Power analysis on the time effect for the longitudinal Rasch model.

Statistics literature in the social, behavioral, and biomedical sciences typically stress the importance of power analysis. Patient Reported Outcomes (PRO) such as quality of life and other perceived health measures (pain, fatigue, stress,...) are increasingly used as important health outcomes in clinical trials or in epidemiological studies. They cannot be directly observed nor measured as other clinical or biological data and they are often collected through questionnaires with binary or polytomous items. The Rasch model is the well known model in the item response theory (IRT) for binary data. The article proposes an approach to evaluate the statistical power of the time effect for the longitudinal Rasch model with two time points. The performance of this method is compared to the one obtained by simulation study. Finally, the proposed approach is illustrated on one subscale of the SF-36 questionnaire.

Thyroglobulin in lymph node fine-needle aspiration washout: a systematic review and meta-analysis of diagnostic accuracy.

CONTEXT: The thyroglobulin measurement in the needle washout after fine-needle aspiration (FNA) has been reported to increase the sensitivity of FNA in identifying lymph node (LN) metastases from differentiated thyroid cancer (DTC). OBJECTIVE: The aim of the study was to estimate the diagnostic accuracy of this technique. DATA SOURCES: To identify eligible studies, we searched electronic databases for original articles in English from 1975 through 2013. STUDY SELECTION: Studies that enrolled participants with suspicious neck LNs during thyroid nodule workup or thyroid cancer follow-up were included. DATA EXTRACTION: Working independently, authors used a standard form to extract data. For quality assessment, QUADAS2 guidelines were applied. DATA SYNTHESIS: Including all the selected studies (24 studies, 2865 LNs) in the pooled analysis, overall sensitivity was 95.0% (95% confidence interval [CI], 93.7-96.0%), specificity was 94.5% (95% CI, 93.2-95.7%), and diagnostic odds ratio (DOR) was 338.91 (95% CI, 164.82-696.88) with significant heterogeneity (inconsistency [I(2)] = 65.7%; heterogeneity, P < .001). Stratifying different populations and including only patients with thyroid gland (410 LNs), pooled sensitivity was 86.2% (95% CI, 80.9-90.5%), specificity was 90.2% (85.1-94.0%), and DOR was 56.621 (22.535-142.26; I(2) = 37.3%; heterogeneity, P = .121). Including only patients after thyroidectomy (1007 LNs), pooled sensitivity was 96.9% (95% CI, 94.9-98.2%), specificity was 94.1% (91.7-96.0%), and DOR was 407.65 (198.67-836.46; I(2) = 0.0%; heterogeneity, P = .673). CONCLUSIONS: Thyroglobulin measurement in washout from LN FNA has high accuracy in early detection of nodal metastases from DTC. The technique is simple, but a better standardization of criteria for patient selection, analytical methods, and cutoff levels is required.

Missing data in longitudinal studies: cross-sectional multiple imputation provides similar estimates to full-information maximum likelihood.

PURPOSE: The aim of this research was to examine, in an exploratory manner, whether cross-sectional multiple imputation generates valid parameter estimates for a latent growth curve model in a longitudinal data set with nonmonotone missingness. METHODS: A simulated longitudinal data set of N = 5000 was generated and consisted of a continuous dependent variable, assessed at three measurement occasions and a categorical time-invariant independent variable. Missing data had a nonmonotone pattern and the proportion of missingness increased from the initial to the final measurement occasion (5%-20%). Three methods were considered to deal with missing data: listwise deletion, full-information maximum likelihood, and multiple imputation. A latent growth curve model was specified and analysis of variance was used to compare parameter estimates between the full data set and missing data approaches. RESULTS: Multiple imputation resulted in significantly lower slope variance compared with the full data set. There were no differences in any parameter estimates between the multiple imputation and full-information maximum likelihood approaches. CONCLUSIONS: This study suggested that in longitudinal studies with nonmonotone missingness, cross-sectional imputation at each time point may be viable and produces estimates comparable with those obtained with full-information maximum likelihood. Future research pursuing the validity of this method is warranted.

Selectivity of attrition in longitudinal studies of cognitive functioning.

