Table 2 Title, author, year, journal and a summary of objectives of the papers included in this review. 2SPS – Two-stage predictor substitution; 2SRI – Two-stage residual inclusion; 2SLS – Two-stage least squares; RCT – Randomised controlled trial; IV – Instrumental variables; ITT – Intention to treat; AT – As treated; PP – Per protocol; CALM – Causal accelerated life model; C-Prophet - Compliers proportional hazards effect of treatment; CHARM - Causal hazard ratio adjustment regression model; CACE – Complier average causal effect; LTGM – Latent treat grizzle model; HR – Hazard ratio
Author(s) | Publication Year | Title | Journal | Main aims and objectives |
|---|---|---|---|---|
Agbla et al. [23] | 2020 | Estimating cluster-level average treatment effect in cluster randomised trials with non-adherence | Statistical Methods in Medical Research | o Comparing alternative estimation strategies for the implementation of 2SLS estimation using cluster-level data. o Demonstrate that using individual-level covariate-adjusted cluster summaries in the (weighted) 2SLS regression can increase efficiency. |
Bang and Davis [31] | 2007 | On estimating treatment effects under non-compliance in randomized clinical trials: Are intent-to-treat or instrumental variables analyses perfect solutions? | Statistics in Medicine | o Compare the performance of four estimators that are conventionally considered for treatment effect estimation under different non-compliance scenarios in a typical clinical trial setting under simulation. |
Cai et al. [32] | 2011 | Two-stage instrumental variable methods for estimating the causal odds ratio: Analysis of bias | Statistics in Medicine | o Present analytical and simulation results for the bias of 2SPS and 2SRI estimators under a causal logistic model expressed in terms of potential outcomes under the principal stratification framework. |
Cuzick et al. [21] | 1997 | Adjusting for non-compliance and contamination in randomized clinical trials | Statistics in Medicine | o Study a method of analysis which estimates the magnitude of the treatment effect among compliers in a randomized study in such a way as to respect the randomization and still be valid even when compliers have a different baseline risk than non-compliers. |
Hampson and Metcalfe [27] | 2012 | Incorporating prognostic factors into causal estimators: A comparison of methods for randomised controlled trials with a time-to-event outcome | Statistics in Medicine | o Discusses the problem of making causal inferences in trials with a survival outcome when a proportion of patients allocated to the active intervention do not receive it, and prognosis in the absence of the intervention differs between those who comply and do not comply. o Focuses on the case where treatment switches occur at baseline. o Compares three estimators of the causal effect of treatment on compliers using simulated data. |
Hossain and Karim [22] | 2022 | Analysis approaches to address treatment nonadherence in pragmatic trials with point-treatment settings: a simulation study | BMC Medical Research Methodology | o Compare the performance of four methods to address non-adherence; two adjusted PP approaches and three versions of the IV-based method in the presence of nonadherence. o Identify which methods are more appropriate to use under the scenarios where their respective assumptions are violated. |
Jimenez et al. [33] | 2017 | Evaluating the effects of treatment switching with randomization as an instrumental variable in a randomized controlled trial. | Communications in Statistics – Simulation and Computation | o Utilises simulated data based on an ongoing RCT to evaluate the effects of treatment switching with randomisation as an instrumental variable at differing levels of treatment crossovers, for continuous and binary outcomes. o Data were analysed using IV, ITT and PP methods. |
Korhonen et al. [28] | 1999 | Correcting for non-compliance in randomized trials: An application to the ATBC study | Statistics in Medicine | o We compare the performance of the ITT, AT and g-estimation approaches under different setting for non-compliance with emphasis on the case where there are unmeasured confounders at baseline affecting both treatment-free survival time and time on active treatment. |
Merrill and McClure [34] | 2015 | Dichotomizing partial compliance and increased participant burden in factorial designs: the performance of four noncompliance methods | Trials | o Using simulations, we assessed the performance of ITT, PP, AT and IV in both the partial compliance setting and in a 2-by-2 factorial design with increased participant burden for those randomised to both active treatments. |
Moerbeek and Schie [24] | 2018 | What are the statistical implications of treatment non-compliance in cluster randomized trials: A simulation study | Statistics in Medicine | o Investigate the statistical implications of non-compliance in cluster randomized trials. o A simulation study was conducted with varying degrees of non-compliance at either the cluster level or subject level, which compared ITT, AT, PP and IV methods. |
Odondi and McNamee [29] | 2010 | Performance of statistical methods for analysing survival data in the presence of non-random compliance | Statistics in Medicine | o Compares methods (CALM, C-Prophet, CHARM, Cox-Reg1, Cox-Reg2) and the ITT (Cox-ITT) method when applied to trials where there is time-dependent unidirectional noncompliance in the active arm. o A principal objective is to compare performance of methods which treat compliance as binary (Cox-Reg1 and C-Prophet) and those which utilize the time when subjects switch to noncompliance (Cox-Reg2, CALM and CHARM). |
Roberts [35] | 2021 | The implications of noncompliance for randomized trials with partial nesting due to treatment group | Statistics in Medicine | o Considers the following questions in the setting of nested clustering, whereby clustering only exists in the intervention arm: 1. How do methods for estimating ITT effects using intended group/actual group perform? 2. Where both are recorded, which should be used for estimating ITT effects? 3. How do methods for estimating the CACE perform? |
Schweig et al. [25] | 2020 | Switching Cluster Membership in Cluster Randomized Control Trials: Implications for Design and Analysis | Psychological Methods | o With a focus on cluster switching that violates treatment assignment, goal of article is to explore the challenges posed for analysis of clustered RCTs and propose a potential solution to these challenges. o Address three research questions using real data as well as a series of Monte Carlo simulations: (a) To what extent can inferences about program effects differ when using as-treated or as-assigned clusters? (b) Under what conditions are choices about modelling clustering consequential? Does it depend on the extend of noncompliance or assumptions about the source(s) of between-cluster variability? (c) Are any approaches preferable to others? |
Soltanian et al. [26] | 2020 | Analysis of crossover clinical trial in the presence of non-compliance: a two-stage latent treat grizzle model | JP Journal of Biostatistics | o Compare the accuracy of three models: ordinary grizzle model, generalised grizzle model and LTGM model under different simulated scenarios. o In this article, have tried to use the effect of baseline variables on patients’ compliance and estimate the treatment effects by maximising the likelihood function. |
Stuart and Jo [36] | 2015 | Assessing the sensitivity of methods for estimating principal causal effects | Statistical Methods in Medical Research | o Discuss and examine two methods that rely on very different assumptions to estimate the CACE. o Details the assumptions underlying each approach, and assess each method’s sensitivity to both its assumptions, and those of the other method using both simulated data and a motivating example. |
Wan et al. [30] | 2015 | Bias in estimating the causal hazard ratio when using two-stage instrumental variable methods | Statistics in Medicine | o Directly compare bias in causal HR estimated by 2SRI and 2SPS using extensive simulations. |
Ye et al. [37] | 2014 | Estimating treatment effects in randomised controlled trials with non-compliance: a simulation study | BMJ Open | o Through simulation, we aim to compare common approaches in analysing non-compliant data under different non-compliant scenarios. o Objectives were to compare the performance of these different approaches and make recommendations on optimal approaches under specific scenarios. |