Small sample characteristics of generalized estimating equations

The aim of this study was to investigate the Type I error rate of hypothesis testing based on generalized estimating equations (GEE) for data characteristic of periodontal clinical trials. The data in these studies consist of a large number of binary responses from each subject and a small number of subjects (Haffajee et al. (1983), Goodson (1986), Jenkins et al. (1988)) Computer simulations were employed to investigate GEE based both on an empirical estimate of the variance-covariance matrix and a model-based estimate. Results from this investigation indicate that hypothesis testing based on GEE resulted in inappropriate Type I error rates when small samples are employed. Only an increase in the number of subjects to the point where it matched the number of observations per subject resulted in appropriate Type I error rates     

Modelling heterogeneity in longitudinal binomial responses by generalized estimating equations

ABSTRACT

Extra-binomial variation in longitudinal/clustered binomial data is frequently observed in biomedical and observational studies. The usual generalized estimating equations method treats the extra-binomial parameter as a constant across all subjects. In this paper, a two-parameter variance function modelling the extraneous variance is proposed to account for heterogeneity among subjects. The new approach allows modelling the extra-binomial variation as a function of the mean and binomial size.     

Bias from the use of generalized estimating equations to analyze incomplete longitudinal binary data

Patient dropout is a common problem in studies that collect repeated binary measurements. Generalized estimating equations (GEE) are often used to analyze such data. The dropout mechanism may be plausibly missing at random (MAR), i.e. unrelated to future measurements given covariates and past measurements. In this case, various authors have recommended weighted GEE with weights based on an assumed dropout model, or an imputation approach, or a doubly robust approach based on weighting and imputation. These approaches provide asymptotically unbiased inference, provided the dropout or imputation model (as appropriate) is correctly specified. Other authors have suggested that, provided the working correlation structure is correctly specified, GEE using an improved estimator of the correlation parameters (‘modified GEE’) show minimal bias. These modified GEE have not been thoroughly examined. In this paper, we study the asymptotic bias under MAR dropout of these modified GEE, the standard GEE, and also GEE using the true correlation. We demonstrate that all three methods are biased in general. The modified GEE may be preferred to the standard GEE and are subject to only minimal bias in many MAR scenarios but in others are substantially biased. Hence, we recommend the modified GEE be used with caution.     

Variable selection in generalized estimating equations via empirical likelihood and Gaussian pseudo-likelihood

AIC and BIC based on either empirical likelihood (EAIC and EBIC) or Gaussian pseudo-likelihood (GAIC and GBIC) are proposed to select variables in longitudinal data analysis. Their performances are evaluated in the framework of the generalized estimating equations via intensive simulation studies. Our findings are: (i) GAIC and GBIC outperform other existing methods in selecting variables; (ii) EAIC and EBIC are effective in selecting covariates only when the working correlation structure is correctly specified; (iii) GAIC and GBIC perform well regardless the working correlation structure is correctly specified or not. A real dataset is also provided to illustrate the findings.     

Revealing age-specific past and future unrelated costs of pneumococcal infections by flexible generalized estimating equations

We aimed to study the excess health-care expenditures for persons with a known positive isolate of Streptococcus pneumoniae. The data set was compiled by linking the database of the largest Belgian Sickness Fund with data obtained from laboratories reporting pneumococcal isolates. We analyzed the age-specific per-patient cumulative costs over time, using generalized estimating equations (GEEs). The mean structure was described by fractional polynomials. The quasi-likelihood under the independence model criterion was used to compare different correlation structures. We show for all age groups that the health-care costs incurred by diagnosed pneumococcal patients are significantly larger than those incurred by undiagnosed matched persons. This is not only the case at the time of diagnosis but also long before and after the time of diagnosis. These findings can be informative for the current debate on unrelated costs in health economic evaluation, and GEEs could be used to estimate these costs for other diseases. Finally, these results can be used to inform policy on the expected budget impact of preventing pneumococcal infections.     

