Effects of categorization method, regression type, and variable distribution on the inflation of Type-I error rate when categorizing a confounding variable

Non-invasive brain stimulation in the treatment of generalized anxiety disorder: A systematic review and meta-analysis

BACKGROUND

Non-invasive brain stimulation (NIBS), including repetitive transcranial magnetic stimulation (rTMS), continuous theta-burst stimulation (cTBS), and transcranial direct current stimulation (tDCS), is an emerging intervention that has been used to treat various mental illnesses. However, previous studies have not comprehensively compared the efficacies of various NIBS modalities in alleviating anxiety symptoms among patients with generalized anxiety disorder (GAD). Therefore, this study conducted a systematic review and meta-analysis to assess the efficacy of NIBS for patients with GAD.

METHODS

A systematic search of four major bibliographic databases (Embase, PubMed, Web of Science and The Cochrane Library) was conducted from inception dates to November 26, 2023 to identify eligible studies. The data were analyzed using a random-effects model.

RESULTS

Seven randomized controlled trials (RCTs) were included in the meta-analysis. Significant differences were found in changes in Hamilton anxiety rating scale (HARS) scores, study-defined response, and remission between the intervention and control groups. Moreover, the intervention groups experienced a significantly higher frequency of headaches.

CONCLUSION

The results revealed that interventions improved GAD compared to control groups. cTBS and rTMS exhibited better treatment efficacy than tDCS, which did not appear to have a significant therapeutic effect. Longer follow-up periods and larger sample sizes are required in future RCTs.

TRIAL REGISTRATION

This meta-analysis was conducted in accordance with PRISMA guidelines and registered at PROSPERO (https://www.crd.york.ac.uk/PROSPERO/, CRD42023466285).

Mental Illness Diagnoses May Not Cause All Mental Symptoms: A Simulation Study for Major Depressive Episodes, Dysthymic Disorder, and Manic Episodes

Objectives This study aims to understand the statistical significance of the associations between diagnoses and symptoms based on simulations that have been used to understand the interpretability of mental illness diagnoses. Methods The symptoms for the diagnosis of major depressive episodes, dysthymic disorder, and manic episodes were extracted from the Diagnostic and Statistical Manual of Mental Disorders, Fourth Edition, Text Revision (DSM-IV-TR, American Psychiatric Association, Philadelphia, Pennsylvania). Without real-world symptom data, we simulated populations using various combinations of symptom prevalence and correlations. Assuming symptoms occurred with similar prevalence and correlations, for each combination of symptom prevalence (0.05, 0.1, 0.3, 0.5, and 0.7) and correlation (0, 0.1, 0.4, 0.7, and 0.9), 100 cohorts with 10,000 individuals were randomly created. Diagnoses were made according to the DSM-IV-TR criteria. The associations between the diagnoses and their input symptoms were quantified with odds ratios and correlation coefficients. P-values from 100 cohorts for each combination of symptom prevalence and correlation were summarized. Results Three mental illness diagnoses were not significantly correlated with their own symptoms in all simulations, particularly when symptoms were not correlated, except for the symptom in the major criteria of major depressive episodes or dysthymic disorder. The symptoms for the diagnosis of major depressive episodes and dysthymic disorder were significantly correlated with these two diagnoses in some simulations, assuming 0.1, 0.4, 0.7, or 0.9 symptom correlations, except for one symptom. The overlap in the input symptoms for the diagnosis of major depressive episodes and dysthymic disorder also leads to significant correlations between these two diagnoses, assuming 0.1, 0.4, 0.7, and 0.9 correlations between input symptoms. Manic episodes are not significantly associated with the input symptoms of major depressive episodes and dysthymic disorder. Conclusion There are challenges to establish the causation between psychiatric symptoms and mental illness diagnoses. There is insufficient prevalence and incidence data to show all psychiatric symptoms exist or can be observed in patients. The diagnostic accuracy of symptoms to detect a disease cause is far from perfect. Assuming the symptoms of three mood disorders may present in patients, three diagnoses are not significantly associated with all psychiatric symptoms used to diagnose them. The diagnostic criteria of the three diagnoses have not been designed to guarantee significant associations between symptoms and diagnoses. Because statistical associations are important for making causal inferences, there may be a lack of causation between diagnoses and symptoms. Previous research has identified factors that lead to insignificant associations between diagnoses and symptoms, including biases due to data processing and a lack of epidemiological evidence to support the design of mental illness diagnostic criteria.