OBJECTIVES: Identify characteristics distinguishing people who do and do not continue to participate in a longitudinal study and determine whether the longitudinal changes for people who continue are representative of the changes that would have occurred had longitudinal data been available from all of the initial participants. METHOD: Moderately large samples of returning (N = 2,082) and nonreturning (N = 1,698) participants across a wide age range (i.e., 18-97 years of age) performed a battery of cognitive tests and completed personality and mood questionnaires. Differences between the groups were examined with multiple regression analyses with age, returner status, and their interaction as predictors. RESULTS: Compared with participants who did not return, returning participants at the initial occasion had higher levels of each cognitive ability and of certain personality characteristics (e.g., agreeableness and openness), but many of the differences were only apparent among adults older than 50 years of age. Importantly, there was no evidence that the longitudinal change for nonreturning participants would have been different from that among the participants who did return. DISCUSSION: The phenomenon of selective attrition is more complex than often assumed, and it may not necessarily limit the generalizability of longitudinal comparisons.

Risk of attrition in a longitudinal study of skin cancer: logistic and survival models can give different results.

OBJECTIVES: Longitudinal studies are a major tool for public health research, but their value can be undermined by attrition. Identification of factors associated with attrition through modeling depends on the efficient use of data and is conditional on modeling assumptions being met. The primary aim of this study was to compare the performance of four models in analyzing attrition risk. STUDY DESIGN AND SETTING: Data from participants who were lost to follow-up from The Nambour Skin Cancer Study between 1992 and 2000 were analyzed using logistic and survival models, for all-cause and nondeath attritions. RESULTS: During follow-up, 321 (19.8%) of 1,621 participants were lost to follow-up; 70 (4.3%) because of death and 251 (15.5%) for other reasons. Using survival models showed skin cancer diagnosis to be associated with increased all-cause attrition (hazard ratio: 2.3; 95% confidence interval [95% CI]: 1.5, 3.4) and nondeath attrition (subhazard ratio: 1.9; 95% CI: 1.0, 3.3). Using logistic regression resulted in inverse associations being observed for both all-cause attrition (odds ratio [OR]: 0.7; 95% CI: 0.5, 1.1) and nondeath attrition (OR: 0.5; 95% CI: 0.3, 1.0). CONCLUSION: These results demonstrate the relative inadequacy of a logistic as opposed to a survival approach when analyzing attrition risk in the presence of time-varying covariates and multiple timepoints.

Evaluation of time-varying and cumulative effects in nursing in a longitudinal study.

BACKGROUND: Traditional statistics in longitudinal data analysis are likely to be insufficient in nursing studies, in which the time varying characteristics of explanatory variables and cumulative effects require additional consideration. OBJECTIVES: The aims of this study were to introduce alternative longitudinal approaches for incorporating time-varying variables and cumulative effects, to discuss their strengths, and to highlight key issues that nursing researchers should recognize before and while undertaking such analyses. RESULTS: The three alternative models provide differing analytical outcomes based on the research focus. The baseline tracking model was used to estimate the stability effect of an intervention program, detecting risk factors early. The temporal sequence of potential cause and effect was incorporated further in the time-dependent model. The cumulative model was used to explore whether cumulative intervention effects existed. CONCLUSION: Nurse researchers should incorporate alternative methods into the longitudinal data analysis tools they commonly use when facing explanatory variables with time variations or cumulative effects on the variable being measured.

Assessing possible selection bias in a national voluntary MS longitudinal study in Australia.

BACKGROUND: Surveying volunteer members of a multiple sclerosis registry is a very cost-effective way of assessing the impact of the disease on life outcomes. However, whether the data from such a study can be generalised to the whole population of persons living with MS in a country or region is unclear. METHODS: Here we compare the demographic and disease characteristics of participants in one such study, the Australian Multiple Sclerosis Longitudinal Study (AMSLS), with two well-characterised MS prevalence studies with near-complete ascertainment of MS in their study regions. RESULTS: Although some differences were found, these largely represented the effects of geography (sex ratios) and local factors (national immunomodulatory therapy prescribing requirements), and the cohorts were otherwise comparable. Overall, despite comprising only 12-16% of MS cases in Australia, the AMSLS is highly representative of the MS population. CONCLUSIONS: Therefore with some minor caveats, the AMSLS data can be generalised to the whole Australasian MS population. Volunteer disease registries such as this can be highly representative and provide an excellent convenience sample when studying rare conditions such as MS.