Multiple imputation compared with restricted pseudo‐likelihood and generalized estimating equations for analysis of binary repeated measures in clinical studies

Non‐likelihood‐based methods for repeated measures analysis of binary data in clinical trials can result in biased estimates of treatment effects and associated standard errors when the dropout process is not completely at random. We tested the utility of a multiple imputation approach in reducing these biases. Simulations were used to compare performance of multiple imputation with generalized estimating equations and restricted pseudo‐likelihood in five representative clinical trial profiles for estimating (a) overall treatment effects and (b) treatment differences at the last scheduled visit. In clinical trials with moderate to high (40–60%) dropout rates with dropouts missing at random, multiple imputation led to less biased and more precise estimates of treatment differences for binary outcomes based on underlying continuous scores. Copyright © 2005 John Wiley & Sons, Ltd.     

Efficiency of generalized estimating equations for binary responses

Summary.  Using standard correlation bounds, we show that in generalized estimation equations (GEEs) the so-called 'working correlation matrix' R ( α ) for analysing binary data cannot in general be the true correlation matrix of the data. Methods for estimating the correlation param-eter in current GEE software for binary responses disregard these bounds. To show that the GEE applied on binary data has high efficiency, we use a multivariate binary model so that the covariance matrix from estimating equation theory can be compared with the inverse Fisher information matrix. But R ( α ) should be viewed as the weight matrix, and it should not be confused with the correlation matrix of the binary responses. We also do a comparison with more general weighted estimating equations by using a matrix Cauchy–Schwarz inequality. Our analysis leads to simple rules for the choice of α in an exchangeable or autoregressive AR(1) weight matrix R ( α ), based on the strength of dependence between the binary variables. An example is given to illustrate the assessment of dependence and choice of α .     

Variance function in regression analysis of longitudinal data using the generalized estimating equation approach

Longitudinal or clustered response data arise in many applications such as biostatistics, epidemiology and environmental studies. The repeated responses cannot in general be assumed to be independent. One method of analysing such data is by using the generalized estimating equations (GEE) approach. The current GEE method for estimating regression effects in longitudinal data focuses on the modelling of the working correlation matrix assuming a known variance function. However, correct choice of the correlation structure may not necessarily improve estimation efficiency for the regression parameters if the variance function is misspecified [Wang YG, Lin X. Effects of variance-function misspecification in analysis of longitudinal data. Biometrics. 2005;61:413–421]. In this connection two problems arise: finding a correct variance function and estimating the parameters of the chosen variance function. In this paper, we study the problem of estimating the parameters of the variance function assuming that the form of the variance function is known and then the effect of a misspecified variance function on the estimates of the regression parameters. We propose a GEE approach to estimate the parameters of the variance function. This estimation approach borrows the idea of Davidian and Carroll [Variance function estimation. J Amer Statist Assoc. 1987;82:1079–1091] by solving a nonlinear regression problem where residuals are regarded as the responses and the variance function is regarded as the regression function. A limited simulation study shows that the proposed method performs at least as well as the modified pseudo-likelihood approach developed by Wang and Zhao [A modified pseudolikelihood approach for analysis of longitudinal data. Biometrics. 2007;63:681–689]. Both these methods perform better than the GEE approach.     

Regression modelling of weighted κ by using generalized estimating equations

In many clinical studies more than one observer may be rating a characteristic measured on an ordinal scale. For example, a study may involve a group of physicians rating a feature seen on a pathology specimen or a computer tomography scan. In clinical studies of this kind, the weighted κ coefficient is a popular measure of agreement for ordinally scaled ratings. Our research stems from a study in which the severity of inflammatory skin disease was rated. The investigators wished to determine and evaluate the strength of agreement between a variable number of observers taking into account patient-specific (age and gender) as well as rater-specific (whether board certified in dermatology) characteristics. This suggested modelling κ as a function of these covariates. We propose the use of generalized estimating equations to estimate the weighted κ coefficient. This approach also accommodates unbalanced data which arise when some subjects are not judged by the same set of observers. Currently an estimate of overall κ for a simple unbalanced data set without covariates involving more than two observers is unavailable. In the inflammatory skin disease study none of the covariates were significantly associated with κ, thus enabling the calculation of an overall weighted κ for this unbalanced data set. In the second motivating example (multiple sclerosis), geographic location was significantly associated with κ. In addition we also compared the results of our method with current methods of testing for heterogeneity of weighted κ coefficients across strata (geographic location) that are available for balanced data sets.     

