Investigating treatment-effect modification by a continuous covariate in IPD meta-analysis: an approach using fractional polynomials.

BACKGROUND: In clinical trials, there is considerable interest in investigating whether a treatment effect is similar in all patients, or that one or more prognostic variables indicate a differential response to treatment. To examine this, a continuous predictor is usually categorised into groups according to one or more cutpoints. Several weaknesses of categorization are well known. To avoid the disadvantages of cutpoints and to retain full information, it is preferable to keep continuous variables continuous in the analysis. To handle this issue, the Subpopulation Treatment Effect Pattern Plot (STEPP) was proposed about two decades ago, followed by the multivariable fractional polynomial interaction (MFPI) approach. Provided individual patient data (IPD) from several studies are available, it is possible to investigate for treatment heterogeneity with meta-analysis techniques. Meta-STEPP was recently proposed and in patients with primary breast cancer an interaction of estrogen receptors with chemotherapy was investigated in eight randomized controlled trials (RCTs). METHODS: We use data from eight randomized controlled trials in breast cancer to illustrate issues from two main tasks. The first task is to derive a treatment effect function (TEF), that is, a measure of the treatment effect on the continuous scale of the covariate in the individual studies. The second is to conduct a meta-analysis of the continuous TEFs from the eight studies by applying pointwise averaging to obtain a mean function. We denote the method metaTEF. To improve reporting of available data and all steps of the analysis we introduce a three-part profile called MethProf-MA. RESULTS: Although there are considerable differences between the studies (populations with large differences in prognosis, sample size, effective sample size, length of follow up, proportion of patients with very low estrogen receptor values) our results provide clear evidence of an interaction, irrespective of the choice of the FP function and random or fixed effect models. CONCLUSIONS: In contrast to cutpoint-based analyses, metaTEF retains the full information from continuous covariates and avoids several critical issues when performing IPD meta-analyses of continuous effect modifiers in randomised trials. Early experience suggests it is a promising approach. TRIAL REGISTRATION: Not applicable.

Kernel hazard estimation for visualisation of the effect of a continuous covariates on time-to-event endpoints.

The problem of associating a continuous covariate, or biomarker, against a time-to-event outcome, is that it often requires categorisation of the covariate. This can lead to bias, loss of information and a poor representation of any underlying relationship. Here, two methods are proposed for estimating the effects of a continuous covariate on a time-to-event endpoint using weighted kernel estimators. The first method aims to estimate a density function for a time-to-event endpoint conditional on some covariate value whilst the second uses a joint density estimator. The results are visualisations in the form of surface plots that show the effects of a covariate without any need for categorisation. Both methods can aid interpretation and analysis of covariates against a time-to-event endpoint.

Effects of data preprocessing on results of the epidemiological analysis of coronary heart disease and behaviour-related risk factors.

BACKGROUND: We carried out this study to demonstrate the effects of outcome sensitivity, participant exclusions, and covariate manipulations on results of the epidemiological analysis of coronary heart disease (CHD) and its behaviour-related risk factors. MATERIAL AND METHODS: Our study population consisted of 1592 54-year-old men, who participated in the Kuopio Ischaemic Heart Disease Risk Factor (KIHD) Study. We used the Cox proportional-hazards model to predict the hazard of CHD and applied different sets of outcomes concerning outcome sensitivity and data preprocessing procedures regarding participant exclusions and covariate manipulations. RESULTS: The mean follow-up time was 23 years, and 730 men received the CHD diagnosis. Cox regressions based on data with no participant exclusions most often discovered statistically significant associations. Loose inclusion criteria for study participants with any CVD during the follow-up and strict exclusion criteria for participants with no CVD were best in discovering the associations between risk factors and CHD. Outcome sensitivity affected the associations, whereas the covariate type, continuous or categorical, did not. CONCLUSIONS: This study suggests that excluding study participants who are not disease-free at baseline is probably unnecessary for epidemiological analyses. Epidemiological research reports should present results based on no data exclusions together with results based on reasoned exclusions.

Choosing an imbalance metric for covariate-constrained randomization in multiple-arm cluster-randomized trials.