Flexible marginalized models for bivariate longitudinal ordinal data.

Random effects models are commonly used to analyze longitudinal categorical data. Marginalized random effects models are a class of models that permit direct estimation of marginal mean parameters and characterize serial correlation for longitudinal categorical data via random effects (Heagerty, 1999). Marginally specified logistic-normal models for longitudinal binary data. Biometrics 55, 688-698; Lee and Daniels, 2008. Marginalized models for longitudinal ordinal data with application to quality of life studies. Statistics in Medicine 27, 4359-4380). In this paper, we propose a Kronecker product (KP) covariance structure to capture the correlation between processes at a given time and the correlation within a process over time (serial correlation) for bivariate longitudinal ordinal data. For the latter, we consider a more general class of models than standard (first-order) autoregressive correlation models, by re-parameterizing the correlation matrix using partial autocorrelations (Daniels and Pourahmadi, 2009). Modeling covariance matrices via partial autocorrelations. Journal of Multivariate Analysis 100, 2352-2363). We assess the reasonableness of the KP structure with a score test. A maximum marginal likelihood estimation method is proposed utilizing a quasi-Newton algorithm with quasi-Monte Carlo integration of the random effects. We examine the effects of demographic factors on metabolic syndrome and C-reactive protein using the proposed models.

The evolution of a clinical registry during 25 years of experience with Gamma Knife radiosurgery in Pittsburgh.

OBJECT: The first North American 201 cobalt-60 source Gamma Knife surgery (GKS) device was introduced at the University of Pittsburgh Medical Center in 1987. The introduction of this innovative and largely untested surgical procedure prompted the desire to study patient outcomes and evaluate the effectiveness of this technique. The parallel advances in computer software and database technology led to the development of a registry to track patient outcomes at this center. The purpose of this study was to describe the registry's evolution and to evaluate its usefulness. METHODS: A team was created to develop a software database and tracking system to organize and retain information on the usage of GKS. All patients undergoing GKS were systematically entered into this database by a clinician familiar with the technology and the clinical indications. Information included patient demographics and diagnosis as well as the anatomical site of the target and details of the procedure. RESULTS: There are currently 11,738 patients in the database, which began to be used in August 1987. The University of Pittsburgh Medical Center has pioneered the evaluation and publication of the GKS technique and outcomes. Data derived from this computer database have facilitated the publication of more than 400 peer-reviewed manuscripts, more than 200 book chapters, 8 books, and more than 300 published abstracts and scientific presentations. The use of GKS has become a well-established surgical technique that has been performed more than 700,000 times around the world. CONCLUSIONS: The development of a patient registry to track and analyze the use of GKS has given investigators the ability to study patient procedures and outcomes. The future of clinical medical research will rely on the ability of clinical centers to store and to share information.

Dynamic semiparametric Bayesian models for genetic mapping of complex trait with irregular longitudinal data.

Many phenomena of fundamental importance to biology and biomedicine arise as a dynamic curve, such as organ growth and HIV dynamics. The genetic mapping of these traits is challenged by longitudinal variables measured at irregular and possibly subject-specific time points, in which case nonnegative definiteness of the estimated covariance matrix needs to be guaranteed. We present a semiparametric approach for genetic mapping within the mixture-model setting by jointly modeling mean and covariance structures for irregular longitudinal data. Penalized spline is used to model the mean functions of individual quantitative trait locus (QTL) genotypes as latent variables, whereas an extended generalized linear model is used to approximate the covariance matrix. The parameters for modeling the mean-covariances are estimated by MCMC, using the Gibbs sampler and the Metropolis-Hastings algorithm. We derive the full conditional distributions for the mean and covariance parameters and compute Bayes factors to test the hypothesis about the existence of significant QTLs. We used the model to screen the existence of specific QTLs for age-specific change of body mass index with a sparse longitudinal data set. The new model provides powerful means for broadening the application of genetic mapping to reveal the genetic control of dynamic traits.

A multilevel simultaneous equations model for within-cluster dynamic effects, with an application to reciprocal parent-child and sibling effects.