Efficacy of Non-Invasive Brain Stimulation for Refractory Obsessive-Compulsive Disorder: A Meta-Analysis of Randomized Controlled Trials

Obsessive-compulsive disorder (OCD) is a neuropsychiatric disorder, with 30−40% of OCD patients being unresponsive to adequate trials of anti-OCD drugs and cognitive behavior therapy. The aim of this paper is to investigate the efficacy of non-invasive brain stimulation (NIBS) on treating refractory OCD. With PubMed, Embase, PsycInfo, and Cochrane Library used on 15 February 2022, 24 randomized controlled trials involving 663 patients were included. According to this analysis, NIBS including repetitive transcranial magnetic stimulation (rTMS), theta-burst stimulation (TBS), and transcranial direct current stimulation (tDCS), had a moderate effect on the reduction of Yale-Brown Obsessive Compulsive Scale (Y-BOCS) scores (SMD = 0.54, 95% CI: 0.26−0.81; p < 0.01). In the subgroup analysis, rTMS seemed to produce a better therapeutic effect (SMD = 0.73, 95% CI: 0.38−1.08; p < 0.01). Moreover, excitatory (SMD = 1.13, 95% CI: 0.24−2.01; p = 0.01) and inhibitory (SMD = 0.81, 95% CI: 0.26−1.36; p < 0.01) stimulation of the dorsolateral prefrontal cortex (DLPFC) both alleviated OCD symptoms. In the secondary outcome of clinical response rates, NIBS treatment led to an increase in response rates (RR = 2.26, 95% CI: 1.57−3.25; p < 0.01).

An objective metric of individual health and aging for population surveys

BACKGROUND

We have previously developed and validated a biomarker-based metric of overall health status using Mahalanobis distance (DM) to measure how far from the norm of a reference population (RP) an individual's biomarker profile is. DM is not particularly sensitive to the choice of biomarkers; however, this makes comparison across studies difficult. Here we aimed to identify and validate a standard, optimized version of DM that would be highly stable across populations, while using fewer and more commonly measured biomarkers.

METHODS

Using three datasets (the Baltimore Longitudinal Study of Aging, Invecchiare in Chianti and the National Health and Nutrition Examination Survey), we selected the most stable sets of biomarkers in all three populations, notably when interchanging RPs across populations. We performed regression models, using a fourth dataset (the Women's Health and Aging Study), to compare the new DM sets to other well-known metrics [allostatic load (AL) and self-assessed health (SAH)] in their association with diverse health outcomes: mortality, frailty, cardiovascular disease (CVD), diabetes, and comorbidity number.

RESULTS

A nine- (DM9) and a seventeen-biomarker set (DM17) were identified as highly stable regardless of the chosen RP (e.g.: mean correlation among versions generated by interchanging RPs across dataset of r = 0.94 for both DM9 and DM17). In general, DM17 and DM9 were both competitive compared with AL and SAH in predicting aging correlates, with some exceptions for DM9. For example, DM9, DM17, AL, and SAH all predicted mortality to a similar extent (ranges of hazard ratios of 1.15-1.30, 1.21-1.36, 1.17-1.38, and 1.17-1.49, respectively). On the other hand, DM9 predicted CVD less well than DM17 (ranges of odds ratios of 0.97-1.08, 1.07-1.85, respectively).

CONCLUSIONS

The metrics we propose here are easy to measure with data that are already available in a wide array of panel, cohort, and clinical studies. The standardized versions here lose a small amount of predictive power compared to more complete versions, but are nonetheless competitive with existing metrics of overall health. DM17 performs slightly better than DM9 and should be preferred in most cases, but DM9 may still be used when a more limited number of biomarkers is available.

Judgement Bias in Miniature Donkeys: Conditioning Factors and Personality Links

Expectation-related bias may configure individuals' perception of their surrounding environment and of the elements present in it. This study aimed to determine the repercussions of environmental (weather elements) or subject-inherent factors (sex, age, or personality features) on judgment bias. A cognitive bias test was performed in eight Miniature jennies and four jacks. Test comprised habituation, training and testing phases during which subjects were trained on how to complete the test and scored based on their latency to approach an ambiguous stimulus. A questionnaire evaluating eleven personality features was parallelly completed by three caretakers, five operators and two care assistants to determine the links between personality features and judgment bias. Adjusted latencies did not significantly differ between sexes (Mann-Whitney test, p > 0.05). Although Miniature donkeys can discriminate positive/negative stimuli, inter-individual variability evidences were found. Such discrimination is evidenced by significant latency differences to approach positive/negative stimuli (33.7 ± 43.1 vs. 145.5 ± 53.1 s) (Mann-Whitney test, p < 0.05). Latencies significantly increased with patience, indicative of an expression of pessimism. Better understanding judgement bias mechanisms and implications may help optimize routine handling practices in the framework of animal welfare.

