Clinical_Trial_Paper

The Purpose of Clinical Research When a pharmaceutical company wants to bring a new drug to market, they need proof that the drug is safe and effective. It's important to realize, however, that no drug is absolutely safe. There is always a risk of an adverse reaction. The FDA uses a cost benefit analysis to determine if the drug can be released to the public. When the FDA determines that the benefits outweigh the risks they consider the drug safe enough to approve. The purpose of clinical research is to determine the safety and efficacy of the Investigational New Drug (IND) for the treatment of a particular disease or condition in humans. Clinical research is divided into three phases in the normal course of testing. The first phase is the Pre-Clinical Research Stage. During this stage, the drug is synthesized and purified. Tests are performed on animals and the institutional review board assesses the studies to make recommendations about how to proceed. If recommendations are positive, then an application is submitted to the FDA and clinical tests begin. Following the Pre-Clinical Research Stage are three phases of clinical studies and trials. These studies look at how the drug effects human volunteer patients. Next a larger clinical trial is conducted on a larger human population for the purpose of learning about short-term side effects or risks. Finally, the phase 3 clinical trials provide more informaiton about the safety of the drug and look at the effects on a larger population with which scientists extrapolate the results to the general population. There are several checks and balances in the process of clinical trials; among them is the use of institutional review boards (IRBs) and advisory committees. IRBs are designed to protect the rights and welfare of people participating in clinical trials both before and during the trials. IRB's are made up of a group of at least five experts and lay people with diverse backgrounds to provide a complete review of clinical proceedings. The CDER uses advisory committees of various experts in order to obtain outside opinions and advice about a new drug. It also provides new information for a previously approved drug, as well as labeling information about a drug, guidelines for developing particular kinds of drugs, or data showing the adequate safety and effects of the drug. The process of developing and testing a new drug is a lengthy one. The FDA estimates that it takes a little over 8 years to test a drug, including early laboratory and animal testing, before there is final approval for use by the general public. Final Actions The FDA's decision whether to approve a new drug for marketing comes down to answering two questions: Do the results of well-controlled studies provide substantial evidence of its effectiveness' Do the results show that the product is safe under the explicit conditions of use in the proposed labeling' Here "safe" is a relative term; it means that the benefits of the drug appear to outweigh its risks. When the review is complete, the FDA writes to the applicant to say the drug is either approved for marketing, is "approvable," provided minor changes are made, or is not approvable because of major problems. In the last case, the applicant can then amend or withdraw the NDA or ask for a hearing. Once its NDA is approved, a drug is on the market as soon as the firm gets its production and distribution systems {text:soft-page-break} going. Bayesian statistics is an approach used to learn from evidence as it accumulates. In clinical trials, traditional (frequentist) statistical methods may use information from previous studies only at the design stage. Then, at the data analysis stage, the information from these studies is not part of, the formal analysis. However, the Bayesian approach uses Bayes’ Theorem to combine prior information with current information. This approach considers the prior information and trial results as part of a continual stream of data, in which inferences are updated each time new data is available. When good prior information on a clinical use exists, the Bayesian approach may enable this information to be incorporated into the statistical analysis of a trial. The FDA believes that Bayesian methods are usually less controversial when the prior information is based on empirical evidence such as data from clinical trials. However, Bayesian methods can be controversial when the prior information is based mainly on personal opinion (often derived by elicitation from “experts”). The Bayesian approach is frequently useful in the absence of prior information. This approach can accommodate adaptive trials and even some trial modifications. Additionally, the Bayesian approach can be useful for analysis of a complex model. Other potential uses include adjustment for missing data, sensitivity analysis, multiple comparisons, and optimal decision making (Bayesian decision theory). The Bayesian approach, when correctly employed, may be less burdensome than a traditional non-Bayesian approach. The Federal Food, Drug, and Cosmetic Act (FFDCA) mandates that FDA shall consider the least burdensome appropriate means of evaluating effectiveness of a device that would have a reasonable likelihood of resulting in approval (see 21 U.S.C. 360c(a)(3)). The information from a current trial is augmented by incorporating prior information in a Bayesian analysis. The Bayesian analysis brings to bear the extra, relevant, prior information, which can help FDA make a decision. Additionally, if the prior information does not agree sufficiently with trial results, then the Bayesian analysis may actually be conservative relative to an analysis that does not incorporate prior information. Planning the design, conduct, and analysis of any trial is always important from a regulatory perspective, but is especially crucial for a Bayesian trial. In a Bayesian trial, decisions have to be made at the design stage regarding: the prior information, the information to be obtained from the trial, and the mathematical model used to combine the two. Different choices of prior information or different choices of model can produce different decisions. As a result, in the regulatory setting, the