There has been substantial interest in the social and health sciences in the reciprocal causal influences that people in close relationships have on one another. Most research has considered reciprocal processes involving only 2 units, although many social relationships of interest occur within a larger group (e.g., families, work groups, peer groups, classrooms). This article presents a general longitudinal multilevel modeling framework for the simultaneous estimation of reciprocal relationships among individuals with unique roles operating in a social group. We use family data for illustrative purposes, but the model is generalizable to any social group in which measurements of individuals in the social group occur over time, individuals have unique roles, and clustering of the data is evident. We allow for the possibility that the outcomes of family members are influenced by a common set of unmeasured family characteristics. The multilevel model we propose allows for residual variation in the outcomes of parents and children at the occasion, individual, and family levels and residual correlation between parents and children due to the unmeasured shared environment, genetic factors, and shared measurement. Another advantage of this method over approaches used in previous family research is it can handle mixed family sizes. The method is illustrated in an analysis of maternal depression and child delinquency using data from the Avon Brothers and Sisters Study.

Cognitive impairment in the preclinical stage of dementia in FTD-3 CHMP2B mutation carriers: a longitudinal prospective study.

OBJECTIVE AND METHODS: A longitudinal study spanning over 8 years and including 17 asymptomatic individuals with CHMP2B mutations was conducted to assess the earliest neuropsychological changes in autosomal dominant neurodegenerative disease frontotemporal dementia (FTD) linked to chromosome 3 (FTD-3). Subjects were assessed with neuropsychological tests in 2002, 2005 and 2010. RESULTS: Cross-sectional analyses showed that the mutation carriers scored lower on tests of psychomotor speed, working memory, executive functions and verbal memory than a control group consisting of not-at-risk family members and spouses. Longitudinal analyses showed a gradual decline in psychomotor speed, working memory capacity and global executive measures in the group of non-demented mutation carriers that was not found in the control group. In contrast, there were no significant group differences in domain scores on memory or visuospatial functions. On an individual level the cognitive changes over time varied considerably. CONCLUSION: Subjects with CHMP2B mutation show cognitive changes dominated by executive dysfunctions, years before they fulfil diagnostic criteria of FTD. However, there is great heterogeneity in the individual cognitive trajectories.

Model uncertainty and multimodel inference in reliability estimation within a longitudinal framework.

Laenen, Alonso, and Molenberghs (2007) and Laenen, Alonso, Molenberghs, and Vangeneugden (2009) proposed a method to assess the reliability of rating scales in a longitudinal context. The methodology is based on hierarchical linear models, and reliability coefficients are derived from the corresponding covariance matrices. However, finding a good parsimonious model to describe complex longitudinal data is a challenging task. Frequently, several models fit the data equally well, raising the problem of model selection uncertainty. When model uncertainty is high one may resort to model averaging, where inferences are based not on one but on an entire set of models. We explored the use of different model building strategies, including model averaging, in reliability estimation. We found that the approach introduced by Laenen et al. (2007, 2009) combined with some of these strategies may yield meaningful results in the presence of high model selection uncertainty and when all models are misspecified, in so far as some of them manage to capture the most salient features of the data. Nonetheless, when all models omit prominent regularities in the data, misleading results may be obtained. The main ideas are further illustrated on a case study in which the reliability of the Hamilton Anxiety Rating Scale is estimated. Importantly, the ambit of model selection uncertainty and model averaging transcends the specific setting studied in the paper and may be of interest in other areas of psychometrics.

Measuring change for a multidimensional test using a generalized explanatory longitudinal item response model.

Even though many educational and psychological tests are known to be multidimensional, little research has been done to address how to measure individual differences in change within an item response theory framework. In this paper, we suggest a generalized explanatory longitudinal item response model to measure individual differences in change. New longitudinal models for multidimensional tests and existing models for unidimensional tests are presented within this framework and implemented with software developed for generalized linear models. In addition to the measurement of change, the longitudinal models we present can also be used to explain individual differences in change scores for person groups (e.g., learning disabled students versus non-learning disabled students) and to model differences in item difficulties across item groups (e.g., number operation, measurement, and representation item groups in a mathematics test). An empirical example illustrates the use of the various models for measuring individual differences in change when there are person groups and multiple skill domains which lead to multidimensionality at a time point.

