Conflict of Interest
Request Clinical TrialsAssociated Trial(s):
- NCT01106625 - A Randomized, Double-Blind, Placebo-Controlled, 3-Arm, Parallel-Group, Multicenter Study to Evaluate the Efficacy, Safety, and Tolerability of Canagliflozin in the Treatment of Subjects With Type 2 Diabetes Mellitus With Inadequate Glycemic Control on Metformin and Pioglitazone Therapy
- NCT01064414 - A Randomized, Double-Blind, Placebo-Controlled, 3-Arm, Parallel-Group, 26-Week, Multicenter Study With a 26-Week Extension, to Evaluate the Efficacy, Safety and Tolerability of Canagliflozin in the Treatment of Subjects With Type 2 Diabetes Mellitus Who Have Moderate Renal Impairment
- NCT01081834 - A Randomized, Double-Blind, Placebo-Controlled, Parallel-Group, Multicenter Study to Evaluate the Efficacy, Safety, and Tolerability of Canagliflozin as Monotherapy in the Treatment of Subjects With Type 2 Diabetes Mellitus Inadequately Controlled With Diet and Exercise
- NCT01106677 - A Randomized, Double-Blind, Placebo and Active-Controlled, 4-Arm, Parallel Group, Multicenter Study to Evaluate the Efficacy, Safety, and Tolerability of Canagliflozin in the Treatment of Subjects With Type 2 Diabetes Mellitus With Inadequate Glycemic Control on Metformin Monotherapy
- NCT00968812 - A Randomized, Double-Blind, 3-Arm Parallel-Group, 2-Year (104-Week), Multicenter Study to Evaluate the Efficacy, Safety, and Tolerability of JNJ-28431754 Compared With Glimepiride in the Treatment of Subjects With Type 2 Diabetes Mellitus Not Optimally Controlled on Metformin Monotherapy
- NCT01106651 - A Randomized, Double-Blind, Placebo-Controlled, Parallel-Group, Multicenter Study to Evaluate the Efficacy, Safety, and Tolerability of Canagliflozin Compared With Placebo in the Treatment of Older Subjects With Type 2 Diabetes Mellitus Inadequately Controlled on Glucose Lowering Therapy
- NCT01106690 - A Randomized, Double-Blind, Placebo-Controlled, 3-Arm, Parallel-Group, Multicenter Study to Evaluate the Efficacy, Safety, and Tolerability of Canagliflozin in the Treatment of Subjects With Type 2 Diabetes Mellitus With Inadequate Glycemic Control on Metformin and Sulphonylurea Therapy
- NCT01137812 - A Randomized, Double-Blind, Active-Controlled, Multicenter Study to Evaluate the Efficacy, Safety, and Tolerability of Canagliflozin Versus Sitagliptin in the Treatment of Subjects With Type 2 Diabetes Mellitus With Inadequate Glycemic Control on Metformin and Sulphonylurea Therapy
Request Clinical Trials
Data Request StatusStatus: Withdrawn/Closed
Project Title: Making Better Use of Randomized Trials: Assessing Applicability and Transporting Causal Effects
Use of RCTs to inform clinical and policy decisions, even when trials are impeccably designed and conducted, requires the application of trial results to patients who differ from trial participants in significant ways.
Our first aim is to develop methods for assessing the applicability of RCTs and transporting causal effects from RCTs to target populations; describe methods for assessing the robustness of inferences based on these methods. Second, to evaluate the performance of methods developed in Aim 1 in simulation studies that will reflect real-world data. Third, to empirically assess the methods developed under Aims 1 and 2 in high-impact clinical topics.
Our approach combines causal inference theory, computer simulation studies, and empirical analyses of trial and observational data on high-impact clinical areas. This comprehensive approach is designed to address the research gaps identified in Section A.3.
Participants include any that meet the inclusion criterion for their respective RCT. This methods study will be compiling a database of RCTS with at least 100 participants to perform the specified methods on.
Main Outcome Measure(s);
The main outcomes measures include any primary or secondary outcomes specified in the protocol of the respective RCT, and will be extrapolated during the analysis.
The PS is the probability of being treated conditional on pre-treatment covariates. We will use logistic regression fit with standard
maximum likelihood methods to estimate the PS.
