Estimates Of Treatment Differences Research Articles

Bayesian approach for a 2 x 2 crossover design with repeated measures: a simulation study

In crossover designs, the subjects receive all treatments, according to the groups of sequences formed. Therefore, if carryover effects are present in the model, inferences about the treatments effects become difficult. Furthermore, repeated measures of the response variable can be taken over time in the same experimental unit; however, these measures may be correlated. In this way, we aimed to analyze a 2 x 2 crossover design with repeated measures within the treatment period, using a Bayesian approach. A simulation study was performed to evaluate the performance. The posterior estimates of the model parameters were obtained under non-informative prior distributions and the normal likelihood function. The model performed well with a sample size of 20 subjects, showing even better results with samples of 100 subjects. With larger samples, exact tests for the differences in carryover effects and time effects were obtained. However, the test of time effect proved to be powerful even with small samples. In turn, considering carryover effects different from zero did not influence the estimates of treatment differences, although biased estimates of the period effect were obtained.

EPHex: a phase 3, double-blind, placebo-controlled, randomized study to evaluate long-term efficacy and safety of Oxalobacter formigenes in patients with primary hyperoxaluria

BackgroundPrimary hyperoxalurias (PHs) are rare genetic diseases that increase the endogenous level of oxalate, a waste metabolite excreted predominantly by the kidneys and also the gut. Treatments aim to improve oxalate excretion, or reduce oxalate generation, to prevent kidney function deterioration. Oxalobacter formigenes is an oxalate metabolizing bacterium. This Phase III, double-blind, placebo-controlled randomized trial investigated the effectiveness of orally administered Oxabact™, a lyophilized O. formigenes formulation, at reducing plasma oxalate levels in patients suffering from PH.MethodsSubjects (≥ 2 years of age) with a diagnosis of PH and maintained but suboptimal kidney function (mean estimated glomerular filtration rate at baseline < 90 mL/min/1.73 m2) were eligible to participate. Subjects were randomized to receive Oxabact or placebo twice daily for 52 weeks. Change from baseline in plasma oxalate concentration at Week 52 was the primary study endpoint.ResultsForty-three subjects were screened, 25 were recruited and one was discontinued. At Week 52, O. formigenes was established in the gut of subjects receiving Oxabact. Despite decreasing plasma oxalate level in subjects treated with Oxabact, and stable/increased levels with placebo, there was no significant difference between groups in the primary outcome (Least Squares mean estimate of treatment difference was − 3.80 μmol/L; 95% CI: − 7.83, 0.23; p-value = 0.064). Kidney function remained stable in both treatments.ConclusionsOxabact treatment may have stabilized/reduced plasma oxalate versus a rise with placebo, but the difference over 12 months was not statistically significant (p = 0.06). A subtle effect observed with Oxabact suggests that O. formigenes may aid in preventing kidney stones.Graphical abstractA higher resolution version of the Graphical abstract is available as Supplementary information.

Efficacy and safety of biosimilar CT-P17 versus reference adalimumab in subjects with rheumatoid arthritis: 24-week results from a randomized study

BackgroundTo demonstrate equivalent efficacy of the proposed high-concentration (100 mg/ml), citrate-free adalimumab biosimilar CT-P17 to European Union-approved adalimumab (EU-adalimumab) in subjects with active rheumatoid arthritis (RA).MethodsThis randomized, double-blind phase III study (ClinicalTrials.gov, NCT03789292) randomized (1:1) subjects with active RA at 52 centers to receive CT-P17 or EU-adalimumab 40 mg subcutaneously every 2 weeks until week 52. Results to week 24 are reported here. The primary endpoint was 20% improvement by American College of Rheumatology criteria (ACR20) response rate at week 24. Equivalence was concluded if the corresponding confidence intervals (CIs) for the estimate of treatment difference were within predefined equivalence margins: − 15 to 15% (95% CI; European Medicines Agency assumption); − 12 to 15% (90% CI; Food and Drug Administration assumption). Additional efficacy, pharmacokinetic, usability, safety, and immunogenicity endpoints were evaluated.Results648 subjects were randomized (324 CT-P17; 324 EU-adalimumab). The ACR20 response rate at week 24 was 82.7% (n = 268/324) in both groups (intention-to-treat population). The 95% CI (− 5.94 to 5.94) and 90% CI (− 4.98 to 4.98) were within predefined equivalence margins for both assumptions and equivalent efficacy was concluded. Additional endpoints and overall safety were comparable between groups. Mean trough serum concentrations of CT-P17 were slightly higher than those of EU-adalimumab. Immunogenicity was slightly lower numerically for the CT-P17 group than for the EU-adalimumab group.ConclusionsCT-P17 and EU-adalimumab have equivalent efficacy and comparable safety and immunogenicity in subjects with active RA. Overall safety of CT-P17 is consistent with the known safety profile of reference adalimumab.Trial registrationClinicalTrials.gov, NCT03789292. Registered 28 December 2018—retrospectively registered.