Joint generalized estimating equations for multivariate longitudinal binary outcomes with missing data: an application to acquired immune deficiency syndrome data

Summary.  In a large, prospective longitudinal study designed to monitor cardiac abnormalities in children born to women who are infected with the human immunodeficiency virus, instead of a single outcome variable, there are multiple binary outcomes (e.g. abnormal heart rate, abnormal blood pressure and abnormal heart wall thickness) considered as joint measures of heart function over time. In the presence of missing responses at some time points, longitudinal marginal models for these multiple outcomes can be estimated by using generalized estimating equations (GEEs), and consistent estimates can be obtained under the assumption of a missingness completely at random mechanism. When the missing data mechanism is missingness at random, i.e. the probability of missing a particular outcome at a time point depends on observed values of that outcome and the remaining outcomes at other time points, we propose joint estimation of the marginal models by using a single modified GEE based on an EM-type algorithm. The method proposed is motivated by the longitudinal study of cardiac abnormalities in children who were born to women infected with the human immunodeficiency virus, and analyses of these data are presented to illustrate the application of the method. Further, in an asymptotic study of bias, we show that, under a missingness at random mechanism in which missingness depends on all observed outcome variables, our joint estimation via the modified GEE produces almost unbiased estimates, provided that the correlation model has been correctly specified, whereas estimates from standard GEEs can lead to substantial bias.     

Using bootstrap method to evaluate the estimates of nicotine equivalents from linear mixed model and generalized estimating equation

Twenty-four-hour urinary excretion of nicotine equivalents, a biomarker for exposure to cigarette smoke, has been widely used in biomedical studies in recent years. Its accurate estimate is important for examining human exposure to tobacco smoke. The objective of this article is to compare the bootstrap confidence intervals of nicotine equivalents with the standard confidence intervals derived from linear mixed model (LMM) and generalized estimation equation. We use percentile bootstrap method because it has practical value for real-life application and it works well with nicotine data. To preserve the within-subject correlation of nicotine equivalents between repeated measures, we bootstrap the repeated measures of each subject as a vector. The results indicate that the bootstrapped estimates in most cases give better estimates than the LMM and generalized estimation equation without bootstrap.     

A generalized estimating equation approach to modelling foetal response in developmental toxicity studies when the number of implants is dose dependent

Summary. In developmental toxicity studies with exposure before implantation, the toxin that is used may interfere with the early reproductive process and prevent the implantation of some foetuses that would otherwise have been formed. A manifestation of this effect is a decreasing trend in the number of implants per dam as the dose level increases. Because unformed foetuses are not observable, Dunson proposed a multiple-imputation approach to estimate the number of missing foetuses. We propose instead to build a model for observable quantities only, namely the observed litter sizes and the observed numbers of deaths or malformations within litters. Using the probabilistic concept of thinning, we express the probability that a foetus fails to implant in terms of the parameters of the litter size distribution. Estimation is by means of generalized estimating equations and takes into account underdispersion of the observed litter sizes and overdispersion of the numbers of foetal deaths or malformations. A combined risk measure that takes into account not just post-implantation death or malformation but also the possibility of failure to implant is proposed and used to determine the virtually safe dose VSD. It is demonstrated numerically and graphically that ignoring the possibility of unformed foetuses leads to estimates of VSD that are too liberal. The proposed generalized estimating equation approach to estimating VSD and finding lower confidence limits is found to work well in a simulation study. We apply the proposed method to two data sets and compare the results that are obtained with those of existing studies.     

Using generalized doubly robust estimator to estimate average treatment effects of multiple treatments in observational studies

The generalized doubly robust estimator is proposed for estimating the average treatment effect (ATE) of multiple treatments based on the generalized propensity score (GPS). In medical researches where observational studies are conducted, estimations of ATEs are usually biased since the covariate distributions could be unbalanced among treatments. To overcome this problem, Imbens [The role of the propensity score in estimating dose-response functions, Biometrika 87 (2000), pp. 706–710] and Feng et al. [Generalized propensity score for estimating the average treatment effect of multiple treatments, Stat. Med. (2011), in press. Available at: http://onlinelibrary.wiley.com/doi/10.1002/sim.4168/abstract] proposed weighted estimators that are extensions of a ratio estimator based on GPS to estimate ATEs with multiple treatments. However, the ratio estimator always produces a larger empirical sample variance than the doubly robust estimator, which estimates an ATE between two treatments based on the estimated propensity score (PS). We conduct a simulation study to compare the performance of our proposed estimator with Imbens’ and Feng et al.’s estimators, and simulation results show that our proposed estimator outperforms their estimators in terms of bias, empirical sample variance and mean-squared error of the estimated ATEs.     

