Search

Background: Several indirect comparison methods, including network meta-analyses (NMAs), using individual patient data (IPD) have been developed to synthesize evidence from a network of trials. Although IPD indirect comparisons are published with increasing frequency in health care literature, there is no guidance on selecting the appropriate methodology and on reporting the methods and results.Methods: In this paper we examine the methods and reporting of indirect comparison methods using IPD. We searched MEDLINE, Embase, the Cochrane Library, and CINAHL from inception until October 2014. We included published and unpublished studies reporting a method, application, or review of indirect comparisons using IPD and at least three interventions.Results: We identified 37 papers, including a total of 33 empirical networks. Of these, only 9 (27 %) IPD-NMAs reported the existence of a study protocol, whereas 3 (9 %) studies mentioned that protocols existed without providing a reference. The 33 empirical networks included 24 (73 %) IPD-NMAs and 9 (27 %) matching adjusted indirect comparisons (MAICs). Of the 21 (64 %) networks with at least one closed loop, 19 (90 %) were IPD-NMAs, 13 (68 %) of which evaluated the prerequisite consistency assumption, and only 5 (38 %) of the 13 IPD-NMAs used statistical approaches. The median number of trials included per network was 10 (IQR 4-19) (IPD-NMA: 15 [IQR 8-20]; MAIC: 2 [IQR 3-5]), and the median number of IPD trials included in a network was 3 (IQR 1-9) (IPD-NMA: 6 [IQR 2-11]; MAIC: 2 [IQR 1-2]). Half of the networks (17; 52 %) applied Bayesian hierarchical models (14 one-stage, 1 two-stage, 1 used IPD as an informative prior, 1 unclear-stage), including either IPD alone or with aggregated data (AD). Models for dichotomous and continuous outcomes were available (IPD alone or combined with AD), as were models for time-to-event data (IPD combined with AD).Conclusions: One in three indirect comparison methods modeling IPD adjusted results from different trials to estimate effects as if they had come from the same, randomized, population. Key methodological and reporting elements (e.g., evaluation of consistency, existence of study protocol) were often missing from an indirect comparison paper. [ABSTRACT FROM AUTHOR]

Background: Even though investigating predictors of intervention success (e.g Cognitive Training, CT) is gaining more and more interest in the light of an individualized medicine, results on specific predictors of intervention success in the overall field are mixed and inconsistent due to different and sometimes inappropriate statistical methods used. Therefore, the present paper gives a guidance on the appropriate use of multiple regression analyses to identify predictors of CT and similar non-pharmacological interventions.Methods: We simulated data based on a predefined true model and ran a series of different analyses to evaluate their performance in retrieving the true model coefficients. The true model consisted of a 2 (between: experimental vs. control group) × 2 (within: pre- vs. post-treatment) design with two continuous predictors, one of which predicted the success in the intervention group and the other did not. In analyzing the data, we considered four commonly used dependent variables (post-test score, absolute change score, relative change score, residual score), five regression models, eight sample sizes, and four levels of reliability.Results: Our results indicated that a regression model including the investigated predictor, Group (experimental vs. control), pre-test score, and the interaction between the investigated predictor and the Group as predictors, and the absolute change score as the dependent variable seemed most convenient for the given experimental design. Although the pre-test score should be included as a predictor in the regression model for reasons of statistical power, its coefficient should not be interpreted because even if there is no true relationship, a negative and statistically significant regression coefficient commonly emerges.Conclusion: Employing simulation methods, theoretical reasoning, and mathematical derivations, we were able to derive recommendations regarding the analysis of data in one of the most prevalent experimental designs in research on CT and external predictors of CT success. These insights can contribute to the application of considered data analyses in future studies and facilitate cumulative knowledge gain. [ABSTRACT FROM AUTHOR]

Background: The sample size required to power a study to a nominal level in a paired comparative diagnostic accuracy study, i.e. studies in which the diagnostic accuracy of two testing procedures is compared relative to a gold standard, depends on the conditional dependence between the two tests - the lower the dependence the greater the sample size required. A priori, we usually do not know the dependence between the two tests and thus cannot determine the exact sample size required. One option is to use the implied sample size for the maximal negative dependence, giving the largest possible sample size. However, this is potentially wasteful of resources and unnecessarily burdensome on study participants as the study is likely to be overpowered. A more accurate estimate of the sample size can be determined at a planned interim analysis point where the sample size is re-estimated.Methods: This paper discusses a sample size estimation and re-estimation method based on the maximum likelihood estimates, under an implied multinomial model, of the observed values of conditional dependence between the two tests and, if required, prevalence, at a planned interim. The method is illustrated by comparing the accuracy of two procedures for the detection of pancreatic cancer, one procedure using the standard battery of tests, and the other using the standard battery with the addition of a PET/CT scan all relative to the gold standard of a cell biopsy. Simulation of the proposed method illustrates its robustness under various conditions.Results: The results show that the type I error rate of the overall experiment is stable using our suggested method and that the type II error rate is close to or above nominal. Furthermore, the instances in which the type II error rate is above nominal are in the situations where the lowest sample size is required, meaning a lower impact on the actual number of participants recruited.Conclusion: We recommend multinomial model maximum likelihood estimation of the conditional dependence between paired diagnostic accuracy tests at an interim to reduce the number of participants required to power the study to at least the nominal level.Trial Registration: ISRCTN ISRCTN73852054 . Registered 9th of January 2015. Retrospectively registered. [ABSTRACT FROM AUTHOR]