PNS339 BAYESIAN NETWORK META-ANALYSES WHERE INFORMATIVE PRIORS WERE USED FOR THE BETWEEN-STUDY HETEROGENEITY; A PRAGMATIC REVIEW

Network meta-analyses (NMAs) are commonly used to pool evidence from randomised controlled trials. Either a fixed effect or a random effect model can be used for analysis. Fixed effect models are often used when there are too few studies with which to estimate the between-study heterogeneity. Random effect models allow for between-study heterogeneity; however, in NMAs with few studies, uncertainty around point estimates are imprecisely estimated. Bayesian NMAs (BNMAs) using a random effect model and incorporating an informative prior for the between-study heterogeneity may offer a good solution. There is a need to understand how available methodologies, including those by Turner and colleagues, were being applied and on what basis the prior distributions were determined. Therefore, for this work, we aimed to review published BNMAs where informative priors were used for the between-study heterogeneity. A pragmatic literature review was conducted using PubMed to include BNMAs, published in the last 10 years, where informative priors were used for the between-study heterogeneity. Twenty-two BNMAs were included. The number of studies included in individual BNMAs ranged from 4 to 96 studies. The rationale for using a prior distribution for the between-study heterogeneity was not stated in most studies. Among studies justifying the approach, informative priors were employed in order to reduce imprecision and improve the estimation of treatment differences. Most studies applied the methodologies proposed by Turner and colleagues. In one study, the choice of prior distribution was based on comparing models of selected prior distributions using the deviance information criterion (DIC). A consistent approach regarding the use of prior distribution to inform between-study heterogeneity was used in the literature. Nevertheless, more advances in prior elicitation methodologies are likely to increase the use of informative priors in future BNMAs.

Opiate Use and Escalation of Care in Hospitalized Adults with Acute Heart Failure and Sleep-disordered Breathing (OpiatesHF Study)

Rationale: Sleep-disordered breathing (SDB) is highly prevalent in adults hospitalized with acute heart failure. Data are limited on the implications of inadvertent opiate use in this population.Objectives: To determine the prevalence and impact of in-hospital opiate use in adults hospitalized for acute heart failure.Methods: From a prospective sleep registry, we selected a sequential group of adult participants who were admitted to the hospital for acute heart failure and received a portable sleep study (PSS) after screening for SDB using the STOP-BANG questionnaire. A retrospective review of charts was performed to assess use of opiates, need for escalation of care (defined as transfer to the intensive care unit [ICU]), 30-day readmission, and length of stay. A logistic regression model was used to calculate propensity scores for each participant with a screening apnea-hypopnea index (AHI) greater than or equal to 10/h. Study endpoints, including escalation of care to the ICU and 30-day hospital readmission, were compared using a χ2 test with stabilized inverse probability-weighted propensity scores to control for potential confounding variables.Results: A total of 301 consecutive adults admitted with acute heart failure between November 2016 and October 2017 underwent PSS after SDB screening. Overall, 125 of 301 (41.5%) received opiates in the hospital, and 149 (49.5%) patients had an AHI greater than or equal to 10/h by PSS (high risk of SDB). In this high-risk group, 47 of 149 (32%) received opiates. Among those with an AHI greater than or equal to 10/h, escalation of care occurred in 12 of 47 (26%) of those who received opiates versus 4 of 102 (4%) of those who did not (P < 0.001; weighted estimate of treatment difference, 23.5%; 95% confidence interval [CI], 9.9 to 37.2). Similarly, readmission within 30 days occurred in 7 of 47 (15%) of those who received opiates versus 9 of 102 (9%) of those who did not (P = 0.14; weighted estimate of treatment difference, 8.3%; 95% CI, -4.0 to 20.6). Mean length of stay (days) did not differ between groups (P = 0.61; weighted estimate of treatment difference, -0.3 d; 95% CI, -1.4 to 0.8).Conclusions: In adults admitted with acute heart failure and found to be at high risk of SDB, opiate use in the hospital was highly prevalent and was associated with a greater likelihood of escalation of care.

