from experimental methods developed for agricultural research. These methods were codified and refined for biomedical research, and have been compiled into several excellent reference sources (Pocock, 1983; Meinert, 1986; Friedman et al., 1996). It is recommended that principal investigators involve a well-qualified statistician from the earliest phases of the caries clinical trial planning and design process.
Simple Randomized Designs
Simple randomized designs are controlled prospective studies with an intervention and a concurrent control group. Patients are assigned to either group by a formal randomization procedure. Simple randomized designs may be extended to incorporate multiple intervention groups (study cells or arms). In analogous fashion, trials may involve more than one concurrent control group, specifically placebo and active control interventions. As was alluded to above, in many cases stratification or blocking procedures are used prior to implementation of randomization to treatment groups.
Crossover Designs
Crossover designs are a special case of randomized trials in which each study participant serves as his/her own control (Brown, 1980; Louis et al., 1984; James et al., 1985; Koch et al., 1989). The simplest case is the two-period crossover design, in which the participants are randomly assigned to begin a course of treatment with either the active or control intervention, and when the planned treatment period is over, and after a "wash-out period", are assigned to the opposite intervention. This method ensures that the order of the interventions assigned to the participants is randomized. Extensions to three-period crossover designs have also been described (Carriere, 1994). While crossover designs offer several well-known advantages and disadvantages, they are rarely used for caries clinical trials because the length of treatment is too long and the problems of the wash-out period are too great.
Factorial Designs
Factorial designs provide the opportunity for simultaneous evaluation of two interventions, singly and in combination, against a single control within the framework of a single study. Factorial designs lend themselves well to caries clinical trials. The major risk in factorial designs lies in the possibility of interactions among interventions. When interaction is present, the two interventions are individually compared with the control, effectively reducing sample size and study power. When interaction is absent, the factorial design becomes a powerful, highly efficient method for evaluating intervention effectiveness, providing more "bang for the buck" than simple randomized designs. Brittain and Wittes (1989) explore the pros and cons of factorial designs in considerable detail.
Designs Involving Group Randomization
In the usual randomized clinical trial, individual subjects are allocated to the experimental and control groups. However, there are caries-preventive measures for which it is possible to randomize groups but not individuals. Clinical trials of school water fluoridation or school-based dental health education programs are but two examples. Three reasons for relying on group randomization are: administrative convenience, therapeutic necessity, or the need to avoid between-subject contamination in the case of preventive programs that involve cognition on the part of study subjects. Frequently, all three considerations apply simultaneously. Group randomization poses some special problems for the investigator. For example, both single- and double-blind protocols are almost impossible to maintain, but with ingenuity, even such a dilemma may be overcome. For example, to control examiner bias in the Kingston-Newburgh water fluoridation trial, single-blind assessment of dental x-rays was carried out which confirmed that the overall findings of the unblinded clinical examiners were relatively unbiased. Furthermore, sample sizes generally have to be increased when group randomization is to be used.
Designs for Studies of Intervention Equivalency
In the caries clinical trial field, it has become increasingly common to undertake studies designed to demonstrate intervention equivalency to a known effective preventive therapy (e.g., fluoride dentifrice). In the statistical literature, such studies are referred to as equivalency or "non-inferiority" trials. These designs are important, since they counter the common yet erroneous earlier practice of assuming that the absence of statistical significance between two or more study groups in a conventional RCT could be interpreted as demonstrating a therapeutically equivalent effect. Just as it is inappropriate to consider absence of a statistical significance in a superiority trial to indicate equivalence among interventions, it is usually inappropriate to convert, post hoc, a standard superiority trial involving a positive control to one that assesses equivalency. Establishing intervention equivalency, or an "at least as good as" standard, is frequently a difficult or contentious first step in these types of designs. In most cases, good equivalency or non-superiority caries trials, within the conventional parallel group trial design, generally means that larger sample sizes must be planned for. Equivalency studies may also carry a larger burden, in that they should incorporate the gathering of collateral or secondary information beyond the main outcome condition. This is to ensure that if equivalency is established in the primary outcome, analysis of the additional factors such as cost, convenience, availability, flavor, stability, etc., may be considered in making recommendations for clinical or community practice. Various theoretical and practical issues relating to equivalency or "non-inferiority" designs for clinical trials have been addressed by numerous investigators (Schuirmann, 1987; Council on Dental Therapeutics, 1988; Chow and Liu, 1992; Kingman, 1992; Proskin et al., 1995; Jones et al., 1996; Wiens and Iglewicz, 2000; Blackwelder, 2001).