Background: This paper explores the importance of electronic medical records (EMR) for predicting 30-day all-cause non-elective readmission risk of patients and presents a comparison of prediction performance of commonly used methods.Methods: The data are extracted from eight Advocate Health Care hospitals. Index admissions are excluded from the cohort if they are observation, inpatient admissions for psychiatry, skilled nursing, hospice, rehabilitation, maternal and newborn visits, or if the patient expires during the index admission. Data are randomly and repeatedly divided into fitting and validating sets for cross validations. Approaches including LACE, STEPWISE logistic, LASSO logistic, and AdaBoost, are compared with sample sizes varying from 2,500 to 80,000.Results: Our results confirm that LACE has moderate discrimination power with the area under receiver operating characteristic curve (AUC) around 0.65-0.66, which can be improved to 0.73-0.74 when additional variables from EMR are considered. These variables include Inpatient in the last six months, Number of emergency room visits or inpatients in the last year, Braden score, Polypharmacy, Employment status, Discharge disposition, Albumin level, and medical condition variables such as Leukemia, Malignancy, Renal failure with hemodialysis, History of alcohol substance abuse, Dementia and Trauma. When sample size is small (≤5000), LASSO is the best; when sample size is large (≥20,000), the predictive performance is similar. The STEPWISE method has a slightly lower AUC (0.734) comparing to LASSO (0.737) and AdaBoost (0.737). More than one half of the selected predictors can be false positives when using a single method and a single division of fitting/validating data.Conclusions: True predictors can be identified by repeatedly dividing data into fitting/validating subsets and referring the final model based on summarizing results. LASSO is a better alternative to the STEPWISE logistic regression, especially when sample size is not large. The evidence for adequate sample size can be explored by fitting models on gradually reduced samples. Our model comparison strategy is not only good for 30-day all-cause non-elective readmission risk predictions, but also applicable to other types of predictive models in clinical studies. [ABSTRACT FROM AUTHOR]

Background: As a part of a larger Health Technology Assessment (HTA), the measurement error of a device used to monitor the hemoglobin concentration of a patient undergoing surgery, as well as its decision consequences, were to be estimated from published data.Methods: A Bayesian hierarchical model of measurement error, allowing the meta-analytic estimation of both central and dispersion parameters (under the assumption of normality of measurement errors) is proposed and applied to published data; the resulting potential decision errors are deduced from this estimation. The same method is used to assess the impact of an initial calibration.Results: The posterior distributions are summarized as mean ± sd (credible interval). The fitted model exhibits a modest mean expected error (0.24 ± 0.73 (-1.23 1.59) g/dL) and a large variability (mean absolute expected error 1.18 ± 0.92 (0.05 3.36) g/dL). The initial calibration modifies the bias (-0.20 ± 0.87 (-1.99 1.49) g/dL), but the variability remains almost as large (mean absolute expected error 1.05 ± 0.87 (0.04 3.21) g/dL). This entails a potential decision error ("false positive" or "false negative") for about one patient out of seven.Conclusions: The proposed hierarchical model allows the estimation of the variability from published aggregates, and allows the modeling of the consequences of this variability in terms of decision errors. For the device under assessment, these potential decision errors are clinically problematic. [ABSTRACT FROM AUTHOR]

Background: There are inherent methodological challenges in the measurement of mental health problems in longitudinal research. There is constant development in definitions, taxonomies and demands concerning the properties of mental health measurements. The aim of this paper was to construct composite measures of mental health problems (according to today's standard) from single questionnaire items devised in the early 1980s, and to evaluate their internal consistency and factorial invariance across the life course using the Northern Swedish Cohort.Methods: All pupils in the last year of compulsory school in Luleå in 1981 (n = 1083) form a prospective cohort study where the participants have been followed with questionnaires from the age of 16 (in 1981) until the age of 43 (in 2008). We created and tested the following composite measures from self-reports at each follow-up: depressive symptoms, anxiety symptoms, functional somatic symptoms, modified GHQ and positive health. Validity and internal consistency were tested by confirmatory factor analysis, including tests of factorial invariance over time.Results: As an overall assessment, the results showed that the composite measures (based on more than 30-year-old single item questions) are likely to have acceptable factorial invariance as well as internal consistency over time.Conclusions: Testing the properties of the mental health measures used in older studies according to the standards of today is of great importance in longitudinal research. Our study demonstrates that composite measures of mental health problems can be constructed from single items which are more than 30 years old and that these measures seem to have the same factorial structure and internal consistency across a significant part of the life course. Thus, it can be possible to overcome some specific inherent methodological challenges in using historical data in longitudinal research. [ABSTRACT FROM AUTHOR]