design of a Bayesian clinical trial involves pre-specification of and agreement on both the prior information and the model. Since reaching this agreement is often an iterative process, we recommend you meet with FDA early to obtain agreement upon the basic aspects of the Bayesian trial design. A change in the prior information or the model at a later stage of the trial may imperil the scientific validity of the trial results. For this reason, formal agreement meetings may be appropriate when using a Bayesian approach. The Bayesian approach can involve extensive mathematical modeling of a clinical trial, including: {text:soft-page-break} the probability distributions chosen to reflect the prior information, the relationships between multiple sources of prior information, the influence of covariates on patient outcomes or missing data, and sensitivity analyses on the model choices. Specific statistical and computational expertise The Bayesian approach often involves specific statistical expertise in Bayesian analysis and computation. Special computational algorithms like MCMC are often used to analyze trial data, check model assumptions, assess prior probabilities at the design stage, perform simulations to assess probabilities of various outcomes, and estimate sample size. The technical and statistical costs involved in successfully designing, conducting, and analyzing a Bayesian trial may be offset by the increased precision on device performance that can be obtained by incorporating prior information, or in the absence thereof, by the benefits of a flexible Bayesian trial design (e.g., smaller expected sample size resulting from interim analysis). Outcomes and Parameters Statistics is concerned with making inferences about unknown quantities of interest. A quantity of interest may be: an outcome associated with a patient or some other experimental or study unit, or a parameter, a quantity that describes a characteristic of the population from which the study is considered to be a sample. For instance, outcomes in device trials include: adverse events (e.g., death, renal failure, bleeding, myocardial infarctions, recurrence), measures of effectiveness (e.g., in cardiac function, visual acuity, patient satisfaction), and diagnostic test results. Parameters might describe: the rate of serious adverse events, the probability of device effectiveness for a patient, a patient’s survival probability, and sensitivity and specificity of a diagnostic device. The Bayesian Paradigm The Bayesian paradigm states that probability is the only measure of one’s uncertainty about an unknown quantity. In a Bayesian clinical trial, uncertainty about a quantity of interest is described according to probabilities, which are updated as information is gathered from the trial. Prior distribution Before a Bayesian trial begins and data are obtained, probabilities are given to all the possible values (or {text:soft-page-break} ranges of values) of an unknown quantity of interest. These probabilities, taken together, constitute the prior distribution for that quantity. In trials undergoing regulatory review, the prior distribution is usually based on data from previous trials (although mathematically they need not actually be temporally ordered). Bayes’ theorem and posterior probabilities After data from the trial become available, the prior distribution is updated according to Bayes’ theorem. This updated distribution is called the posterior distribution, from which one obtains the probabilities for values of the unknown quantity after data are observed. This approach is a scientifically valid way of combining previous information (the prior probabilities) with current data. The approach can be used in an iterative fashion as knowledge accumulates: today’s posterior probabilities become tomorrow’s prior probabilities. Bayesian inferences are based on the posterior distribution. For example, a Bayesian decision procedure might rule out a set of parameter values if the posterior probability of the parameter values (given the observed data) is small. Decision rules The pre-market evaluation of medical devices aims to demonstrate reasonable assurance of safety and effectiveness of a new device, often through pre-specified decision rules. Traditional hypothesis testing is an example of a type of decision rule. For Bayesian trials, one common type of decision rule considers that a hypothesis has been demonstrated (with reasonable assurance) if its posterior probability is large enough (“large enough” will be discussed later). The Bayesian approach encompasses a number of key concepts, some of which are not part of the traditional statistical approach. Below, we briefly discuss these concepts and contrast the Bayesian and frequentist approaches. 3.6 What is a predictive distribution' The Bayesian approach allows for the derivation of a special type of posterior probability; namely, the probability of unobserved outcomes (future or missing) given what has already been observed. This probability is called the predictive probability. Collectively, the probabilities for all possible values of the unobserved outcome are called the predictive distribution. Predictive distributions have many uses, including: determining when to stop a trial (based on predicting outcomes > for patients not yet observed), helping a physician and patient make decisions by predicting > the patient’s clinical outcome, given the observed outcomes > of patients in the clinical trial, predicting a clinical outcome from an earlier or more easily > measured outcome for that patient, augmenting trial results for missing data (imputation), > and model checking. Bayesian trials start with a sound clinical trial design Parts of a comprehensive trial protocol include: objectives of the trial, endpoints to be evaluated, conditions under which the trial will be conducted, population that will be investigated, and planned statistical analysis. Randomization minimizes bias that can be introduced in the selection of which patients get which treatment. Randomization allows concrete statements about the probability of imbalances in covariates due to chance alone. For reasonably large sample sizes, randomization ensures some