|
SAMPLE SIZE
|
|---|
The question of adequate sample size for caries clinical trials is one that still receives too little attention in spite of notable contributions to the literature by McClendon et al.(1972) and Kingman (1977, 1978). The main and continuing problem is that too many clinical trials are undertaken without regard to the sample size requirements needed to reduce the Type II error to acceptable levels, conventionally 0.20 or 0.10. Trials embarked upon with inadequate statistical power from the outset stand a good chance of generating indeterminate results and contributing little to the question being asked. This is particularly true of caries trials involving schoolchildren because mean annual decayed surface increments and associated variances appear significantly smaller than was the case 25 or more years ago.
The proper estimation of sample sizes likely to be required for a specific caries trial should be a very early consideration in study planning. Many aspects must be considered, including the type of dental caries to be studied (permanent or primary teeth, coronal caries only, root caries only, both), the likely base caries increment level, participant compliance, and participant attrition from the study, as well as other factors. Sample size estimation must also take into account whether the clinical trial is intended to test preventive equivalency or "non-inferiority" performance (Makuch and Simon, 1978; Blackwelder, 1982; Blackwelder and Chang, 1984). The likely base caries increment level is particularly important, since decades of experience suggest that caries increment in control groups is usually lower than anticipated by investigators at the beginning of the study. Moreover, when this lower caries increment is combined with the frequently lower-than-anticipated caries-preventive effect in the active treatment group, the main study results tend toward the null, and the overall outcome may well fall short of study expectations. A judgment about contemporary clinical significance also has to be made.
There are numerous references that give exquisite guidance on the question of sample size determination (Donner et al., 1981; Lachin, 1981; Blackwelder and Chang, 1984; Whitehead, 1986). In addition, however, an accurate if basic understanding of the principles of hypothesis testing, significance level, and power will be critical for applying a sound approach to sample size estimation. In many conventional caries trials, sample sizes will be estimated with alpha = 0.05 and a power of 1-beta = 0.90 (although others may regard 0.80 as a conventional level). The calculation of estimated sample sizes required for a caries trial may be readily extended to studies incorporating extensive pairing or matching, multiple treatment groups, repeated outcome measures, and time-to-failure outcomes (Friedman et al., 1996). When group rather individual randomization is to be used, relatively simple procedures for determining sample size have been devised for the case where the size of the group to be randomized is small (Donner et al., 1981), or where it is large (Cornfield, 1978). Good reviews of group randomization procedures exist (Gillum et al., 1980; Buck and Donner, 1982).
|
PARTICIPANT ACCRUAL
|
|---|
One of the most common challenges in caries clinical trials is the accrual of study participants on a timely basis (Weinstein et al., 1995). Frequently, the principal investigator will underestimate the difficulty of enrolling and retaining participants in a caries trial, particularly one directed at pre-school or adult individuals. The magnitude of this problem is also related to the choice of inclusion and exclusion criteria that are adopted for a particular caries trial. For example, restricting enrollment to persons who meet certain "high caries risk" inclusion criteria may eliminate a surprising number of potential participants, with deleterious consequences for patient accrual. Also, investigators are often disappointed by the low level of interest expressed by various administrative entities for participating in caries clinical trials. Inappropriately aggressive efforts to overcome this type of reluctance may lead to ethical concerns as discussed earlier.