Background: Surveys are commonly used in health research to assess patient satisfaction with hospital care. Achieving an adequate response rate, in the face of declining trends over time, threatens the quality and reliability of survey results. This paper evaluates a strategy to increase the response rate in a postal satisfaction survey with women who had recently given birth.Methods: A sample of 2048 Australian women who had recently given birth at seven maternity units in New South Wales were invited to participate in a postal survey about their recent experiences with maternity care. The study design included a randomised controlled trial that tested two types of pre-notification letter (with or without the option of opting out of the survey). The study also explored the acceptability of a request for consent to link survey data with existing routinely collected health data (omitting the latter data items from the survey reduced survey length and participant burden). This consent was requested of all women.Results: The survey had an overall response rate of 46% (913 completed surveys returned, total sample 1989). Women receiving the pre-notification letter with the option of opting out of the survey were more likely to actively decline to participate than women receiving the letter without this option, although the overall numbers of women declining were small (27 versus 12). Letter type was not significantly associated with the return of a completed survey. Among women who completed the survey, 97% gave consent to link their survey data with existing health data.Conclusions: The two types of pre-notification letters used in our study did not influence the survey response rate. However, seeking consent for record linkage was highly acceptable to women who completed the survey, and represents an important strategy to add to the arsenal for designing and implementing effective surveys. In addition to aspects of survey design, future research should explore how to more effectively influence personal constructs that contribute to the decision to participate in surveys. [ABSTRACT FROM AUTHOR]

Background: Although some 300 million Indian children travel to school every day, little is known about how they get there. This information is important for transport planners and public health authorities. This paper presents the development of a self-administered questionnaire and examines its reliability and validity in estimating distance and mode of travel to school in a low resource urban setting.Methods: We developed a questionnaire on children's travel to school. We assessed test re-test reliability by repeating the questionnaire one week later (n = 61). The questionnaire was improved and re-tested (n = 68). We examined the convergent validity of distance estimates by comparing estimates based on the nearest landmark to children's homes with a 'gold standard' based on one-to-one interviews with children using detailed maps (n = 50).Results: Most questions showed fair to almost perfect agreement. Questions on usual mode of travel (κ 0.73- 0.84) and road injury (κ 0.61- 0.72) were found to be more reliable than those on parental permissions (κ 0.18- 0.30), perception of safety (κ 0.00- 0.54), and physical activity (κ -0.01- 0.07). The distance estimated by the nearest landmark method was not significantly different than the in-depth method for walking , 52 m [95 % CI -32 m to 135 m], 10 % of the mean difference, and for walking and cycling combined, 65 m [95 % CI -30 m to 159 m], 11 % of the mean difference. For children who used motorized transport (excluding private school bus), the nearest landmark method under-estimated distance by an average of 325 metres [95 % CI -664 m to 1314 m], 15 % of the mean difference.Conclusions: A self-administered questionnaire was found to provide reliable information on the usual mode of travel to school, and road injury, in a small sample of children in Hyderabad, India. The 'nearest landmark' method can be applied in similar low-resource settings, for a reasonably accurate estimate of the distance from a child's home to school. [ABSTRACT FROM AUTHOR]

Background: Inconsistent responding is a type of invalid responding, which occurs on self-report surveys and threatens the reliability and validity of study results. This secondary analysis evaluated the utility of identifying inconsistent responses as a real-time, direct method to improve quality during data collection for an Internet-based RCT.Methods: The cannabis subscale of the Alcohol, Smoking and Substance Involvement Screening Test (ASSIST) was administered as part of eligibility screening for the RCT. Following the consent procedure, the cannabis subscale was repeated during the baseline interview. Responses were automatically compared and individuals with inconsistent responses were screened out.Results: Nearly half of those initially eligible for the RCT were subsequently screened out for data quality issues (n = 626, 45.3%). Between-group bivariate analysis found that those screened out (OUT) were significantly older (OUT = 39.5 years (SD = 13.9), IN = 35.7 years (SD = 12.9), p < .001), more had annual incomes less than $20,000CND (OUT = 58.3%, IN = 53.0%, p = .047), used cannabis less often in the past 30 days (OUT = 23.3 days (SD = 9.7), IN = 24.8 days (SD = 11.3), p < .006), and had lower total ASSIST scores at screener (OUT = 19.3 (SD = 8.0), IN = 23.8 (SD = 10.4), p < .001) and baseline (OUT = 17.5 (SD = 7.9), IN = 23.3 (SD = 10.3), p < .001) compared to participants who were screened in to the RCT.Conclusion: Inconsistent responding may occur at high rates in Internet research and direct methods to identify invalid responses are needed. Comparing responses for consistency can be programmed in Internet surveys to automatically screen participants during recruitment and reduce the need for post-hoc data cleaning. [ABSTRACT FROM AUTHOR]