Brief Project Background and Statement of Project Significance: The randomized clinical trial (RCT) is widely recognized as the preferred study design for assessing the efficacy of treatments and treatment policies. The advantages of RCTs stem from randomization (which renders the compared treatment groups ?similar?), but also from the use of a priori specified criteria to select patients for participation, and the use of carefully planned followup procedures for ascertaining outcomes. In addition, randomization justifies simple, assumption-free approaches for the statistical analysis of trial data.2 However, for practical and ethical reasons, RCTs cannot be conducted in all target populations where specific interventions are considered for use.3 Moreover, stringent selection criteria and patients? reluctance to participate in trials render RCT populations unrepresentative of those seen in clinical practice (e.g., women, ethnic/racial minorities, the elderly, and patients with multiple comorbidities are often underrepresented). Numerous systematic studies have demonstrated that most trials lack external validity and enroll patients who differ in various important characteristics compared with patients to whom trial findings are applied. Study design-based approaches for improving the external validity of RCTs include the conduct of randomized experiments in representative samples (?population-based survey experiments?) and pragmatic trials (large trials with simple inclusion criteria and treatment protocols). Survey experiments are exceedingly rare for medical interventions because of their limited feasibility in clinical care settings. Pragmatic trials, despite their strengths, do not fully address issues of generalizability because questions about the effectiveness of interventions in populations different from those enrolled in a pragmatic trial always arise (e.g., how do the results of a large simple trial conducted in Europe apply in the U.S.?) and because they are logistically challenging to conduct. Furthermore, inclusion criteria that aim to make trial samples representative of a population may not be the most efficient way for addressing certain research questions (e.g., for confirmatory studies of heterogeneity of treatment effects, HTE, it is often more efficient to sample participants in a way that maximizes the study?s statistical power to detect between-subgroup differences). In fact, it has been argued that representativeness in ?designed? epidemiological studies (including RCTs) that address causal relationships should be avoided in most cases.12-14 After all, even if patients invited to participate in a trial are representative of the target population, those who consent to randomization may still be unrepresentative of the target. Thus, even when RCTs have high internal validity, ?real world? decisions always have to rely on the extrapolations results to a target population. Use of RCTs to inform clinical and policy decisions, even when trials are impeccably designed and conducted, requires the application of trial results to patients who differ from trial participants in significant ways.
Specific Aims of the Project:
Our long-term objective is to optimize the translation of clinical trial results into clinical practice.
By leveraging advances in causal inference and considerable existing data resources we will achieve the following
Aim 1: Develop methods for assessing the applicability of RCTs and transporting causal effects from RCTs to target
populations; describe methods for assessing the robustness of inferences based on these methods.
Aim 2: Evaluate the performance of methods developed in Aim 1 in simulation studies that will reflect real-world data.
Aim 3: Empirically assess the methods developed under Aims 1 and 2 in two high-impact clinical topics: (a) thrombolysis
for the treatment of acute ischemic stroke; and (b) revascularization strategies for chronic coronary artery disease (CAD).
What is the purpose of the analysis being proposed? Please select all that apply.: New research question to examine treatment effectiveness on secondary endpoints and/or within subgroup populations
Data Source and Inclusion/Exclusion Criteria to be used to define the patient sample for your study:
To address Aim 1 we will perform the following tasks:
(1) compile a dataset of RCTs;
(2) design observational studies corresponding to the RCTs;
(3) analyze the RCT data;
(4) analyze the observational data;
(5) compare the results of RCT and observational data analyses and synthesize findings across topics. We detail the
methods for each of these tasks below.
Task 1, Compile a dataset of RCTs: We have obtained individual patient data from 30 RCTs* in
cerebrovascular and cardiovascular disease (see preliminary work, Box 1). We have collected details of the
patient selection criteria through associated publications and we have been able to reconstruct the original
papers? analysis plans and (numerically) reproduce their results.
We will expand our database by including non-cerebrovascular/cardiovascular trials and using a
sample size cut-off of 500 enrolled patients. During the funding period we will continue to monitor our sources
(NHLBI, NIDDK and Trials) to identify any new studies that meet our criteria. Furthermore, we have identified
additional sources of RCT databases.
* the trials requested through the YODA Project are in addition to the 30 identified RCTS
Primary and Secondary Outcome Measure(s) and how they will be categorized/defined for your study:
We will also evaluate and report the distribution of predicted risk in treated and untreated patients, in the observational and
randomized data. We will summarize predicted baseline risk using non-parametric estimates of the empirical
distribution of outcome risk and descriptive statistics (mean, median, minimum, maximum, and 25th and 75th
percentile). We will also compare the ratio of the outcome risk in the extreme quartiles, the median-to-mean
risk ratio, and use Pearson?s median skewness coefficient.