A Placebo-controlled Study of PF-06473871 (Anti-Connective Tissue Growth Factor Antisense Oligonucleotide) in Reducing Hypertrophic Skin Scarring.

Connective tissue growth factor (CTGF) is a matricellular protein that plays a key role in wound healing and scar formation. Inhibition of CTGF by a specific antisense oligonucleotide significantly reduced scarring and fibrosis in animal models. This study examined whether an antisense oligonucleotide that inhibits human CTGF expression could reduce the severity of hypertrophic scar formation in patients following surgical revision of preexisting breast scars. This study was a 24-week multicenter, randomized, double-blind, within-subject, placebo-controlled phase 2b study evaluating the efficacy and safety of PF-06473871 in 2 regimens of either 3 or 4 intradermal injections (postsurgery weeks 2, 5, 8, and 11) of 5 mg/cm adjacent to the new surgical incision. One hundred subjects with bilateral hypertrophic scars resulting from prior breast surgery were randomized. Efficacy was determined by the Patient and Observer Scar Assessment Scale (POSAS). The Physician/Observer POSAS overall opinion score at (week 24) for the 4-injection regimen demonstrated a statistically significant (P = 0.022) treatment difference from placebo of 0.68, and the treatment difference for the 3-injection regimen was nonsignificant (P = 0.4). Physician evaluation of scar severity at (week 24) with the photo-guide in the 4-injection regimen had a significant reduction (point estimate of treatment difference of 0.43 favoring PF-06473871). The surgical effect was approximately 2.0 at week 24 and was nearly 3 times greater than the treatment effect. Patient evaluations using the POSAS and photo-guide were not significantly improved with either dose regimen. PF-06473871 was generally well tolerated systemically and locally. The 4-dose regimen of PF-06473871 provided statistically significant improvement, inhibiting severity of hypertrophic scar formation based on physician assessment. However, the effect of revision surgery alone is significant and may dominate the treatment effect of PF-06473871.

Use of weighted multivariate estimates in trials of multi-serotype vaccines to simplify interpretation of treatment differences

BackgroundMany vaccines contain multiple components. Licensed pneumococcal conjugate vaccines (PCV) contain polysaccharides from 7, 10, or 13 different serotypes of Streptococcus pneumoniae. The main outcomes in randomised trials of pneumococcal vaccines are serotype-specific antibody measures. Comparisons are made between groups for each serotype, resulting in multiple separate comparisons of treatment effects which can be complicated to interpret. We investigated methods for computing the overall difference between vaccine groups across all serotypes.MethodsPneumococcal antibody concentrations were obtained from a randomised controlled trial of ten-valent pneumococcal vaccine, conducted in Kathmandu, Nepal. Infants received either 2 priming doses of vaccine at 6 and 14 weeks of age followed by a booster (2+1), or 3 priming doses at 6, 10, and 14 weeks of age with no booster (3+0). The overall difference between vaccine schedules across all serotypes was computed at each visit using a multivariate linear model with equal weights for each serotype. Alternative weights were derived from invasive pneumococcal disease cases in Nepal, Bangladesh and Pakistan, and from estimates of the relative invasiveness of each serotype and used in sensitivity analyses.ResultsWhen 10 separate estimates of treatment differences were computed the ratio of antibody responses for each serotype in the 2+1 group compared with the 3+0 group at 10 months of age varied greatly, with serotype-specific GMRs ranging from 2.80 for serotype 14, to 9.14 for serotype 18C. Using equal weights for each serotype, the overall geometric mean ratio (GMR) was 5.02 (95% CI 4.06−6.22) at 10 months of age, and 1.46 (95% CI 1.14−1.88) at 3 years of age. Using weights based on disease incidence gave GMRs ranging from 5.15 to 6.63 at 10 months of age, and 1.47 to 1.78 at 3 years of age. Using weights based on relative invasiveness gave estimates of 6.81 and 1.59, at 10 months and 3 years respectively.ConclusionPCV clinical trial data have a multivariate structure with correlated outcomes for different serotypes. When analysing each serotype separately, the multiple estimates of the treatment effect can complicate the interpretation of trial results. Reporting a single overall estimate which accounts for the correlation between outcomes can simplify such interpretation. Treatment effects can be weighted equally or alternative weights derived from independent data can be used.Many modern vaccines have multiple components, such as quadrivalent meningococcal group ACWY vaccine or four-component group B meningococcal vaccine, thus these methods are widely applicable.