Assessing discrimination of risk prediction rules in a clustered data setting.

The AUC (area under ROC curve) is a commonly used metric to assess discrimination of risk prediction rules; however, standard errors of AUC are usually based on the Mann-Whitney U test that assumes independence of sampling units. For ophthalmologic applications, it is desirable to assess risk prediction rules based on eye-specific outcome variables which are generally highly, but not perfectly correlated in fellow eyes [e.g. progression of individual eyes to age-related macular degeneration (AMD)]. In this article, we use the extended Mann-Whitney U test (Rosner and Glynn, Biometrics 65:188-197, 2009) for the case where subunits within a cluster may have different progression status and assess discrimination of different prediction rules in this setting. Both data analyses based on progression of AMD and simulation studies show reasonable accuracy of this extended Mann-Whitney U test to assess discrimination of eye-specific risk prediction rules.

Continuous time modelling with individually varying time intervals for oscillating and non-oscillating processes.

When designing longitudinal studies, researchers often aim at equal intervals. In practice, however, this goal is hardly ever met, with different time intervals between assessment waves and different time intervals between individuals being more the rule than the exception. One of the reasons for the introduction of continuous time models by means of structural equation modelling has been to deal with irregularly spaced assessment waves (e.g., Oud & Delsing, 2010). In the present paper we extend the approach to individually varying time intervals for oscillating and non-oscillating processes. In addition, we show not only that equal intervals are unnecessary but also that it can be advantageous to use unequal sampling intervals, in particular when the sampling rate is low. Two examples are provided to support our arguments. In the first example we compare a continuous time model of a bivariate coupled process with varying time intervals to a standard discrete time model to illustrate the importance of accounting for the exact time intervals. In the second example the effect of different sampling intervals on estimating a damped linear oscillator is investigated by means of a Monte Carlo simulation. We conclude that it is important to account for individually varying time intervals, and encourage researchers to conceive of longitudinal studies with different time intervals within and between individuals as an opportunity rather than a problem.

A new linear model-based approach for inferences about the mean area under the curve.

Outcome versus time data are commonly encountered in biomedical and clinical research. A common strategy adopted in analyzing such longitudinal data is to condense the repeated measurements on each individual into a single summary statistic such as the area under the response versus time curve. Standard parametric or non-parametric methods are then applied to perform inferences on the conditional area under the curve distribution. Disadvantages of this approach include the disregard of the within-subject variation in the longitudinal profile. We propose a general linear model approach, accounting for the within-subject variance, for estimation and hypothesis tests about the mean areas. Inferential properties of our approach are compared with those from standard methods of analysis using Monte Carlo simulation studies. The impact of missing data, within-subject heterogeneity and homogeneity of variance, are also evaluated. A real working example is used to illustrate the methodology. It is seen that the proposed approach is associated with a significant power advantage over traditional methods, especially when missing data are encountered.

Flexible parametric joint modelling of longitudinal and survival data.

The joint modelling of longitudinal and survival data is a highly active area of biostatistical research. The submodel for the longitudinal biomarker usually takes the form of a linear mixed effects model. We describe a flexible parametric approach for the survival submodel that models the log baseline cumulative hazard using restricted cubic splines. This approach overcomes limitations of standard parametric choices for the survival submodel, which can lack the flexibility to effectively capture the shape of the underlying hazard function. Numerical integration techniques, such as Gauss-Hermite quadrature, are usually required to evaluate both the cumulative hazard and the overall joint likelihood; however, by using a flexible parametric model, the cumulative hazard has an analytically tractable form, providing considerable computational benefits. We conduct an extensive simulation study to assess the proposed model, comparing it with a B-spline formulation, illustrating insensitivity of parameter estimates to the baseline cumulative hazard function specification. Furthermore, we compare non-adaptive and fully adaptive quadrature, showing the superiority of adaptive quadrature in evaluating the joint likelihood. We also describe a useful technique to simulate survival times from complex baseline hazard functions and illustrate the methods using an example data set investigating the association between longitudinal prothrombin index and survival of patients with liver cirrhosis, showing greater flexibility and improved stability with fewer parameters under the proposed model compared with the B-spline approach. We provide user-friendly Stata software.