Instrumental variable meta-analysis of individual patient data: application to adjust for treatment non-compliance

A systematic review found that deviations from intention-to-treat are common in randomized trials and systematic reviews

OBJECTIVES

To describe the characteristics, and estimate the incidence, of trials included in systematic reviews deviating from the intention-to-treat (ITT) principle.

STUDY DESIGN AND SETTING

A 5% random sample of reviews were selected (Medline 2006-2010). Trials from reviews were classified based on the ITT: (1) ITT trials (trials reporting standard ITT analyses); (2) modified ITT (mITT) trials (modified ITT; trials deviating from standard ITT); or (3) no ITT trials.

RESULTS

Of 222 reviews, 81 (36%) included at least one mITT trial. Reviews with mITT trials were more likely to contain trials that used placebo, that investigated drugs, and that reported favorable results. The incidence of reviews with mITT trial ranged from 29% (17/58) to 48% (23/48). Of the 2,349 trials, 597 (25.4%) were classified as ITT trials, 323 (13.8%) as mITT trials, and 1,429 (60.8%) as no ITT trials. The mITT trials were more likely to have reported exclusions compared to studies classified as ITT trials and to have received funding.

CONCLUSION

The reporting of the type of ITT may differ according to the clinical area and the type of intervention. Deviation from ITT in randomized controlled trials is a widespread phenomenon that significantly affects systematic reviews.

Effect of a Behavioral Self-Regulation Intervention on Patient Adherence to Fluid-Intake Restrictions in Hemodialysis: a Randomized Controlled Trial

OBJECTIVE

The purpose of this study is to evaluate the efficacy of a behavioral self-regulation intervention vs. active control condition using a parallel-group randomized clinical trial with a sample of center hemodialysis patients with chronic kidney disease.

METHOD

Participants were recruited from 8 hemodialysis treatment centers in the Midwest. Eligible patients were (a) fluid nonadherent as defined by an interdialytic weight gain >2.5 kg over a 4-week period, (b) >18 years of age, (c) English-speaking without severe cognitive impairment, (d) treated with center-based hemodialysis for >3 months, and (e) not living in a care facility in which meals were managed. Medical records were used to identify eligible patients. Patients were randomly assigned to either a behavioral self-regulation intervention or active control condition in which groups of 3-8 patients met for hour-long, weekly sessions for 7 weeks at their usual hemodialysis clinic. Primary analyses were intention-to-treat.

RESULTS

Sixty-one patients were randomized to the intervention while 58 were assigned to the attention-placebo support and discussion control. Covariate-adjusted between-subjects analyses demonstrated no unique intervention effect for the primary outcome, interdialytic weight gain (β = 0.13, p = 0.48). Significant within-subjects improvement over time was observed for the intervention group (β = -0.32, p = 0.014).

CONCLUSIONS

The present study found that participation in a behavioral self-regulation intervention resulted in no unique intervention effect on a key indicator of adherence for those with severe chronic kidney disease. There was, however, modest within-subjects improvement in interdialytic weight gain for the intervention group which meshes with other evidence showing the utility of behavioral interventions in this patient population. ClinicalTrials.gov Identifier: NCT01066949.

Adjustment for Variable Adherence Under Hierarchical Structure: Instrumental Variable Modeling Through Compound Residual Inclusion

BACKGROUND

Variable adherence to assigned conditions is common in randomized clinical trials.

OBJECTIVES

A generalized modeling framework under longitudinal data structures is proposed for regression estimation of the causal effect of variable adherence on outcome, with emphasis upon adjustment for unobserved confounders.

RESEARCH DESIGN

A nonlinear, nonparametric random-coefficients modeling approach is described. Estimates of local average treatment effects among compliers can be obtained simultaneously for all assigned conditions to which participants are randomly assigned within the trial. Two techniques are combined to address time-varying and time-invariant unobserved confounding-residual inclusion and nonparametric random-coefficients modeling. Together these yield a compound, 2-stage residual inclusion, instrumental variables model.

SUBJECTS

The proposed method is illustrated through a set of simulation studies to examine small-sample bias and in application to neurocognitive outcome data from a large, multicenter, randomized clinical trial in sleep medicine for continuous positive airway pressure treatment of obstructive sleep apnea.

RESULTS

Results of simulation studies indicate that, relative to a standard comparator, the proposed estimator reduces bias in estimates of the causal effect of variable adherence. Bias reductions were greatest at higher levels of residual variance and when confounders were time varying.

CONCLUSIONS

The proposed modeling framework is flexible in the distributions of outcomes that can be modeled, applicable to repeated measures longitudinal structures, and provides effective reduction of bias due to unobserved confounders.

Adjustment for compliance behavior in trials of epidural analgesia in labor using instrumental variable meta-analysis

OBJECTIVES

Intention-to-treat (ITT) analysis of randomized controlled trials (RCTs) may cause bias when compliance is poor. Noncompliance describes failure to comply with allocation in the intervention arm, and contamination describes uptake of the intervention in the control arm. Instrumental variable (IV) analysis can be applied in addition to the primary ITT analysis to estimate the causal effect adjusted for noncompliance and contamination, assuming that noncompliers would have had the same treatment benefit as compliers. We aimed to compare ITT and IV meta-analysis of the association between epidural analgesia in labor and cesarean section.

STUDY DESIGN AND SETTING

The study was restricted to 27 trials in a Cochrane Systematic Review. The association between epidural analgesia in labor and cesarean section was calculated using ITT and IV analyses. Pooled risk ratios (RRs) were calculated using fixed-effects meta-analysis.

RESULTS

In 18 trials with compliance data, noncompliance was 23% and contamination was 27%. In 10 trials with outcome data stratified by compliance, the pooled RR for cesarean section following epidural analgesia was 1.37 [95% confidence interval (CI): 1.00, 1.89; P = 0.049] using IV compared with 1.19 (95% CI: 0.93, 1.51; P = 0.16) using ITT.

CONCLUSION

ITT meta-analysis underestimates the effect of receiving epidural analgesia in labor on cesarean section compared with IV meta-analysis.

Randomised controlled trials and clinical maternity care: moving on from intention-to-treat and other simplistic analyses of efficacy

BACKGROUND

The obstetrical literature is dominated by Randomised Controlled Trials (RCTs), with the vast majority being analysed using an intention-to-treat (ITT) approach. Whilst this approach may reflect well the consequence of assignment to therapy and hence the 'trialists'perspective', it may fail to address the consequence of actually receiving therapy (the patient's perspective).

DISCUSSION

This review questions the ubiquitous adherence to the ITT approach, and gives examples of where this may have misled the maternity care professions. It gives an overview of techniques to overcome potential deficiencies in result presentation, using method effectiveness models such as 'Per Protocol' (PP) or 'As-Treated' (AT) that may give more accurate clinical meaning to the presentation of obstetrical results. It then proceeds to cover the added benefits, considerations and potential pitfalls of the use of Instrumental Variable (IV) models in order to better reflect the clinical context.

SUMMARY

While ITT may achieve statistical purity, it frequently fails to address the true clinical or patient's perspective. Though more complex and potentially beset by problems of their own, alternative methods of result presentation may better serve the latter aim. Each of the other methods may rely on untestable assumptions and therefore it is wisest that study results are presented in multiple formats to allow for informed reader evaluation.