degree of balance for all covariates, including those not measured in the study. In some cases, a Bayesian analysis of a new trial may salvage some information obtained in a previous non-Bayesian clinical trial that deviated from the original protocol. The information provided by such a trial may be represented by a prior distribution to be used in a prospective Bayesian clinical trial. Endpoints (also called parameters) are the measures of safety and effectiveness used to support a certain claim. Ideally, endpoints are: clinically relevant, directly observable, related to the claims for the device, and important to the patient. For example, an endpoint may be a measure of the average change in an important outcome (mortality, morbidity, quality of life) observed in the trial. The objective of a clinical trial is to gather information from the patients in the trial to make inferences about these unknown endpoints or parameters. Appropriate prior information should be carefully selected and incorporated into the analysis correctly. The FDA recommends identifying as many sources of good prior information as possible. Possible sources of prior information include: clinical trials conducted overseas, patient registries, clinical data on very similar products, and pilot studies. A Bayesian analysis of a current study of a new device may include prior information from: the new device, the control device, or both devices. The sample size in a clinical trial depends on: {text:soft-page-break} variability of the sample, prior information, mathematical model used in analysis, distributions of parameters in the analytic model, and specific decision criteria. Effective Sample Size A useful summary that can be computed for any simulations using posterior variance information is the effective sample size in the new trial. That is, Effective sample size (ESS) is given by: ESS = n * V1/V2, Where n = the sample size in the new trial V1 = the variance of the parameter of interest without borrowing (computed using a non-informative prior distribution) V2 = the variance of the parameter of interest with borrowing (computed using the proposed informative prior) Then, the quantity (ESS – n) can be interpreted as the number of patients “borrowed” from the previous trial. This summary is useful in quantifying the efficiency you are gaining from using the prior information. It is also useful for gauging if the prior is too informative. For example, if the number of patients “borrowed” is larger than the sample size for the trial, then the prior may be too informative and may need to be modified. At any point before or during a Bayesian clinical trial, you can obtain the predictive distribution for the sample size. Therefore, at any point in the trial, you can compute the expected additional number of observations needed to meet the stopping criterion. In other words, the predictive distribution for the sample size is continuously updated as the trial goes on. Because the sample size is not explicitly part of the stopping criterion, the trial can be ended at the precise point where enough information has been gathered to answer the important questions. Because of the inherent flexibility in the design of a Bayesian clinical trial, a thorough evaluation of the operating characteristics should be part of the trial planning. This includes evaluation of: probability of erroneously approving an ineffective or unsafe device (type I error rate), probability of erroneously disapproving a safe and effective device (type II error rate), power (the converse of type II error rate: the probability of appropriately approving a safe and effective device), sample size distribution (and expected sample size), prior probability of claims for the device, and if applicable, probability of stopping at each interim look. A trial design that is adaptive to information as it is accrued can allow for a decision to be made in an efficient way. According to the Bayesian approach, information from a trial should be gathered until it is deemed sufficient for a decision to be made. However, before the trial starts, there is a great deal of uncertainty on the sample size that will provide enough information for a decision to be made. If the variability of the sample is larger than expected, a larger sample size will be needed whereas if the variability is smaller than expected, a smaller sample size will suffice. {text:soft-page-break} Identification of each discrete outcome from your statistical analysis, providing statistical rationale for each one. Analyzing a Bayesian Clinical Trial 5.1 Summaries of the posterior distribution The posterior distribution contains all information from the prior distribution, combined with the results from the trial via the likelihood function. Conclusions from a Bayesian trial are based only on the posterior distribution. Statistical inference may include hypothesis testing, interval estimation, or both. For Bayesian hypothesis testing, you may use the posterior distribution to calculate the probability that a particular hypothesis is true, given the observed data. 5.3 Interval estimation Bayesian interval estimates are based on the posterior distribution and are called credible intervals. If the posterior probability that an endpoint lies in an interval is 0.95, then this interval is called a 95 percent credible interval. FDA strongly encourages reporting of credible intervals for Bayesian trials in the labeling. For an example on how credible intervals are reported, see the Summary of Safety and Effectiveness for InFuse Bone Graft / LT-CAGE ™ 9. Two types of credible intervals are highest posterior density (HPD) intervals (Lee, 1997) and central posterior intervals. For construction of credible intervals, see Chen & Shao (1999) and Irony (1992)_. _ 5.4 Predictive probabilities Uses of predictive probabilities (Section 3.6) include the following: Deciding when to stop a trial If it is part of the clinical trial plan, you may use a predictive probability at an interim point as the rule for stopping your trial. If the predictive probability that the trial will be successful is sufficiently high (based on results at the interim point), you may be able to stop the trial and