Aside from the above technical considerations, probably the most common reasons for not enrolling the projected study participants on a timely basis include inadequate planning, failure to get the study under way on time, or lack of managerial ability or commitment. Even if participant enrollment is eventually achieved, failure to meet patient accrual within the targeted time frame may create negative consequences such as turnover in examiners, increased training costs, drifting clinical examination standards, and the like. Finally, just as careful and conservative sample size estimation is crucial, the very best sample estimation techniques can be rendered moot by inattention to the challenges of patient accrual and retention. Thus, the timely launching and careful management of patient recruitment are fundamental for the conduct of successful caries clinical trials.
|
CARIES DEFINITION AND MEASUREMENT
|
|---|
For centuries, the defining feature of dental caries was cavitation in tooth surface structure ascribable to acids formed in the bacterial biofilm on the tooth surface. It had long been recognized that initial bacterial acid dissolution of tooth enamel could lead to defined, non-cavitated white lesions that either proceeded to the cavitation stage, or that would become arrested and, in cases, discolored over time. But following clinical dental practice at the time, as well as the early conventions of dental epidemiologists, classic caries clinical trials adopted physical softening and cavitation of tooth structure, supplemented with radiological assessment, as both the definition and measurement of caries (Radike, 1972). Combining the decayed (D) surface with missing and filled surfaces generated the Decayed, Missing and Filled Tooth/Surface (DMFT/S) index (Klein et al., 1938). When the DMFS index is assessed longitudinally in the same patient, a DMFS increment measurement may be generated, and this serves as the prime outcome variable for most caries clinical trials to this day. It is beyond the scope of this paper to deal with the analytical issues linked to alternative measures of caries activity or progression and the impact that such approaches might have on the direction of clinical trials.
Significant efforts are under way to reduce or eliminate vertical probing of demineralized tooth surfaces with sharp dental explorers. Within the context of clinical trials, preliminary initiatives are being made to demonstrate that tactile caries examination procedures can be substituted with visual assessment methods for caries. Experience in conducting caries trials based largely on the reliance on visual diagnostic methods is still in an early stage, and more data need to be accumulated and more validation will need to occur. Wider training and calibration of caries examiners using the emerging visual diagnostic standards for caries would be beneficial.
For over a century, efforts have been made to develop technologies that would refine the detection and measurement of caries. Beyond the dental explorer, these include the x-ray (1890s), optical transillumination (1920s), electrical conductance (1960s), dyes, a variety of advanced light refraction technologies, magnetic resonance imaging (MRI), tuned aperture computed tomography (TACT), and optical coherence tomography (OCT), as well as others. Over time, each of those original diagnostic technologies has been considerably refined, extended, and improved. In recent years, digitization has further enhanced the capture and display of signals from many of these technologies. Virtually all these new technologies are aimed at better measurement of the severity and extent of the initial stages of tooth structure demineralization as a direct measure of the early caries process. The increasingly advanced caries measurement technologies being developed today promise to have a major impact on the design and conduct of caries clinical trials in the near future.
|
RELIABILITY OF CLINICAL CARIES MEASUREMENTS IN RCTS
|
|---|
The ascertainment of caries is subject to variability in the detection, measurement, and recording parameters that together bear on the quality of the caries data that are analyzed in conventional caries clinical trials. The quality of caries data can be influenced in many ways that bear on either measurement bias and/or measurement precision, consequences that influence both the analysis and inference associated with the trial. Aside from the use of valid, standardized, and more sensitive measurement techniques, the issues of examiner reliability, single vs. multiple examiners, and clinical examiner training and calibration arise (Fleiss et al., 1979; Bell and Klein, 1984; Hunt, 1986). Maintaining a valid, sensitive, and highly reproducible caries measurement system throughout a caries trial is an important element in ensuring a high-quality clinical study.