Clinical similarity of biosimilar ABP 501 to adalimumab in the treatment of patients with moderate to severe plaque psoriasis: A randomized, double-blind, multicenter, phase III study

ABP 501 is a biosimilar of adalimumab. We sought to compare the efficacy and safety of ABP 501 with adalimumab. This 52-week, double-blind study randomized patients with moderate to severe psoriasis to ABP 501 or adalimumab. At week 16, those with 50% or more improvement in Psoriasis Area and Severity Index score from baseline on ABP 501 continued the same treatment, whereas adalimumab-treated patients were rerandomized to adalimumab or ABP 501. Clinical similarity in Psoriasis Area and Severity Index percent improvement from baseline to week 16 (primary end point) was established if the point estimate of treatment difference and its 2-sided 95% confidence interval between groups was within equivalence margin of ±15. Patients, including those undergoing a single transition at week 16, were evaluated for safety and immunogenicity. Psoriasis Area and Severity Index percent improvement at week 16 was 80.9 for ABP 501 and 83.1for adalimumab (least-square mean difference -2.18 [95% confidence interval -7.39 to 3.02]). Adverseevents (67.2% [117/174] vs 63.6% [110/173]) and antidrug antibody incidence (55.2% [96/174] vs 63.6% [110/173]) for ABP 501 vs adalimumab were similar. Safety, including immunogenicity, was similar among groups after single transition (week 20). The 52-week data are not reported here. ABP 501 was shown to be clinically similar to adalimumab. Safety and immunogenicity were not impacted immediately after single transition (adalimumab to ABP 501).

Alirocumab in patients with heterozygous familial hypercholesterolaemia undergoing lipoprotein apheresis: the ODYSSEY ESCAPE trial.

Aim To evaluate the effect of alirocumab on frequency of standard apheresis treatments [weekly or every 2 weeks (Q2W)] in heterozygous familial hypercholesterolaemia (HeFH). Methods and results ODYSSEY ESCAPE (NCT02326220) was a double-blind study in 62 HeFH patients undergoing regular weekly or Q2W lipoprotein apheresis. Patients were randomly assigned (2:1, respectively) to receive alirocumab 150 mg (n = 41) or placebo (n = 21) Q2W subcutaneously for 18 weeks. From day 1 to week 6, apheresis rate was fixed according to the patient’s established schedule; from weeks 7 to 18, apheresis rate was adjusted based on the patient’s low-density lipoprotein cholesterol (LDL-C) response in a blinded fashion. Apheresis was not performed when the LDL-C value was ≥30% lower than the baseline (pre-apheresis) value. The primary efficacy endpoint was the rate of apheresis treatments over 12 weeks (weeks 7–18), standardized to number of planned treatments. In the alirocumab group the least square (LS) mean ± SE (95% confidence interval [CI]) per cent change in pre-apheresis LDL-C from baseline at week 6 was −53.7 ± 2.3 (−58.2 to − 49.2) compared with 1.6 ± 3.1 (–4.7 to 7.9) in the placebo group. The primary efficacy endpoint showed statistically significant benefit in favour of alirocumab (Hodges–Lehmann median estimate of treatment difference: 0.75; 95% CI 0.67–0.83; P < 0.0001). Therefore, alirocumab-treated patients had a 0.75 (75%) additional reduction in the standardized rate of apheresis treatments vs. placebo-treated patients. During this period, 63.4% of patients on alirocumab avoided all and 92.7% avoided at least half of the apheresis treatments. Adverse event rates were similar (75.6% of patients on alirocumab vs. 76.2% on placebo). Conclusions Lipoprotein apheresis was discontinued in 63.4% of patients on alirocumab who were previously undergoing regular apheresis, and the rate was at least halved in 92.7% of patients. Alirocumab was generally safe and well tolerated.