Main Predictor/Independent Variable and how it will be categorized/defined for your study: When information on the predicted response for individuals is not reliable, our methods may still be able to provide answers for the target population and its subgroups: Some of the proposed methods do not rely on modeling response to treatment and may be useful when predictive outcome modeling is not feasible, e.g., due to the low number of events in the trial or the lack of background information on outcome predictors. These methods rely instead on the predicted probability of participating in the trial (and thus have a greater number of outcomes equal to the number of participants in the trial) and can be used to obtain population and subgroup-specific estimates of the treatment effect in the target population. Importantly, because the two approaches rely on different statistical modeling assumptions (even though the underlying structural models are the same), they can be used as stability analyses for addressing the same research question when data are available for implementing both.
Other Variables of Interest that will be used in your analysis and how they will be categorized/defined for your study:
Statistical Analysis Plan:
Transporting causal effects: We will transport causal effects using (1) probability of participation-based estimators (both IPPW and stratification); (2) outcome-model based estimators; (3) and doubly robust estimators, with a focus on the best performing estimators identified in simulation studies (Aim 2).
(a) Empirical evaluation of methods across trials: for groups of trials of sufficiently similar interventions, comparators, outcomes, and patient population we will use each trial as a possible ?target? population. We will then apply probability of participation- and outcome model-based, and doubly robust estimators in the remaining trials to transport their effects to the target (the left-one-out trial). Importantly, this analysis allows for formal empirical evaluation because in the left-out trial we can estimate treatment effects using information on the actual exposure and outcome data, and we can compare this estimate with the ?transported? estimates. This analysis is somewhat weakened by the fact that trials on the same topic area are not identical, but we believe that trials deemed similar enough to be included in the same meta-analyses are at least as similar between them, as each of them is similar to an observational study representing a target population.44,93 Results from these analyses will be interpreted qualitatively, with reference to the particular characteristics of the trials and in view of the results of our simulations.
(b) Assessment of performance using both randomized trials and observational studies: We will apply the same methods to transport the effects from each randomized study to a common target population represented by the registry data. Again, comparisons will be interpreted qualitatively, with reference to the particular characteristics of the trials and target populations, and in view of the results of the simulation studies in Aim 2. Analyses will examine (1) the average causal effect in the overall target population (in all analyses); (2) the average causal effect on the treated (when transporting results to target populations represented by observational data; this causal effect is equal with the average causal effect in the overall population when transporting results to target populations represented by an observational study as in (a) above); (3) the conditional average causal effect in subgroups of the population (in all analyses). For analyses focusing on conditional average effects, we will examine the effect in subgroups defined by sex (male vs. female), age (above or below median in the trial), and race/ethnicity (depending on data availability; to be determined at the protocol stage of each analysis). Comparisons between subgroups will be based on formal interaction tests and estimates of relative treatment effects (comparing subgroups); results will be reported in a standardized manner. If additional exploratory analyses are undertaken they will be reported in full (for descriptive purposes).
Narrative Summary: Use of randomized clinical trials (RCTs) to inform clinical and policy decisions?even when trials are impeccably designed and conducted?requires the application of trial results to patients who differ from trial participants in significant ways. We propose to develop and evaluate methods for assessing the applicability of RCT results and transporting trial results to broader target populations. A stakeholder panel of patients, clinicians, methodologists, guideline developers, and representatives of federal agencies will participate in the design and conduct of this project.
Table 3: Project deliverables and delivery dates. Stakeholder calls are not included in this table.
Finalize list of participants
Finalized list of stakeholder partners (including those currently marked as TBN).
November 30, 2015
Develop a toolbox for assessing applicability and transporting causal effects
Formulate an overarching framework for assessing applicability and transporting causal effects from trials to populations. Develop estimators for transporting causal effects that rely on different modeling assumptions, describe methods for sensitivity analysis. Manuscripts to disseminate the framework.
October 31, 2016
Signed DUAs for all RCT and observational datasets to be used for Aim 3.
October 31, 2016
Evaluate proposed methods in simulation studies
Completed simulation studies comparing alternative methods for assessing applicability and alternative estimators for transporting causal effects from trials to target populations. Manuscripts describing the simulation studies.
October 31, 2017
Empirical analyses for trials of acute stroke ? applicability
Completed analyses on the applicability of randomized clinical trials of stroke thrombolysis to the target population (OSR). Manuscript describing the analyses.
October 31, 2017
This study combines characteristics of studies of observational data, RCT analyses, and evidence syntheses.