|
USE OF RADIOGRAPHS IN CARIES TRIALS
|
|---|
The potential role of dental radiographs in enhancing investigators’ ability to separate treatment groups in caries trials deserves more study. Toward this end, knowledge-based image analysis systems have been developed, and digital subtraction radiology for caries detection has received attention (Firestone et al., 1998; Wenzel et al., 2000). It should be emphasized that the key issue is not simply the increased diagnostic sensitivity of radiographs for interproximal caries detection, but rather the actual impact of radiograph utilization on caries trial efficiency. [For fuller discussions, see Williams et al., 1967; Clark et al., 1982; Klein et al., 1984; Downer and Worthington, 1992.] Unless the advantages of using radiographs in caries clinical trials are clear and unambiguous, the ethical concerns associated with the use of ionizing radiation will become a barrier to routine use of radiographs in caries trials (Leske and Ripa, 1980).
|
DATA ACQUISITION
|
|---|
Given a properly developed caries trial protocol, the data acquisition process consists of two components. One is the technology to be used for recording and entering the data into computer files; the second is the quality control procedures that ensure the highest standard in the eventual database. With respect to technology considerations, optical scanning forms, direct data entry into PC-based custom programs, and direct data entry onto custom-designed Web-based forms appear to be the current standards. All such systems require careful prior planning and development, but all three approaches provide the investigator team with more efficient approaches to data capture. Direct data entry into the computer has the additional advantage of allowing for a variety of superior, online error-checking procedures. Web-based systems also encourage online data backup and easier data distribution approaches. Web-based data acquisition systems represent the future standard for registering caries clinical trial data. A comprehensive review of these newer approaches is provided by Marks (2004).
Quality control procedures must be used explicitly and continuously to ensure a caries trial dataset that meets the highest standards. Having the field personnel work from a first-rate protocol and a detailed, high-quality procedures manual is a key initial step. Proper and sufficient training of the personnel working to measure and record the clinical outcome information is mandatory. Adequate pre-testing of all measurement and data recording procedures is essential, as is the incorporation into the procedures manual of the insights gained during the pre-testing phase. Additional quality control steps include minimizing missing data, ensuring conscientious data entry, and incorporating timely and attentive quality-monitoring processes. The monitoring of the data quality should be done on a frequent basis so that procedural discrepancies can be corrected quickly before too many data are collected and entered under a faulty scheme. Subtle discrepancies between examiners’ diagnostic standards are a common problem and can creep into any caries clinical trial, even when initial calibration sessions were well-conducted. Frequent monitoring during the data acquisition phase will pay many dividends in ensuring a high-quality dataset for the eventual statistical analysis.
|
DATA ANALYSIS
|
|---|
Data analysis for caries clinical trials has evolved immensely over the past two decades. The progress in statistical theory and methods, the recognition of academically interesting analytical issues in caries clinical trials, and the increasingly challenging interplay between clinical and statistical issues have drawn biostatisticians to the caries trials field, where they have contributed substantially to the higher standards in caries trials analysis. Additionally, the more stringent demands for first-rate statistical work by regulatory agencies, such as the FDA in the United States, have encouraged caries trials’ sponsors to invest more heavily in statistical support for their studies. The evolution of the main design and data analysis issues involved in traditional caries clinical trials fields is well-summarized in several publications (Fleiss, 1984; Grainger et al., 1984; Geary et al., 1992; Hujoel et al., 1994; Caplan et al., 1999).
While the classic caries trial was well-served by statistical techniques based largely upon analysis of variance and its many refinements, the decline in caries experience during the past 20 years, the smaller incremental benefits from recent-generation caries-preventives, and the desire to demonstrate a product’s clinical equivalence to an existing and effective preventive therapy have encouraged the development of newer analytical techniques and strategies, including survival analysis methods (Hannigan et al., 2001). The increased methodological sophistication in caries detection and measurement, study design, and statistical approaches makes it urgent that new data analysis methods be continuously explored and published (Hannigan, 2004; Imrey and Kingman, 2004; Johnson, 2004; Katz and Huntington, 2004; Mancl et al., 2004).