Background: AMSTAR-2 ('A Measurement Tool to Assess Systematic Reviews, version 2') and ROBIS ('Risk of Bias in Systematic Reviews') are independent instruments used to assess the quality of conduct of systematic reviews/meta-analyses (SR/MAs). The degree of overlap in methodological constructs together with the reliability and any methodological gaps have not been systematically assessed and summarized in the field of nutrition.Methods: We performed a systematic survey of MEDLINE, EMBASE, and the Cochrane Library for SR/MAs published between January 2010 and November 2018 that examined the effects of any nutritional intervention/exposure for cancer prevention. We followed a systematic review approach including two independent reviewers at each step of the process. For AMSTAR-2 (16 items) and ROBIS (21 items), we assessed the similarities, the inter-rater reliability (IRR) and any methodological limitations of the instruments. Our protocol for the survey was registered in PROSPERO (CRD42019121116).Results: We found 4 similar domain constructs based on 11 comparisons from a total of 12 AMSTAR-2 and 14 ROBIS items. Ten comparisons were considered fully overlapping. Based on Gwet's agreement coefficients, six comparisons provided almost perfect (> 0.8), three substantial (> 0.6), and one a moderate level of agreement (> 0.4). While there is considerable overlap in constructs, AMSTAR-2 uniquely addresses explaining the selection of study designs for inclusion, reporting on excluded studies with justification, sources of funding of primary studies, and reviewers' conflict of interest. By contrast, ROBIS uniquely addresses appropriateness and restrictions within eligibility criteria, reducing risk of error in risk of bias (RoB) assessments, completeness of data extracted for analyses, the inclusion of all necessary studies for analyses, and adherence to predefined analysis plan.Conclusions: Among the questions on AMSTAR-2 and ROBIS, 70.3% (26/37 items) address the same or similar methodological constructs. While the IRR of these constructs was moderate to perfect, there are unique methodological constructs that each instrument independently addresses. Notably, both instruments do not address the reporting of absolute estimates of effect or the overall certainty of the evidence, items that are crucial for users' wishing to interpret the importance of SR/MA results. [ABSTRACT FROM AUTHOR]

Background: Network meta-analysis (NMA) has attracted growing interest in evidence-based medicine. Consistency between different sources of evidence is fundamental to the reliability of the NMA results. The purpose of the present study was to estimate the prevalence of evidence of inconsistency and describe its association with different NMA characteristics.Methods: We updated our collection of NMAs with articles published up to July 2018. We included networks with randomised clinical trials, at least four treatment nodes, at least one closed loop, a dichotomous primary outcome, and available arm-level data. We assessed consistency using the design-by-treatment interaction (DBT) model and testing all the inconsistency parameters globally through the Wald-type chi-squared test statistic. We estimated the prevalence of evidence of inconsistency and its association with different network characteristics (e.g., number of studies, interventions, intervention comparisons, loops). We evaluated the influence of the network characteristics on the DBT p-value via a multivariable regression analysis and the estimated Pearson correlation coefficients. We also evaluated heterogeneity in NMA (consistency) and DBT (inconsistency) random-effects models.Results: We included 201 published NMAs. The p-value of the design-by-treatment interaction (DBT) model was lower than 0.05 in 14% of the networks and lower than 0.10 in 20% of the networks. Networks including many studies and comparing few interventions were more likely to have small DBT p-values (less than 0.10), which is probably because they yielded more precise estimates and power to detect differences between designs was higher. In the presence of inconsistency (DBT p-value lower than 0.10), the consistency model displayed higher heterogeneity than the DBT model.Conclusions: Our findings show that inconsistency was more frequent than what would be expected by chance, suggesting that researchers should devote more resources to exploring how to mitigate inconsistency. The results of this study highlight the need to develop strategies to detect inconsistency (because of the relatively high prevalence of evidence of inconsistency in published networks), and particularly in cases where the existing tests have low power. [ABSTRACT FROM AUTHOR]

Background: Null hypothesis significance testing (NHST) is among the most frequently employed methods in the biomedical sciences. However, the problems of NHST and p-values have been discussed widely and various Bayesian alternatives have been proposed. Some proposals focus on equivalence testing, which aims at testing an interval hypothesis instead of a precise hypothesis. An interval hypothesis includes a small range of parameter values instead of a single null value and the idea goes back to Hodges and Lehmann. As researchers can always expect to observe some (although often negligibly small) effect size, interval hypotheses are more realistic for biomedical research. However, the selection of an equivalence region (the interval boundaries) often seems arbitrary and several Bayesian approaches to equivalence testing coexist.Methods: A new proposal is made how to determine the equivalence region for Bayesian equivalence tests based on objective criteria like type I error rate and power. Existing approaches to Bayesian equivalence testing in the two-sample setting are discussed with a focus on the Bayes factor and the region of practical equivalence (ROPE). A simulation study derives the necessary results to make use of the new method in the two-sample setting, which is among the most frequently carried out procedures in biomedical research.Results: Bayesian Hodges-Lehmann tests for statistical equivalence differ in their sensitivity to the prior modeling, power, and the associated type I error rates. The relationship between type I error rates, power and sample sizes for existing Bayesian equivalence tests is identified in the two-sample setting. Results allow to determine the equivalence region based on the new method by incorporating such objective criteria. Importantly, results show that not only can prior selection influence the type I error rate and power, but the relationship is even reverse for the Bayes factor and ROPE based equivalence tests.Conclusion: Based on the results, researchers can select between the existing Bayesian Hodges-Lehmann tests for statistical equivalence and determine the equivalence region based on objective criteria, thus improving the reproducibility of biomedical research. [ABSTRACT FROM AUTHOR]