We will report our findings according to rigorous reporting standards, as applicable. Immediately available avenues for dissemination: Although PCORI guidelines specify that we are not expected to undertake dissemination efforts for the proposed work, we believe that our research team is well positioned
to disseminate research findings widely and across different research communities. Dahabreh is the Associate
Director of the AHRQ-funded Brown Evidence-based Practice Center (EPC), one of 11 centers funded by the
AHRQ to conduct rigorous systematic reviews (of both RCT and observational evidence), which are used to
inform healthcare policy and clinical practice guideline development. Through his affiliation with the EPC
program Dahabreh can disseminate our findings to a group of leading researchers (and other EPC
investigators) engaged in the evaluation of clinical research evidence. For example, from September 2013,
Dahabreh will be leading an cross-EPC workgroup aiming to develop guidance for the integration of
observational and RCT evidence. Please see page 21 of ResearchPlan_Dahabreh.
PCORI Research Plan ? Improving Methods
1. Greenland S, Robins JM. Identifiability, exchangeability, and epidemiological confounding. International journal of epidemiology 1986;15:413-9.
2. Fisher RA. The design of experiments. 1935.
3. Dahabreh IJ, Kent DM. Can the learning health care system be educated with observational data? JAMA 2014;312:129-30.
4. Steg PG, Lopez-Sendon J, Lopez de Sa E, et al. External validity of clinical trials in acute myocardial infarction. Arch Intern Med 2007;167:68-73.
5. Lee PY, Alexander KP, Hammill BG, Pasquali SK, Peterson ED. Representation of elderly persons and women in published randomized trials of acute coronary syndromes. JAMA 2001;286:708-13.
6. Bjorklund E, Lindahl B, Stenestrand U, et al. Outcome of ST-elevation myocardial infarction treated with thrombolysis in the unselected population is vastly different from samples of eligible patients in a large-scale clinical trial. Am Heart J 2004;148:566-73.
7. Bahit MC, Cannon CP, Antman EM, et al. Direct comparison of characteristics, treatment, and outcomes of patients enrolled versus patients not enrolled in a clinical trial at centers participating in the TIMI 9 Trial and TIMI 9 Registry. Am Heart J 2003;145:109-17.
8. Yuasa H, Kurita K, Westesson PL. External validity of a randomised clinical trial of temporomandibular disorders: analysis of the patients who refused to participate in research. Br J Oral Maxillofac Surg 2003;41:129-31.
9. Travers J, Marsh S, Williams M, et al. External validity of randomised controlled trials in asthma: to whom do the results of the trials apply? Thorax 2007;62:219-23.
10. Travers J, Marsh S, Caldwell B, et al. External validity of randomized controlled trials in COPD. Respir Med 2007;101:1313-20.
11. Rothman K, Hatch E, Gallacher J. Representativeness is not helpful in studying heterogeneity of effects across subgroups. International journal of epidemiology 2014;43:633-4.
12. Rothman KJ, Gallacher JE, Hatch EE. Why representativeness should be avoided. International journal of epidemiology 2013;42:1012-4.
13. Elwood JM. Commentary: on representativeness. International journal of epidemiology 2013;42:1014-5.
14. Richiardi L, Pizzi C, Pearce N. Commentary: Representativeness is usually not necessary and often should be avoided. International journal of epidemiology 2013;42:1018-22.
15. Rothwell PM. External validity of randomised controlled trials: “to whom do the results of this trial apply?”. Lancet 2005;365:82-93.
16. Dekkers OM, von Elm E, Algra A, Romijn JA, Vandenbroucke JP. How to assess the external validity of therapeutic trials: a conceptual approach. Int J Epidemiol 2010;39:89-94.
17. Bickel PJ. A distribution free version of the Smirnov two sample test in the p-variate case. The Annals of Mathematical Statistics 1969:1-23.
18. Rosenbaum PR. An exact distribution?free test comparing two multivariate distributions based on adjacency. Journal of the Royal Statistical Society: Series B (Statistical Methodology) 2005;67:515-30.
19. Stuart EA, Cole SR, Bradshaw CP, Leaf PJ. The use of propensity scores to assess the generalizability of results from randomized trials. J R Stat Soc Ser A Stat Soc 2001;174:369-86.
20. Steel D. Across the boundaries: Extrapolation in biology and social science: Oxford University Press Oxford; 2008.
21. Kent DM, Rothwell PM, Ioannidis JP, Altman DG, Hayward RA. Assessing and reporting heterogeneity in treatment effects in clinical trials: a proposal. Trials 2010;11:85.
22. Shadish WR, Cook TD, Campbell DT. Experimental and quasi-experimental designs for generalized causal inference: Wadsworth Cengage learning; 2002.