The data analysis in the traditional caries trial focused on study participants who completed the study. In some cases, the data analysis plan excluded participants who were unavailable for interim examinations, or who provided evidence of "non-compliance", i.e., falling below a defined level of adherence to the regimen criterion. In recent years, agencies such as the US FDA have encouraged clinical trials to incorporate the "intent-to-treat" (ITT) standard into the data analysis (D’Agostino, 2004). The compelling advantage of the ITT approach to clinical trial analysis is that it removes bias that could arise from non-random loss of study participants and/or from non-random compromise in adherence to the treatment regimen. Use of the ITT principle requires that, once enrolled into the study, a participant’s sub-optimal compliance should not lead to excluding his/her data in the analysis of the trial. To prevent bias from participant loss in the study, the ITT analysis might implement the "last observation carried forward" technique (Pledger, 1992). Since intent-to-treat analyses are frequently associated with reduction in demonstrated treatment effect, study power may also be adversely affected. Planned, effective effort to maintain study participants within the study and to reduce non-adherence with the study protocol are the best ways of countering the analytical disadvantage brought by the ITT analysis. A second approach to compensate for the potential loss of power with ITT analysis is to increase the sample size a priori, which, however, is a potentially expensive approach. Statisticians are also examining additional analytical techniques to deal with subject attrition, including increasingly sophisticated methods for data imputation.
|
MULTICENTER TRIALS
|
|---|
A multicenter trial is a collaborative effort involving more than one independent center for the enrollment and follow-up of study subjects. The main rationale, design, execution, and analysis principles have been addressed by numerous investigators (Meinert, 1981, 1986, 1998; Fleiss, 1982, 1986; Gould, 1998; Senn, 1998). In medicine, the relative number of multicenter trials has risen considerably during the past 15 years (Friedman et al., 1996). This trend may well be advised for clinical dental studies for several reasons. First, multicenter caries trials may be one way of overcoming the increasing difficulty in recruiting study populations that are both sufficiently large and meet specified criteria on prognostic variables, particularly minimum, age-specific DMFS levels. Second, multicenter caries trials may offer more representative sampling of the population, thereby permitting stronger inference from the study results to the intended target population.
As discussed in several of the citations in the preceding paragraph, multicenter trials also have potential disadvantages that must be confronted before undertaking such a project. First, there will be a need for a substantial logistical organization. This means that appreciable resources will have to be devoted to various management functions to ensure a quality undertaking. Second, the multicenter trial runs a risk that one among the participating centers may not adhere to the aims and protocol as originally laid out. Third, unexpected data problems and analytical difficulties can arise that may well complicate the interpretation of the combined trial results (Senn, 1998).
|
CARIES CLINICAL TRIAL MONITORING
|
|---|
A key but often underappreciated aspect of conducting a caries clinical trial is the application of appropriate and sufficient monitoring to the actual conduct of the trial. As the size and complexity of caries trials have grown, and as the various standards required by regulatory agencies have risen, the practical management functions necessary to ensure a quality caries trial have become more onerous. A similar phenomenon has been observed on the medical side, where the trials are even larger and the stakes are oftentimes greater. The monitoring referred to in this section deals with the procedures designed to ensure that the day-to-day clinical trial management meets the standards as described in the relevant protocol. An exquisite summary of the problems related to inadequate trial monitoring has been provided by Cohen (1994), writing in Science. Published evidence that better monitoring is needed for caries trials is difficult to find, because the results from compromised trials are not published, or references to specific procedural problems are not included in reports and manuscripts. Yet there are anecdotal reports of caries examiner drift, loss of examiner calibration, inadequate rigor in maintaining examiner/subject assignments, inexplicable differences in baseline caries scores among study subgroups, diagnostically inadequate radiographs, and lower-than-desired regimen compliance. In most cases, such problems reveal inattention to quality control and other managerial aspects of the study.
It is important to distinguish study monitoring as described above from the biostatistical procedure of interim analyses. An excellent report characterizing this important distinction has been provided by Enas et al.(1989).
|
ADVERSE EFFECTS ASCERTAINMENT AND REPORTING
|
|---|
TOP
INTRODUCTION
CLINICAL TRIAL DEFINITION