Missing CD4+ cell response in randomized clinical trials of maraviroc and dolutegravir

Background:Missing data can compromise inferences from clinical trials, yet the topic has received little attention in the clinical trial community. Shortcomings in commonly used methods used to analyze studies with missing data (complete case, last- or baseline-observation carried forward) have been highlighted in a recent Food and Drug Administration-sponsored report. This report recommends how to mitigate the issues associated with missing data. We present an example of the proposed concepts using data from recent clinical trials.Methods:CD4+ cell count data from the previously reported SINGLE and MOTIVATE studies of dolutegravir and maraviroc were analyzed using a variety of statistical methods to explore the impact of missing data. Four methodologies were used: complete case analysis, simple imputation, mixed models for repeated measures, and multiple imputation. We compared the sensitivity of conclusions to the volume of missing data and to the assumptions underpinning each method.Results:Rates of missing data were greater in the MOTIVATE studies (35%–68% premature withdrawal) than in SINGLE (12%–20%). The sensitivity of results to assumptions about missing data was related to volume of missing data. Estimates of treatment differences by various analysis methods ranged across a 61 cells/mm3 window in MOTIVATE and a 22 cells/mm3 window in SINGLE.Conclusions:Where missing data are anticipated, analyses require robust statistical and clinical debate of the necessary but unverifiable underlying statistical assumptions. Multiple imputation makes these assumptions transparent, can accommodate a broad range of scenarios, and is a natural analysis for clinical trials in HIV with missing data.

Psychotherapist effects in meta-analyses: How accurate are treatment effects?

Psychotherapists are known to vary in their effectiveness with their clients, in randomized clinical trials as well as naturally occurring treatment settings. The fact that therapists matter has 2 effects in psychotherapy studies. First, if therapists are not randomly assigned to modalities (which is rare) this may bias the estimation of the treatment effects, as the modalities may have therapists of differing skill. In addition, if the data are analyzed at the client level (which is virtually always the case) then the standard errors for the effect sizes will be biased due to a violation of the assumption of independence. Thus, the conclusions of many meta-analyses may not reflect true estimates of treatment differences. We reexamined 20 treatment effects selected from 17 meta-analyses. We focused on meta-analyses that found statistically significant differences between treatments for a variety of disorders by correcting the treatment effects according to the variability in outcomes known to be associated with psychotherapists. The results demonstrated that after adjusting the results based on most small estimates of therapist effects, ∼80% of the reported treatment effects would still be statistically significant. However, at larger estimates, only 20% of the treatment effects would still be statistically significant after controlling for therapist effects. Although some meta-analyses were consistent in their estimates for treatment differences, the degree of certainty in the results was considerably reduced after considering therapist effects. Practice implications for understanding treatment effects, namely, therapist effects, in meta-analyses and original studies are provided.

Number of imputations needed to stabilize estimated treatment difference in longitudinal data analysis.

Multiple imputation procedures replace each missing value with a set of plausible values based on the posterior predictive distribution of missing data given observed data. In many applications, as few as five imputations are adequate to achieve high efficiency relative to an infinite number of imputations. However, substantially more imputations are often needed to stabilize imputation-based inference at the analysis stage. Imputation-based inference at the analysis stage is considered stable if the conditional variability of the multiple imputation estimator, half-width of 95% confidence interval, test statistic, and estimated fraction of missing information given observed data is within specified thresholds for simulation error. For the estimation of treatment difference at study end for normally distributed responses in longitudinal trials, we calculate the multiple imputation quantities for an infinite number of imputations analytically and use simulations to assess the variability of the number of imputations needed at the analysis stage in repeated sampling.