Background: In an inter-rater agreement study, if two raters tend to rate considering different aspects of the subject of interest or have different experience levels, a grey zone occurs among the levels of a square contingency table showing the inter-rater agreement. These grey zones distort the degree of agreement between raters and negatively impact the decisions based on the inter-rater agreement tables. In this sense, it is important to know how the existence of a grey zone impacts the inter-rater agreement coefficients to choose the most reliable agreement coefficient against the grey zones to reach out with more reliable decisions.Methods: In this article, we propose two approaches to create grey zones in simulations setting and conduct an extensive Monte Carlo simulation study to figure out the impact of having grey zones on the weighted inter-rater agreement measures for ordinal tables over a comprehensive simulation space.Results: The weighted inter-rater agreement coefficients are not reliable against the existence of grey zones. Increasing sample size and the number of categories in the agreement table decreases the accuracy of weighted inter-rater agreement measures when there is a grey zone. When the degree of agreement between the raters is high, the agreement measures are not significantly impacted by the existence of grey zones. However, if there is a medium to low degree of inter-rater agreement, all the weighted coefficients are more or less impacted.Conclusions: It is observed in this study that the existence of grey zones has a significant negative impact on the accuracy of agreement measures especially for a low degree of true agreement and high sample and tables sizes. In general, Gwet's AC2 and Brennan-Prediger's κ with quadratic or ordinal weights are reliable against the grey zones. [ABSTRACT FROM AUTHOR]

Background: Studies of agreement examine the distance between readings made by different devices or observers measuring the same quantity. If the values generated by each device are close together most of the time then we conclude that the devices agree. Several different agreement methods have been described in the literature, in the linear mixed modelling framework, for use when there are time-matched repeated measurements within subjects.Methods: We provide a tutorial to help guide practitioners when choosing among different methods of assessing agreement based on a linear mixed model assumption. We illustrate the use of five methods in a head-to-head comparison using real data from a study involving Chronic Obstructive Pulmonary Disease (COPD) patients and matched repeated respiratory rate observations. The methods used were the concordance correlation coefficient, limits of agreement, total deviation index, coverage probability, and coefficient of individual agreement.Results: The five methods generated similar conclusions about the agreement between devices in the COPD example; however, some methods emphasized different aspects of the between-device comparison, and the interpretation was clearer for some methods compared to others.Conclusions: Five different methods used to assess agreement have been compared in the same setting to facilitate understanding and encourage the use of multiple agreement methods in practice. Although there are similarities between the methods, each method has its own strengths and weaknesses which are important for researchers to be aware of. We suggest that researchers consider using the coverage probability method alongside a graphical display of the raw data in method comparison studies. In the case of disagreement between devices, it is important to look beyond the overall summary agreement indices and consider the underlying causes. Summarising the data graphically and examining model parameters can both help with this. [ABSTRACT FROM AUTHOR]

Background: The comparison of the performance of two binary diagnostic tests is an important topic in Clinical Medicine. The most frequent type of sample design to compare two binary diagnostic tests is the paired design. This design consists of applying the two binary diagnostic tests to all of the individuals in a random sample, where the disease status of each individual is known through the application of a gold standard. This article presents an R program to compare parameters of two binary tests subject to a paired design.Results: The "compbdt" program estimates the sensitivity and the specificity, the likelihood ratios and the predictive values of each diagnostic test applying the confidence intervals with the best asymptotic performance. The program compares the sensitivities and specificities of the two diagnostic tests simultaneously, as well as the likelihood ratios and the predictive values, applying the global hypothesis tests with the best performance in terms of type I error and power. When the global hypothesis test is significant, the causes of the significance are investigated solving the individual hypothesis tests and applying the multiple comparison method of Holm. The most optimal confidence intervals are also calculated for the difference or ratio between the respective parameters. Based on the data observed in the sample, the program also estimates the probability of making a type II error if the null hypothesis is not rejected, or estimates the power if the if the alternative hypothesis is accepted. The "compbdt" program provides all the necessary results so that the researcher can easily interpret them. The estimation of the probability of making a type II error allows the researcher to decide about the reliability of the null hypothesis when this hypothesis is not rejected. The "compbdt" program has been applied to a real example on the diagnosis of coronary artery disease.Conclusions: The "compbdt" program is one which is easy to use and allows the researcher to compare the most important parameters of two binary tests subject to a paired design. The "compbdt" program is available as supplementary material. [ABSTRACT FROM AUTHOR]

