Clinical Trial Data Analysis Using R Pdf Download
Shanae Maerz
Background: Research using data-driven cluster analysis has proposed five subgroups of diabetes with differences in diabetes progression and risk of complications. We aimed to compare the clinical utility of this subgroup-based approach for predicting patient outcomes with an alternative strategy of developing models for each outcome using simple patient characteristics.
Methods: We identified five clusters in the ADOPT trial (n=4351) using the same data-driven cluster analysis as reported by Ahlqvist and colleagues. Differences between clusters in glycaemic and renal progression were investigated and contrasted with stratification using simple continuous clinical features (age at diagnosis for glycaemic progression and baseline renal function for renal progression). We compared the effectiveness of a strategy of selecting glucose-lowering therapy using clusters with one combining simple clinical features (sex, BMI, age at diagnosis, baseline HbA1c) in an independent trial cohort (RECORD [n=4447]).
Findings: Clusters identified in trial data were similar to those described in the original study by Ahlqvist and colleagues. Clusters showed differences in glycaemic progression, but a model using age at diagnosis alone explained a similar amount of variation in progression. We found differences in incidence of chronic kidney disease between clusters; however, estimated glomerular filtration rate at baseline was a better predictor of time to chronic kidney disease. Clusters differed in glycaemic response, with a particular benefit for thiazolidinediones in patients in the severe insulin-resistant diabetes cluster and for sulfonylureas in patients in the mild age-related diabetes cluster. However, simple clinical features outperformed clusters to select therapy for individual patients.
Interpretation: The proposed data-driven clusters differ in diabetes progression and treatment response, but models that are based on simple continuous clinical features are more useful to stratify patients. This finding suggests that precision medicine in type 2 diabetes is likely to have most clinical utility if it is based on an approach of using specific phenotypic measures to predict specific outcomes, rather than assigning patients to subgroups.
A necessary companion to well-designed clinical trial is its appropriate statistical analysis. Assuming that a clinical trial will produce data that could reveal differences in effects between two or more interventions, statistical analyses are used to determine whether such differences are real or are due to chance. Data analysis for small clinical trials in particular must be focused. In the context of a small clinical trial, it is especially important for researchers to make a clear distinction between preliminary evidence and confirmatory data analysis. When the sample population is small, it is important to gather considerable preliminary evidence on related subjects before the trial is conducted to define the size needed to determine a critical effect. It may be that statistical hypothesis testing is premature. Thus, testing of a null hypothesis might be particularly challenging in the context of a small clinical trial. Thus, in some cases it might be important to focus on evidence rather than to test a hypothesis (Royall, 1997). This is because a small clinical trial is less likely to be self-contained, providing all of the necessary evidence to effectively test a particular hypothesis. Instead, it might be necessary to summarize all of the evidence from the trial and combine it with other evidence available from other trials or laboratory studies. A single large clinical trial is often insufficient to answer a biomedical research question, and it is even more unlikely that a single small clinical trial can do so. Thus, analyses of data must consider the limitations of the data at hand and their context in comparison with those of other similar or related studies.
Since data analysis for small clinical trials inevitably involves a number of assumptions, it is logical that several different statistical analysis be conducted. If these analysis give consistent results under different assumptions, one can be more confident that the results are not due to unwarranted assumptions. In general, certain types of analysis (see Box 3-1) are more amenable to small studies. Each is briefly described in the sections that follow.
Sequential analysis refers to an analysis of the data as they accumulate, with a view toward stopping the study as soon as the results become statistically compelling. This is in contrast to a sequential design (see Chapter 2), in which the probability that a participant is assigned to a particular intervention is changed depending on the accumulating results. In sequential analysis the probabilty of assignment to an intervention is constant across the study.