FOLFOX4 plus cetuximab administered weekly or every second week in the first-line treatment of patients with KRAS wild-type metastatic colorectal cancer: a randomized phase II CECOG study

BackgroundThis randomized phase II study investigated first-line chemotherapy plus cetuximab administered every second week in KRAS wild-type metastatic colorectal cancer. Patients and methodsPatients received FOLFOX4 plus either standard weekly cetuximab (arm 1) or cetuximab (500 mg/m2) every second week (arm 2), until disease progression or unacceptable toxicity. Primary end point was the objective response rate (ORR). Progression-free survival (PFS), overall survival (OS), disease control rate (DCR) and safety were also investigated. The study was not powered to establish non-inferiority, but aimed at the estimation of treatment differences. ResultsOf 152 randomized eligible patients, 75 were treated in arm 1 and 77 in arm 2; ORRs [53% versus 62%, odds ratio 1.40, 95% confidence interval (CI) 0.74–2.66], PFS [median 9.5 versus 9.2 months, hazard ratio (HR) 0.92, 95% CI 0.63–1.34], OS (median 25.8 versus 23.0 months, HR 0.86, 95% CI 0.56–1.30) and DCR (87%) were comparable. HRs adjusted for baseline factors were 1.01 and 0.99 for PFS and OS, respectively. Frequencies of grade 3/4 adverse events in arms 1 versus 2 were similar: most common were neutropenia (28% versus 34%) and rash (15% versus 17%). ConclusionsActivity and safety of FOLFOX4 plus either cetuximab administered weekly or every second week were similar.

FRI0199 In patients with established RA, abatacept efficacy is independent of baseline annual radiographic progression rate

BackgroundIn the Phase III, double-blind, placebo (PBO)-controlled AIM study, abatacept (ABA)+ MTX significantly inhibited structural damage progression vs PBO+MTX in pts with established RA and inadequate response to MTX.1Sustained inhibition...

Assessing the effects of agricultural management practices on carbon fluxes: Spatial variation and the need for replicated estimates of Net Ecosystem Exchange

Replicated measurements of Net Ecosystem CO 2 Exchange (NEE) were made on four arable plots using two eddy covariance (EC) towers that were moved at regular intervals. Replicated plots were created by dividing a large arable field into four approximately equal areas (∼2.5 ha). Regular adjustments of the EC measurement height (1.5–1.9 m) due to seasonal changes in canopy height ensured that the analyzed fluxes were derived from the individual plots. This was sufficient to make high-quality flux measurements with the energy balance accounting for up to 96% of the available energy in the system. The replicated experiment included two treatment plots managed with conventional tillage practices (CON) and two plots managed with non-inversion tillage (NIT). The mean annual NEE (aNEE) over all plots and years showed a carbon uptake of –163 g [C] m −2 y −1. Plot to plot variability ranged between 9 and 70 g [C] m −2 y −1. Differences in leaf area index (LAI) were the most likely cause of plot to plot variability. For the majority of the experiment spatial variation was greater than treatment effects and provided over half the total annual uncertainty, demonstrating the need to address spatial variability when assessing aNEE using EC techniques. Interannual variation in the measured fluxes was greater than the plot-to-plot or the treatment variations, most likely due to differences in water availability, although this could be confounded by changes in spatial variability due to other environmental factors or management operations. The proportionally large uncertainty associated with spatial and interannual variability when compared to the measured fluxes, suggests that NEE estimates are, in the short term, unlikely to result in sufficiently accurate estimates of net biome productivity (NBP) for assessing the slow accumulation/loss of SOC associated with altered tillage management without a much larger number of replicated plots. Whilst the approach used has limited statistical power it could be modified to accommodate a large number of replicates to provide more robust estimates of treatment differences.

Meta-analysis of individual patient data versus aggregate data from longitudinal clinical trials