Background: Prognostic models often show poor performance when applied to independent validation data sets. We illustrate how treatment use in a validation set can affect measures of model performance and present the uses and limitations of available analytical methods to account for this using simulated data.Methods: We outline how the use of risk-lowering treatments in a validation set can lead to an apparent overestimation of risk by a prognostic model that was developed in a treatment-naïve cohort to make predictions of risk without treatment. Potential methods to correct for the effects of treatment use when testing or validating a prognostic model are discussed from a theoretical perspective.. Subsequently, we assess, in simulated data sets, the impact of excluding treated individuals and the use of inverse probability weighting (IPW) on the estimated model discrimination (c-index) and calibration (observed:expected ratio and calibration plots) in scenarios with different patterns and effects of treatment use.Results: Ignoring the use of effective treatments in a validation data set leads to poorer model discrimination and calibration than would be observed in the untreated target population for the model. Excluding treated individuals provided correct estimates of model performance only when treatment was randomly allocated, although this reduced the precision of the estimates. IPW followed by exclusion of the treated individuals provided correct estimates of model performance in data sets where treatment use was either random or moderately associated with an individual's risk when the assumptions of IPW were met, but yielded incorrect estimates in the presence of non-positivity or an unobserved confounder.Conclusions: When validating a prognostic model developed to make predictions of risk without treatment, treatment use in the validation set can bias estimates of the performance of the model in future targeted individuals, and should not be ignored. When treatment use is random, treated individuals can be excluded from the analysis. When treatment use is non-random, IPW followed by the exclusion of treated individuals is recommended, however, this method is sensitive to violations of its assumptions. [ABSTRACT FROM AUTHOR]

Background: Composite endpoints comprising hospital admissions and death are the primary outcome in many cardiovascular clinical trials. For statistical analysis, a Cox proportional hazards model for the time to first event is commonly applied. There is an ongoing debate on whether multiple episodes per individual should be incorporated into the primary analysis. While the advantages in terms of power are readily apparent, potential biases have been mostly overlooked so far.Methods: Motivated by a randomized controlled clinical trial in heart failure patients, we use directed acyclic graphs (DAG) to investigate potential sources of bias in treatment effect estimates, depending on whether only the first or multiple episodes are considered. The biases first are explained in simplified examples and then more thoroughly investigated in simulation studies that mimic realistic patterns.Results: Particularly the Cox model is prone to potentially severe selection bias and direct effect bias, resulting in underestimation when restricting the analysis to first events. We find that both kinds of bias can simultaneously be reduced by adequately incorporating recurrent events into the analysis model. Correspondingly, we point out appropriate proportional hazards-based multi-state models for decreasing bias and increasing power when analyzing multiple-episode composite endpoints in randomized clinical trials.Conclusions: Incorporating multiple episodes per individual into the primary analysis can reduce the bias of a treatment's total effect estimate. Our findings will help to move beyond the paradigm of considering first events only for approaches that use more information from the trial and augment interpretability, as has been called for in cardiovascular research. [ABSTRACT FROM AUTHOR]

Background: The analysis of correlated binary data is commonly addressed through the use of conditional models with random effects included in the systematic component as opposed to generalized estimating equations (GEE) models that addressed the random component. Since the joint distribution of the observations is usually unknown, the conditional distribution is a natural approach. Our objective was to compare the fit of different binary models for correlated data in Tabaco use. We advocate that the joint modeling of the mean and dispersion may be at times just as adequate. We assessed the ability of these models to account for the intraclass correlation. In so doing, we concentrated on fitting logistic regression models to address smoking behaviors.Methods: Frequentist and Bayes' hierarchical models were used to predict conditional probabilities, and the joint modeling (GLM and GAM) models were used to predict marginal probabilities. These models were fitted to National Longitudinal Study of Adolescent to Adult Health (Add Health) data for Tabaco use.Results: We found that people were less likely to smoke if they had higher income, high school or higher education and religious. Individuals were more likely to smoke if they had abused drug or alcohol, spent more time on TV and video games, and been arrested. Moreover, individuals who drank alcohol early in life were more likely to be a regular smoker. Children who experienced mistreatment from their parents were more likely to use Tabaco regularly.Conclusions: The joint modeling of the mean and dispersion models offered a flexible and meaningful method of addressing the intraclass correlation. They do not require one to identify random effects nor distinguish from one level of the hierarchy to the other. Moreover, once one can identify the significant random effects, one can obtain similar results to the random coefficient models. We found that the set of marginal models accounting for extravariation through the additional dispersion submodel produced similar results with regards to inferences and predictions. Moreover, both marginal and conditional models demonstrated similar predictive power. [ABSTRACT FROM AUTHOR]