declare success. If the predictive probability that the trial will be successful is small, you may stop the trial for futility and cut losses. Exchangeability is a key issue here; these predictions are reasonable only if you can assume the patients who have not been observed are exchangeable with the patients who have. This assumption is difficult to formally evaluate but may be more plausible in some instances (e.g., administrative censoring) than others (e.g., high patient drop-out). Predicting outcomes for future patients You may also calculate the predictive probability of the outcome of a future patient, given the observed outcomes of the patients in a clinical trial, provided the current patient is exchangeable with the patients in the trial. In fact, that probability answers the following questions: Given the results of the clinical trial, what is the probability that a new patient receiving the experimental {text:soft-page-break} treatment will be successful' What would that probability be if the patient were treated in the control group' We recommend that you consider the usefulness of this information in helping physicians and patients make decisions regarding treatment options and whether it should be included in the device labeling. Predicting (imputing) missing data You may use predictive probabilities to predict (or impute) missing data, and trial results can be adjusted accordingly. There are also frequentist methods for missing data imputation. Regardless of the method, the adjustment depends on the assumption that patients with missing outcomes follow the same statistical model as patients with observed outcomes. This means the missing patients are exchangeable with the non-missing patients, or that data are missing at random. If this assumption is questionable, FDA recommends you conduct a sensitivity analysis using the prediction model. For examples of missing data adjustments and sensitivity analysis, see the Summary and Safety Effectiveness for PMA P980048, BAK/Cervical Interbody Fusion System, by Sulzer Spine-Tech.10 When the missing data are suspected not to be missing at random, a sensitivity analysis can be built into the computational code to determine what degree of departure from the missing at random assumption is needed for conclusions to change (reference cf. Sulzer-Spine PMA SSED), which can help in alleviating the concern that the missing data may lead to invalid conclusions. Predicting a clinical outcome from earlier measurements If patients have measurements of the same outcome at earlier and later follow-up visits, you may make predictions for the later follow-up visit (even before the follow-up time has elapsed). Basing predictions on measures at the earlier visit requires that: some patients have results from both follow-up visits there is sufficiently high correlation between the early and the later measurement. In this example, the outcome at the first time point is being used to predict the outcome at the second. This type of prediction was used to justify early stopping for the clinical trial of the INTERFIX Intervertebral Body Fusion Device.11 The earlier measurement may also be on a different outcome; for example, for breast implants, rupture may be predictive of an adverse health outcome later. 5.5 Interim analyses FDA recommends you specify your method for analyzing interim results in the trial design and ensure FDA agrees in advance of the trial. The following describes two specific Bayesian interim analysis methods: Applying posterior probability One method stops the trial early if the posterior probability of a hypothesis at the interim look is large enough. In other words, the same Bayesian hypothesis test is repeated during the course of the trial. Applying predictive distribution Sensitivity Analysis Sensitivity analysis is used to investigate the effects of deviations from your statistical model and its assumptions. FDA may recommend that you submit a sensitivity analysis. This might include investigations of: deviations from the distributional assumptions (i.e. deviations from the assumptions regarding the stochastic elements of your model), alternative functional forms for the relationships in your model, alternative prior distributions, alternative “hyperprior” parameters in hierarchical models, or deviations from any “missing at random” assumptions for missing data. 5.8 Decision analysis Decision analysis is a prescriptive approach that enables a decision maker to logically analyze a problem in order to choose the best course of experimentation and action in the presence of uncertainty. A decision analysis method considers the consequences of decisions and experimentation to obtain optimum courses of action. Courses of action might include decisions to stop or continue a study or to approve or not approve a medical device. In any regulatory review of trial results, the consequences of possible decisions should be evaluated and reflected in the values of the posterior probabilities that will be used in order to make the claims. The more critical the endpoint, the more information about that endpoint will be necessary, and the sharper should be its posterior distribution. For example, if there is a possibility of death or serious disability with the use of the device, the Agency will want to be sure that the chance of occurrence of such event is small. A reasonable degree of certainty could be 90%, 95% or even 99%. These values will be reflected in the pre-specified posterior probabilities that the endpoints are below a certain target, or that the difference between the chances of death when using the new device and the control is smaller than a pre-specified value. A decision analysis method might in principle be used to develop an interim analysis plan. Carlin, Kadane, & Gelfand (1998) propose a method to approximate a decision analysis approach in interim analyses. 7.5 Calculations Almost all quantities of interest in a Bayesian analysis involve the calculation of a mathematical integral. All the following are expressed as an integral involving the posterior distribution: the posterior mean, the posterior standard deviation, the credible interval, and the posterior probability of a hypothesis.