In clinical trials following individuals over a period of time, the same assessment may be made at a number of time points during the course of the trial. Our review of current practice for handling longitudinal data in Cochrane systematic reviews shows that the most frequently used approach is to ignore the correlation between repeated observations and to conduct separate meta-analyses at each of a number of time points. The purpose of this paper is to show the link between repeated measurement models used with aggregate data and those used when individual patient data (IPD) are available, and provide guidance on the methods that practitioners might use for aggregate data meta-analyses, depending on the type of data available. We discuss models for the meta-analysis of longitudinal continuous outcome data when IPD are available. In these models time is included either as a factor or as a continuous variable, and account is taken of the correlation between repeated observations. The meta-analysis of IPD can be conducted using either a one-step or a two-step approach: the latter involves analysing the IPD separately in each study and then combining the study estimates taking into account their covariance structure. We discuss the link between models for use with aggregate data and the two-step IPD approach, and the problems which arise when only aggregate data are available. The methods are applied to IPD from 5 trials in Alzheimer's disease. Two major issues for the meta-analysis of aggregate data are the lack of information about correlation coefficients and the effect of missing data at the patient-level. Application to the Alzheimer's disease data set shows that ignoring correlation can lead to different pooled estimates of the treatment difference and their standard errors. Furthermore, the amount of missing data at the patient level can affect these estimates. The models assume fixed treatment effects across studies, and that any missing data is missing at random, both at the patient-level and the study level. It is preferable to obtain IPD from all studies to correctly account for the correlation between repeated observations. When IPD are not available, the ideal aggregate data are model-based estimates of treatment difference and their variance and covariance estimates. If covariance estimates are not available, sensitivity analyses should be undertaken to investigate the robustness of the results to different amounts of correlation.

A Modification of the 50%-Conditional Power Approach for Increasing the Sample Size Based on an Interim Estimate of Treatment Difference

Recently, flexible approaches with updating of sample size during the course of clinical trials have been proposed; the weighted Z-statistic approach and the 50 %-conditional power approach. In this paper, we propose a modification of the 50 %-conditional power approach, which increases the sample size only when the conditional power based on the unblinded interim results is greater than 50 %. Our method can control the type I error rate due to the restriction on the minimum required sample size ratio under the decision of increasing sample size. Simulation studies showed that the proposed method increased power about 10 % compared with the fixed sample size design and attained higher power than the original 50 %-conditional power approach. Compared with the weighted Z-statistic approach, the proposed method had several promising operating characteristics; a substantial gain in conditional power given the decision of sample size adjustment, a low probability of reaching the maximum sample size, a substantial decrease in the conditional type II error rate given the maximum sample size, and a conservative property of not increasing sample size erroneously under no treatment effect.

Randomization of neighbour balanced generalized Youden designs

If a crossover design with more than two treatments is carryover balanced, then the usual randomization of experimental units and periods would destroy the neighbour structure of the design. As an alternative, Bailey [1985. Restricted randomization for neighbour-balanced designs. Statist. Decisions Suppl. 2, 237–248] considered randomization of experimental units and of treatment labels, which leaves the neighbour structure intact. She has shown that, if there are no carryover effects, this randomization validates the row–column model, provided the starting design is a generalized Latin square. We extend this result to generalized Youden designs where either the number of experimental units is a multiple of the number of treatments or the number of periods is equal to the number of treatments. For the situation when there are carryover effects we show for so-called totally balanced designs that the variance of the estimates of treatment differences does not change in the presence of carryover effects, while the estimated variance of this estimate becomes conservative.

Multiple imputation compared with restricted pseudo‐likelihood and generalized estimating equations for analysis of binary repeated measures in clinical studies

AbstractNon‐likelihood‐based methods for repeated measures analysis of binary data in clinical trials can result in biased estimates of treatment effects and associated standard errors when the dropout process is not completely at random. We tested the utility of a multiple imputation approach in reducing these biases. Simulations were used to compare performance of multiple imputation with generalized estimating equations and restricted pseudo‐likelihood in five representative clinical trial profiles for estimating (a) overall treatment effects and (b) treatment differences at the last scheduled visit. In clinical trials with moderate to high (40–60%) dropout rates with dropouts missing at random, multiple imputation led to less biased and more precise estimates of treatment differences for binary outcomes based on underlying continuous scores. Copyright © 2005 John Wiley &amp; Sons, Ltd.

Estimation of Treatment Difference and Standard Deviation with Blinded Data in Clinical Trials

AbstractIn an ongoing clinical trial and while the randomized treatment codes remain blinded, it is often desirable to estimate the treatment difference and standard deviation for a normally‐distributed response variable. This is particularly useful for estimating the sample size for future trials or for adjusting the sample size for the ongoing trial. We describe the limitations of an available EM algorithm‐based procedure to reestimate the standard deviation without unblinding the codes for the two treatments. We introduce a new procedure and propose a clinical trial design for estimating both the treatment difference and standard deviation without unblinding. The performance of the proposed procedure is evaluated in a simulation study.