Sequential analysis methods were first used in the context of industrial quality control in the late 1920s (Dodge and Romig, 1929). The use of sequential analysis in clinical trials has been extensively described by Armitage (1975), Heitjan (1997), and Whitehead (1999). Briefly, the data are analyzed as the results for each participant are obtained. After each observation, the decision is made to (1) continue the study by enrolling additional participants, (2) stop the study with the conclusion that there is a statistically significant difference between the treatments, or (3) stop the study and conclude that there is not a statistically significant difference between the interventions. The boundaries for the decision-making process are constructed by using considerations of power and size needed to determine an effect size similar to those used to determine sample size (see, for example Whitehead [1999]). Commercially available software can be used to construct the boundaries.
In sequential analysis, the final sample size is not known at the beginning of the study. On average, sequential analysis will lead to a smaller average sample size than that in an equivalently powered study with a fixed-sample-size design. This is a major advantage to sequential analysis and is a reason that it should be given consideration when one is planning and analyzing a small clinical trial. For example, take the case study of sickle cell disease introduced in Chapter 1 and consider the analysis of the clinical design problem introduced in Box 1-4 as an example of sequential analysis (Box 3-2).
Data from a clinical trial accumulate gradually over a period of time that can extend to months or even years. Thus, results for patients recruited early in the study are available for interpretation while patients are still being recruited and allocated to treatment. This feature allows the emerging evidence to be used to decide when to stop the study. In particular, it may be desirable to stop the study if a clear treatment difference is apparent, thereby avoiding the allocation of further patients to the less successful therapy. Investigators may also want to stop a study that no longer has much chance of demonstrating a treatment difference (Whitehead, 1992, 1997).
For example, consider the analysis of an intervention (countermeasure) to prevent the loss of bone mineral density in sequentially treated groups of astronauts resulting from their exposure to microgravity during space travel (Figure 3-1). The performance index is the bone mineral density (in grams per square centimeter) of the calcaneus. S refers to success, where p is the probability of success and p* is the cumulative mean. F refers to failure, where q is the probability of failure and q* is the cumulative mean. The confidence intervals for p and q are obtained after each space mission, that is, for p, (P1, P2), and for q, (q1, q2). The sequential accumulation of data then allows one to accept the countermeasure if p1 is greater than p* and q2 is less than q* or reject the countermeasure if p2 is less than p* or q1 is greater than q*. Performance indices will be acceptable when success S, a gain or mild loss, occurs on at least 75 percent (p* = 0.75) of the cases (astronaut missions) and when F, severe bone mineral density loss, occurs in no more than 5 percent (q* = 0.05) of the cases. Unacceptable performance indices occur with less than a 75 percent success rate or more than a 5 percent failure rate. As the number of performance indices increases, level 1 performance criteria can be set; for example, S is equal to a gain or no worse than 1 percent loss of bone mineral density relative to that at baseline. Indeterminate (I) is equal to a moderate loss of 1 to 2 percent from that at the baseline. F is equal to the severe loss of 2 percent or more from that at the baseline (Feiveson, 2000). See Box 1-2 for an alternate design discussion of this case study.
For example, in a randomized clinical trial for investigation of the effect of a selenium nutritional supplement on the prevention of skin cancer, it is determined that plasma selenium levels are not rising as expected in some patients in the supplemented group, indicating a possible noncompliance problem. In this case, the failure of some subjects to receive the prescribed amount of selenium supplement would have led to a loss of power to detect a significant benefit, if one was present. One could then initiate a prestudy treatment period in which potential noncompliers could be identified and eliminated from the study before randomization (Jennison and Turnbull, 1983).
Another reason for early examination of study results is to check the assumptions made when designing the trial. For example, in an experiment where the primary response variable is quantitative, the sample size is often set assuming this variable to be normally distributed with a certain variance. For binary response data, sample size calculations rely on an assumed value for the background incidence rate; for time-to-event data when individuals enter the trial at staggered intervals, an estimate of the subject accrual rate is important in determining the appropriate accrual period. An early interim analysis can reveal inaccurate assumptions in time for adjustments to be made to the design (Jennison and Turnbull, 1983).
760c119bf3