Comparison of Several Multiple Imputation Strategies for Repeated Measures Analysis of Clinical Scales: To Truncate or Not To?

Gestational Weight Gain and Long-term Maternal Obesity Risk: A Multiple-Bias Analysis

BACKGROUND

Lifecourse research provides an important framework for chronic disease epidemiology. However, data collection to observe health characteristics over long periods is vulnerable to systematic error and statistical bias. We present a multiple-bias analysis using real-world data to estimate associations between excessive gestational weight gain and mid-life obesity, accounting for confounding, selection, and misclassification biases.

METHODS

Participants were from the multiethnic Study of Women's Health Across the Nation. Obesity was defined by waist circumference measured in 1996-1997 when women were age 42-53. Gestational weight gain was measured retrospectively by self-recall and was missing for over 40% of participants. We estimated relative risk (RR) and 95% confidence intervals (CI) of obesity at mid-life for presence versus absence of excessive gestational weight gain in any pregnancy. We imputed missing data via multiple imputation and used weighted regression to account for misclassification.

RESULTS

Among the 2,339 women in this analysis, 937 (40%) experienced obesity in mid-life. In complete case analysis, women with excessive gestational weight gain had an estimated 39% greater risk of obesity (RR = 1.4, CI = 1.1, 1.7), covariate-adjusted. Imputing data, then weighting estimates at the guidepost values of sensitivity = 80% and specificity = 75%, increased the RR (95% CI) for obesity to 2.3 (2.0, 2.6). Only models assuming a 20-point difference in specificity between those with and without obesity decreased the RR.

CONCLUSIONS

The inference of a positive association between excessive gestational weight gain and mid-life obesity is robust to methods accounting for selection and misclassification bias.

Points to consider for analyzing efficacy outcomes in long-term extension clinical trials

This article focuses on 2 objectives in the analysis of efficacy in long-term extension studies of chronic diseases: (1) defining and discussing estimands of interest in such studies and (2) evaluating the performance of several multiple imputation methods that may be useful in estimating some of these estimands. Specifically, 4 estimands are defined and their clinical utility and inferential ramifications discussed. The performance of several multiple imputation methods and approaches were evaluated using simulated data. Results suggested that when interest is in a binary outcome derived from an underlying continuous measurement, it is preferable to impute the underlying continuous value that is subsequently dichotomized rather than to directly impute the binary outcome. Results also demonstrated that multivariate Gaussian models with Markov chain Monte Carlo imputation and sequential regression have minimal bias and the anticipated confidence interval coverage, even in settings with ordinal data where departures from normality are a concern. These approaches are further illustrated using a long-term extension study in psoriasis.

Choice of estimand and analysis methods in diabetes trials with rescue medication

The analysis of clinical trials aiming to show symptomatic benefits is often complicated by the ethical requirement for rescue medication when the disease state of patients worsens. In type 2 diabetes trials, patients receive glucose-lowering rescue medications continuously for the remaining trial duration, if one of several markers of glycemic control exceeds pre-specified thresholds. This may mask differences in glycemic values between treatment groups, because it will occur more frequently in less effective treatment groups. Traditionally, the last pre-rescue medication value was carried forward and analyzed as the end-of-trial value. The deficits of such simplistic single imputation approaches are increasingly recognized by regulatory authorities and trialists. We discuss alternative approaches and evaluate them through a simulation study. When the estimand of interest is the effect attributable to the treatments initially assigned at randomization, then our recommendation for estimation and hypothesis testing is to treat data after meeting rescue criteria as deterministically 'missing' at random, because initiation of rescue medication is determined by observed in-trial values. An appropriate imputation of values after meeting rescue criteria is then possible either directly through multiple imputation or implicitly with a repeated measures model. Crucially, one needs to jointly impute or model all markers of glycemic control that can lead to the initiation of rescue medication. An alternative for hypothesis testing only are rank tests with outcomes from patients 'requiring rescue medication' ranked worst, and non-rescued patients ranked according to final visit values. However, an appropriate ranking of not observed values may be controversial.