Background: The attributable risk (AR) measures the proportion of disease cases that can be attributed to an exposure in the population. Several definitions and estimation methods have been proposed for survival data.Methods: Using simulations, we compared four methods for estimating AR defined in terms of survival functions: two nonparametric methods based on Kaplan-Meier's estimator, one semiparametric based on Cox's model, and one parametric based on the piecewise constant hazards model, as well as one simpler method based on estimated exposure prevalence at baseline and Cox's model hazard ratio. We considered a fixed binary exposure with varying exposure probabilities and strengths of association, and generated event times from a proportional hazards model with constant or monotonic (decreasing or increasing) Weibull baseline hazard, as well as from a nonproportional hazards model. We simulated 1,000 independent samples of size 1,000 or 10,000. The methods were compared in terms of mean bias, mean estimated standard error, empirical standard deviation and 95% confidence interval coverage probability at four equally spaced time points.Results: Under proportional hazards, all five methods yielded unbiased results regardless of sample size. Nonparametric methods displayed greater variability than other approaches. All methods showed satisfactory coverage except for nonparametric methods at the end of follow-up for a sample size of 1,000 especially. With nonproportional hazards, nonparametric methods yielded similar results to those under proportional hazards, whereas semiparametric and parametric approaches that both relied on the proportional hazards assumption performed poorly. These methods were applied to estimate the AR of breast cancer due to menopausal hormone therapy in 38,359 women of the E3N cohort.Conclusion: In practice, our study suggests to use the semiparametric or parametric approaches to estimate AR as a function of time in cohort studies if the proportional hazards assumption appears appropriate. [ABSTRACT FROM AUTHOR]

Background: In stark contrast to network-centric view for complex disease, regression-based methods are preferred in disease prediction, especially for epidemiologists and clinical professionals. It remains a controversy whether the network-based methods have advantageous performance than regression-based methods, and to what extent do they outperform.Methods: Simulations under different scenarios (the input variables are independent or in network relationship) as well as an application were conducted to assess the prediction performance of four typical methods including Bayesian network, neural network, logistic regression and regression splines.Results: The simulation results reveal that Bayesian network showed a better performance when the variables were in a network relationship or in a chain structure. For the special wheel network structure, logistic regression had a considerable performance compared to others. Further application on GWAS of leprosy show Bayesian network still outperforms other methods.Conclusion: Although regression-based methods are still popular and widely used, network-based approaches should be paid more attention, since they capture the complex relationship between variables. [ABSTRACT FROM AUTHOR]

Background: When data needed to inform parameters in decision models are lacking, formal elicitation of expert judgement can be used to characterise parameter uncertainty. Although numerous methods for eliciting expert opinion as probability distributions exist, there is little research to suggest whether one method is more useful than any other method. This study had three objectives: (i) to obtain subjective probability distributions characterising parameter uncertainty in the context of a health technology assessment; (ii) to compare two elicitation methods by eliciting the same parameters in different ways; (iii) to collect subjective preferences of the experts for the different elicitation methods used.Methods: Twenty-seven clinical experts were invited to participate in an elicitation exercise to inform a published model-based cost-effectiveness analysis of alternative treatments for prostate cancer. Participants were individually asked to express their judgements as probability distributions using two different methods - the histogram and hybrid elicitation methods - presented in a random order. Individual distributions were mathematically aggregated across experts with and without weighting. The resulting combined distributions were used in the probabilistic analysis of the decision model and mean incremental cost-effectiveness ratios and the expected values of perfect information (EVPI) were calculated for each method, and compared with the original cost-effectiveness analysis. Scores on the ease of use of the two methods and the extent to which the probability distributions obtained from each method accurately reflected the expert's opinion were also recorded.Results: Six experts completed the task. Mean ICERs from the probabilistic analysis ranged between £162,600-£175,500 per quality-adjusted life year (QALY) depending on the elicitation and weighting methods used. Compared to having no information, use of expert opinion decreased decision uncertainty: the EVPI value at the £30,000 per QALY threshold decreased by 74-86 % from the original cost-effectiveness analysis. Experts indicated that the histogram method was easier to use, but attributed a perception of more accuracy to the hybrid method.Conclusions: Inclusion of expert elicitation can decrease decision uncertainty. Here, choice of method did not affect the overall cost-effectiveness conclusions, but researchers intending to use expert elicitation need to be aware of the impact different methods could have. [ABSTRACT FROM AUTHOR]

Background: Wastewater-based epidemiology (WBE) is a novel approach in drug use epidemiology which aims to monitor the extent of use of various drugs in a community. In this study, we investigate functional principal component analysis (FPCA) as a tool for analysing WBE data and compare it to traditional principal component analysis (PCA) and to wavelet principal component analysis (WPCA) which is more flexible temporally.Methods: We analysed temporal wastewater data from 42 European cities collected daily over one week in March 2013. The main temporal features of ecstasy (MDMA) were extracted using FPCA using both Fourier and B-spline basis functions with three different smoothing parameters, along with PCA and WPCA with different mother wavelets and shrinkage rules. The stability of FPCA was explored through bootstrapping and analysis of sensitivity to missing data.Results: The first three principal components (PCs), functional principal components (FPCs) and wavelet principal components (WPCs) explained 87.5-99.6 % of the temporal variation between cities, depending on the choice of basis and smoothing. The extracted temporal features from PCA, FPCA and WPCA were consistent. FPCA using Fourier basis and common-optimal smoothing was the most stable and least sensitive to missing data.Conclusion: FPCA is a flexible and analytically tractable method for analysing temporal changes in wastewater data, and is robust to missing data. WPCA did not reveal any rapid temporal changes in the data not captured by FPCA. Overall the results suggest FPCA with Fourier basis functions and common-optimal smoothing parameter as the most accurate approach when analysing WBE data. [ABSTRACT FROM AUTHOR]