Review of guidelines and literature for handling missing data in longitudinal clinical trials with a case study

Missing data in clinical trials are inevitable. We highlight the ICH guidelines and CPMP points to consider on missing data. Specifically, we outline how we should consider missing data issues when designing, planning and conducting studies to minimize missing data impact. We also go beyond the coverage of the above two documents, provide a more detailed review of the basic concepts of missing data and frequently used terminologies, and examples of the typical missing data mechanism, and discuss technical details and literature for several frequently used statistical methods and associated software. Finally, we provide a case study where the principles outlined in this paper are applied to one clinical program at protocol design, data analysis plan and other stages of a clinical trial.     

A class of semiparametric transormation frailty models with application to estimating treatment effects and heterogenity in aids clinical traials

Muitivariate failure time data are common in medical research; com¬monly used statistical models for such correlated failure-time data include frailty and marginal models. Both types of models most often assume pro¬portional hazards (Cox, 1972); but the Cox model may not fit the data well This article presents a class of linear transformation frailty models that in¬cludes, as a special case, the proportional hazards model with frailty. We then propose approximate procedures to derive the best linear unbiased es¬timates and predictors of the regression parameters and frailties. We apply the proposed methods to analyze results of a clinical trial of different dose levels of didansine (ddl) among HIV-infected patients who were intolerant of zidovudine (ZDV). These methods yield estimates of treatment effects and of frailties corresponding to patient groups defined by clinical history prior to entry into the trial.     

Evaluating the effects of treatment switching with randomization as an instrumental variable in a randomized controlled trial

The purpose of this study was to utilize simulated data based on an ongoing randomized clinical trial (RCT) to evaluate the effects of treatment switching with randomization as an instrumental variable (IV) at differing levels of treatment crossovers, for continuous and binary outcomes. Data were analyzed using IV, intent-to-treat (ITT), and per protocol (PP) methods. The IV method performed the best, since it provided the most unbiased point estimates, and it had equal or higher power and higher coverage probabilities compared to the ITT estimates, and because a PP analysis can be biased due to its exclusion of non-compliant patients.     

Use of a historical control group in a noninferiority trial assessing a new antibacterial treatment: A case study and discussion of practical implementation aspects

When recruitment into a clinical trial is limited due to rarity of the disease of interest, or when recruitment to the control arm is limited due to ethical reasons (eg, pediatric studies or important unmet medical need), exploiting historical controls to augment the prospectively collected database can be an attractive option. Statistical methods for combining historical data with randomized data, while accounting for the incompatibility between the two, have been recently proposed and remain an active field of research. The current literature is lacking a rigorous comparison between methods but also guidelines about their use in practice. In this paper, we compare the existing methods based on a confirmatory phase III study design exercise done for a new antibacterial therapy with a binary endpoint and a single historical dataset. A procedure to assess the relative performance of the different methods for borrowing information from historical control data is proposed, and practical questions related to the selection and implementation of methods are discussed. Based on our examination, we found that the methods have a comparable performance, but we recommend the robust mixture prior for its ease of implementation.     

A distribution free test to detect general dependence between a response variable and a covariate in the presence of heteroscedastic treatment effects

In this paper, we present a test of independence between the response variable, which can be discrete or continuous, and a continuous covariate after adjusting for heteroscedastic treatment effects. The method involves first augmenting each pair of the data for all treatments with a fixed number of nearest neighbours as pseudo‐replicates. Then a test statistic is constructed by taking the difference of two quadratic forms. The statistic is equivalent to the average lagged correlations between the response and nearest neighbour local estimates of the conditional mean of response given the covariate for each treatment group. This approach effectively eliminates the need to estimate the nonlinear regression function. The asymptotic distribution of the proposed test statistic is obtained under the null and local alternatives. Although using a fixed number of nearest neighbours pose significant difficulty in the inference compared to that allowing the number of nearest neighbours to go to infinity, the parametric standardizing rate for our test statistics is obtained. Numerical studies show that the new test procedure has robust power to detect nonlinear dependency in the presence of outliers that might result from highly skewed distributions. The Canadian Journal of Statistics 38: 408–433; 2010 © 2010 Statistical Society of Canada