BACKGROUND: In cluster-randomized controlled trials (C-RCTs), covariate-constrained randomization (CCR) methods efficiently control imbalance in multiple baseline cluster-level variables, but the choice of imbalance metric to define the subset of "adequately balanced" possible allocation schemes for C-RCTs involving more than two arms and continuous variables is unclear. In an ongoing three-armed C-RCT, we chose the min(three Kruskal-Wallis [KW] test P values) > 0.30 as our metric. We use simulation studies to explore the performance of this and other metrics of baseline variable imbalance in CCR. METHODS: We simulated three continuous variables across three arms under varying allocation ratios and assumptions. We compared the performance of min(analysis of variance [ANOVA] P value) > 0.30, min(KW P value) > 0.30, multivariate analysis of variance (MANOVA) P value > 0.30, min(nine possible t test P values) > 0.30, and min(Wilcoxon rank-sum [WRS] P values) > 0.30. RESULTS: Pairwise comparison metrics (t test and WRS) tended to be the most conservative, providing the smallest subset of allocation schemes (10%-13%) meeting criteria for acceptable balance. Sensitivity of the min(t test P values) > 0.30 for detecting non-trivial imbalance was 100% for both hypothetical and resampled simulation scenarios. The KW criterion maintained higher sensitivity than both the MANOVA and ANOVA criteria (89% to over 99%) but was not as sensitive as pairwise criteria. CONCLUSIONS: Our criterion, the KW P value > 0.30, to signify "acceptable" balance was not the most conservative, but it appropriately identified imbalance in the majority of simulations. Since all are related, CCR algorithms involving any of these imbalance metrics for continuous baseline variables will ensure robust simultaneous control over multiple continuous baseline variables, but we recommend care in determining the threshold of "acceptable" levels of (im)balance. TRIAL REGISTRATION: This trial is registered on ClinicalTrials.gov (initial post: December 1, 2016; identifier: NCT02979444 ).

Model selection for univariable fractional polynomials.

Since Royston and Altman's 1994 publication (Journal of the Royal Statistical Society, Series C 43: 429-467), fractional polynomials have steadily gained popularity as a tool for flexible parametric modeling of regression relationships. In this article, I present fp_select, a postestimation tool for fp that allows the user to select a parsimonious fractional polynomial model according to a closed test procedure called the fractional polynomial selection procedure or function selection procedure. I also give a brief introduction to fractional polynomial models and provide examples of using fp and fp_select to select such models with real data.

Interaction of treatment with a continuous variable: simulation study of power for several methods of analysis.

In a large simulation study reported in a companion paper, we investigated the significance levels of 21 methods for investigating interactions between binary treatment and a continuous covariate in a randomised controlled trial. Several of the methods were shown to have inflated type 1 errors. In the present paper, we report the second part of the simulation study in which we investigated the power of the interaction procedures for two sample sizes and with two distributions of the covariate (well and badly behaved). We studied several methods involving categorisation and others in which the covariate was kept continuous, including fractional polynomials and splines. We believe that the results provide sufficient evidence to recommend the multivariable fractional polynomial interaction procedure as a suitable approach to investigate interactions of treatment with a continuous variable. If subject-matter knowledge gives good arguments for a non-monotone treatment effect function, we propose to use a second-degree fractional polynomial approach, but otherwise a first-degree fractional polynomial (FP1) function with added flexibility (FLEX3) is the method of choice. The FP1 class includes the linear function, and the selected functions are simple, understandable, and transferable. Furthermore, software is available. We caution that investigation of interactions in one dataset can only be interpreted in a hypothesis-generating sense and needs validation in new data.

A smooth covariate rank transformation for use in regression models with a sigmoid dose-response function.

We consider how to represent sigmoid-type regression relationships in a practical and parsimonious way. A pure sigmoid relationship has an asymptote at both ends of the range of a continuous covariate. Curves with a single asymptote are also important in practice. Many smoothers, such as fractional polynomials and restricted cubic regression splines, cannot accurately represent doubly asymptotic curves. Such smoothers may struggle even with singly asymptotic curves. Our approach to modeling sigmoid relationships involves applying a preliminary scaled rank transformation to compress the tails of the observed distribution of a continuous covariate. We include a step that provides a smooth approximation to the empirical cumulative distribution function of the covariate via the scaled ranks. The procedure defines the approximate cumulative distribution transformation of the covariate. To fit the substantive model, we apply fractional polynomial regression to the outcome with the smoothed, scaled ranks as the covariate. When the resulting fractional polynomial function is monotone, we have a sigmoid function. We demonstrate several practical applications of the approximate cumulative distribution transformation while also illustrating its ability to model some unusual functional forms. We describe a command, acd, that implements it.