Background: Action research (AR) and randomized controlled trials (RCTs) are usually considered to be theoretically and practically incompatible. However, we argue that their respective strengths and weaknesses can be complementary. We illustrate our argument from a recent study assessing the effect of telemonitoring on health-related quality of life, self-care, hospital use, costs and the experiences of patients, informal carers and health care professionals in two urban hospital services and one remote rural primary care service in New Zealand.Methods: Data came from authors' observations and field notes of discussions with three groups: the healthcare providers and healthcare consumers who participated in the research, and a group of 17 researchers and collaborators. The consumers had heart failure (Site A, urban), airways disease (Site B, urban), and diabetes (Site C, rural). The research ran from 2008 (project inception) until 2012 (project close-off). Researchers came from a wide range of disciplines. Both RCT and AR methods were recognised from early in the process but often worked in parallel rather than together. In retrospect, we have mapped our observed research processes to the AR cycle characteristics (creation of communicative space, democracy and participation, iterative learning and improvement, emergence, and accommodation of different ways of knowing).Results: We describe the context, conduct and outcomes of the telemonitoring trial, framing the overall process in the language of AR. Although not fully articulated at the time, AR processes made the RCT sensitive to important context, e.g. clinical processes. They resulted in substantive changes to the design and conduct of the RCT, and to interpretation and uptake of findings, e.g. a simpler technology procurement process emerged. Creating a communicative space enabled co-design between the researcher group and collaborators from the provider participant group, and a stronger RCT design.Conclusions: It appears possible to enhance the utility of RCTs by explicitly embedding them in an AR framework to shape stronger RCT design. The AR process and characteristics may enable researchers to evaluate telehealth while enhancing rather than compromising the quality of an RCT, where research results are returned to practice as part of the research process.Trial Registration: Australian New Zealand Clinical Trials Registry, reference ACTRN12610000269033 . [ABSTRACT FROM AUTHOR]

Background: Traditional analyses of standard case-control studies using logistic regression do not allow estimation of time-varying associations between exposures and the outcome. We present two approaches which allow this. The motivation is a study of vaccine efficacy as a function of time since vaccination.Methods: Our first approach is to estimate time-varying exposure-outcome associations by fitting a series of logistic regressions within successive time periods, reusing controls across periods. Our second approach treats the case-control sample as a case-cohort study, with the controls forming the subcohort. In the case-cohort analysis, controls contribute information at all times they are at risk. Extensions allow left truncation, frequency matching and, using the case-cohort analysis, time-varying exposures. Simulations are used to investigate the methods.Results: The simulation results show that both methods give correct estimates of time-varying effects of exposures using standard case-control data. Using the logistic approach there are efficiency gains by reusing controls over time and care should be taken over the definition of controls within time periods. However, using the case-cohort analysis there is no ambiguity over the definition of controls. The performance of the two analyses is very similar when controls are used most efficiently under the logistic approach.Conclusions: Using our methods, case-control studies can be used to estimate time-varying exposure-outcome associations where they may not previously have been considered. The case-cohort analysis has several advantages, including that it allows estimation of time-varying associations as a continuous function of time, while the logistic regression approach is restricted to assuming a step function form for the time-varying association. [ABSTRACT FROM AUTHOR]

Background: Sample size planning for longitudinal data is crucial when designing mediation studies because sufficient statistical power is not only required in grant applications and peer-reviewed publications, but is essential to reliable research results. However, sample size determination is not straightforward for mediation analysis of longitudinal design.Methods: To facilitate planning the sample size for longitudinal mediation studies with a multilevel mediation model, this article provides the sample size required to achieve 80% power by simulations under various sizes of the mediation effect, within-subject correlations and numbers of repeated measures. The sample size calculation is based on three commonly used mediation tests: Sobel's method, distribution of product method and the bootstrap method.Results: Among the three methods of testing the mediation effects, Sobel's method required the largest sample size to achieve 80% power. Bootstrapping and the distribution of the product method performed similarly and were more powerful than Sobel's method, as reflected by the relatively smaller sample sizes. For all three methods, the sample size required to achieve 80% power depended on the value of the ICC (i.e., within-subject correlation). A larger value of ICC typically required a larger sample size to achieve 80% power. Simulation results also illustrated the advantage of the longitudinal study design. The sample size tables for most encountered scenarios in practice have also been published for convenient use.Conclusions: Extensive simulations study showed that the distribution of the product method and bootstrapping method have superior performance to the Sobel's method, but the product method was recommended to use in practice in terms of less computation time load compared to the bootstrapping method. A R package has been developed for the product method of sample size determination in mediation longitudinal study design. [ABSTRACT FROM AUTHOR]