Adjusting for drop-out in clinical trials with repeated measures: design and analysis issues

Shared random parameter models: A legacy of the biostatistics program at the National Heart, Lung, and Blood Institute

Shared random parameter models (SRPMs) were first introduced by researchers at the National Heart Lung and Blood Institute (NHLBI) Biostatistics Branch for analyzing longitudinal data with informative dropout (Wu and Carroll, 1987; Wu and Bailey, 1988; Follmann and Wu, 1995; Albert and Follmann, 2000; Albert et al, 2002). This work was all focused on characterizing the longitudinal data process in the presence of an informative missing data mechanism that is treated as a nuisance. Shared random parameter modeling approaches have also been developed from the perspective of characterizing the relationship between longitudinal data and a subsequent outcome that may be an event time, a dichotomous measurement, or another longitudinal outcome. This article will review the early contributions of the NHLBI biostatisticians on SRPMs for analyzing longitudinal data with dropout and demonstrate how these ideas have, more recently, been applied in these other areas of biostatistics. Rather than focus on technical details or specific analyses, this article presents a conceptual framework for SRPMs within a historical context.

Using a Counting Process Method to Impute Censored Follow-Up Time Data

Censoring occurs when complete follow-up time information is unavailable for patients enrolled in a clinical study. The process is considered to be informative (non-ignorable) if the likelihood function for the model cannot be partitioned into a set of response parameters that are independent of the censoring parameters. In such cases, estimated survival time probabilities may be biased, prompting the need for special statistical methods to remedy the situation. The problem is especially salient when censoring occurs early in a study. In this paper, we describe a method to impute censored follow-up times using a counting process method.

Handling dropout and clustering in longitudinal multicentre clinical trials

Many clinical trials enrol patients from different medical centres. Multi-centre studies are particularly helpful in cancer research as they allow researchers to evaluate the efficacy of a therapy in a variety of patients and settings, making it possible to investigate the effect of treatments in those cases when it is difficult, or even impossible, for a single centre to recruit the required number of patients. It is often argued, however, that despite agreement among different centres to follow common standardized protocols, variation may occur in both baseline characteristics of the recruited patients and in treatment effects. This heterogeneity should be detected and, if present, accounted for in the data analysis. Furthermore, the longitudinal nature of these types of experimental studies raises the problem of attrition, that is, patients may dropout of the study for a number of reasons mainly death or disease progression. In this paper, we consider the health related quality of life of advanced melanoma patients in a longitudinal multi-centre randomized clinical trial comparing two different anti-tumoural treatments. We propose a Heckman type model to account for the possibility that patients dropout according to a non-ignorable mechanism. The model is extended to a multilevel setting to account both for the longitudinal nature and the multi-centre structure of the design. We found a strong variation across centres in the quality of life evaluation. The effect of centres on the dropout was not found to be relevant in the considered data although dropout does depend on patient′s characteristics.

Suggested statistical reporting guidelines for clinical trials data

Statistical methods are widely accepted tools for analyzing data and drawing appropriate conclusions. However, inappropriate statistical methods may lead to unsuitable or false conclusions. Statistical quality of articles published in Indian medical journals presents with statistical fallacies and very few attempts have been done to document it till date. Issues related to inappropriate use of statistic in interpreting clinical trial raises serious doubt about the validity of the results published and there generalization to patient population. Clinical trials are considered to have one of the highest evidences in terms of efficacy of any intervention and hence results reported in these clinical trials help in appropriate decision making in clinical practices. Present manuscript is an attempt to have an insight into issues related to inappropriate use of statistic during clinical trials and suggests possible methods by which such fallacies can be averted.

Tracking inner city substance users from the emergency department: how many contacts does it take?

BACKGROUND

Longitudinal studies of substance users report difficulty in locating and completing 12-month interviews, which may compromise study validity.

OBJECTIVES

This study examined rates and predictors of contact difficulty and in-person follow-up completion among patients presenting with cocaine-related chest pain to an inner-city emergency department (ED). The authors hypothesize that less staff effort in contacting patients and lower follow-up rates would bias subsequent substance use analysis by missing those with heavier substance misuse.

METHODS

A total of 219 patients aged 19 to 60 years (65% males; 78% African American) with cocaine-related chest pain were interviewed in the ED and then in person at 3, 6, and 12 months. Demographics, substance use measures, and amount/type of research staff contacts (telephone, letters, home visits, and locating patient during return ED visits) were recorded. Poisson and negative binomial regression analyses were conducted to predict quantity of patient contacts for the 12-month follow-up.

RESULTS

Interview completion rates at 3, 6, and 12 months were 78, 82, and 80%, respectively. Average contact attempts to obtain each interview were 10 at 3 months (range 3-44), 8 at 6 months (1-31), and 8 at 12 months (1-49); 13% of patients required a home visit to complete the 12-month interview. Participants requiring more contact attempts by staff were younger and reported more frequent binge drinking at baseline (p < 0.05), but were less likely to meet criteria for substance abuse or dependence (p < 0.5), or to report prior mental health treatment (p < 0.05). Comparisons of parallel regressions predicting contact difficulty based on the entire sample, the low-effort group, and the difficult-to-reach group showed variation in findings.

CONCLUSIONS

This study demonstrates that substantial staff effort is required to achieve adequate retention over 12 months of patients with substance misuse. Without these extensive efforts at follow-up, longitudinal analyses may be biased.

Prevention of missing data in clinical research studies

Missing data is a problem that is ubiquitous to all clinical studies and a source of multiple problems from an analytic point of view (reduced statistical power, increased the type I error, bias) Statistical approaches have been developed to analyze data collected from trials with missing data. Understanding and implementing the appropriate statistical technique is essential but should be differentiated from preventive approaches that are designed to reduce rates of missing data In this article, we draw attention to these preventive efforts. Seven steps to minimizing the amount of missing data are defined as documentation, training, monitoring reports, patient contact, data entry and management, pilot studies, and communication. Although the implementation of these approaches is time consuming and costly, the overall quality of the study is increased. Despite efforts devoted to areas, no study is without missing data. Once the study is completed, it is essential to assess the pattern of missing data and apply the appropriate statistical analysis.

Guidelines for the design of clinical trials with longitudinal outcomes

A common objective of longitudinal clinical trials is to compare rates of change in a continuous response variable between two groups. The power realized for such a study is a function of both the number of people recruited and the planned number of measurements for each participant. By varying these two quantities in opposite directions, power can be kept at the desired level. We consider the problem of how best to choose the sample size and frequency of measurement, with a view to minimizing either the total number of measurements or the cost of a study. Some general guidelines are first developed for the situation in which all participants have complete observations. In practice, however, longitudinal studies often suffer from dropout, where a participant leaves the study permanently so that no further observations are possible. We therefore consider the impact of unanticipated dropout on power and also ways of allowing for dropout at the design stage. Based on our results, we propose some general design guidelines for longitudinal trials comparing rates of change when dropout is present.

Problems in dealing with missing data and informative censoring in clinical trials

Acommon problem in clinical trials is the missing data that occurs when patients do not complete the study and drop out without further measurements. Missing data cause the usual statistical analysis of complete or all available data to be subject to bias. There are no universally applicable methods for handling missing data. We recommend the following: (1) Report reasons for dropouts and proportions for each treatment group; (2) Conduct sensitivity analyses to encompass different scenarios of assumptions and discuss consistency or discrepancy among them; (3) Pay attention to minimize the chance of dropouts at the design stage and during trial monitoring; (4) Collect post-dropout data on the primary endpoints, if at all possible; and (5) Consider the dropout event itself an important endpoint in studies with many.