Interpretable Trials: Is Interpretability a Reason Why Clinical Trials Fail?

Background: There are clinical trials using composite measures, indices, or scales as proxy for independent variables or outcomes. Interpretability of derived measures may not be satisfying. Adopting indices of poor interpretability in clinical trials may lead to trial failure. This study aims to understand the impact of using indices of different interpretability in clinical trials.

Methods: The interpretability of indices was categorized as: fair-to-poor, good, and unknown. In the literature, frailty indices were considered fair to poor interpretability. Body mass index (BMI) was highly interpretable. The other indices were of unknown interpretability. The trials were searched at clinicaltrials.gov on October 2, 2018. The use of indices as conditions/diseases or other terms was searched. The trials were grouped as completed, terminated, active, and other status. We tabulated the frequencies of frailty, BMI, and other indices.

Results: There were 263,928 clinical trials found and 155,606 were completed or terminated. Among 2,115 trials adopting indices or composite measures as condition or disease, 244 adopted frailty and 487 used BMI without frailty indices. Significantly higher proportions of trials of unknown status used indices as conditions/diseases or other terms, compared to completed and terminated trials. The proportions of active trials using frailty indices were significantly higher than those of completed or terminated trials.

Discussion: Clinical trial databases can be used to understand why trials may fail. Based on the findings, we suspect that using indices of poor interpretability may be associated with trial failure. Interpretability has not been conceived as an essential criterion for outcomes or proxy measures in trials. We will continue verifying the findings in other databases or data sources and apply this research method to improve clinical trial design. To prevent patients from experiencing trials likely to fail, we suggest further examining the interpretability of the indices in trials.

Patient reported outcome measures in clinical trials should be initially analyzed as continuous outcomes for statistical significance and responder analyses should be reserved as secondary analyses

OBJECTIVE

To evaluate the power of responder analyses in a randomized controlled trial.

STUDY DESIGN AND SETTING

Simulations were based on the Chronic Kidney Disease Antidepressant Sertraline Trial (CAST), which compared sertraline to placebo for the treatment of depression in kidney disease. Baseline disease severity, placebo response, effect size, and the proportion of responders were varied across 72 scenarios. Power was assessed using a t-test for change scores, and the chi-square test for dichotomized outcomes of the minimal important difference (MID), improvement and remission in 10,000 datasets with a fixed sample size of 193.

RESULTS

The t-test had >80% power except for scenarios with the lowest sertraline effect size. The chi-square test using the MID had <7% power in all scenarios while improvement and remission of achieved >80% power only at higher effect sizes and/or when the proportion of responders was highest at 0.5. The chi-square test for improvement had marginal power increases compared to the t-test (4/72 scenarios = 5.6%) and that for remission did not outperform the t-test in any scenario.

CONCLUSIONS

The t-test outperforms the chi-square test for dichotomized outcomes regardless of baseline disease severity, placebo response, effect size and the proportion of responders to the intervention.

Simulation study to demonstrate biases created by diagnostic criteria of mental illnesses: major depressive episodes, dysthymia, and manic episodes

OBJECTIVES

Composite diagnostic criteria alone are likely to create and introduce biases into diagnoses that subsequently have poor relationships with input symptoms. This study aims to understand the relationships between the diagnoses and the input symptoms, as well as the magnitudes of biases created by diagnostic criteria and introduced into the diagnoses of mental illnesses with large disease burdens (major depressive episodes, dysthymic disorder, and manic episodes).

SETTINGS

General psychiatric care.

PARTICIPANTS

Without real-world data available to the public, 100 000 subjects were simulated and the input symptoms were assigned based on the assumed prevalence rates (0.05, 0.1, 0.3, 0.5 and 0.7) and correlations between symptoms (0, 0.1, 0.4, 0.7 and 0.9). The input symptoms were extracted from the diagnostic criteria. The diagnostic criteria were transformed into mathematical equations to demonstrate the sources of biases and convert the input symptoms into diagnoses.

PRIMARY AND SECONDARY OUTCOMES

The relationships between the input symptoms and diagnoses were interpreted using forward stepwise linear regressions. Biases due to data censoring or categorisation introduced into the intermediate variables, and the three diagnoses were measured.

RESULTS

The prevalence rates of the diagnoses were lower than those of the input symptoms and proportional to the assumed prevalence rates and the correlations between the input symptoms. Certain input or bias variables consistently explained the diagnoses better than the others. Except for 0 correlations and 0.7 prevalence rates of the input symptoms for the diagnosis of dysthymic disorder, the input symptoms could not fully explain the diagnoses.

CONCLUSIONS

There are biases created due to composite diagnostic criteria and introduced into the diagnoses. The design of the diagnostic criteria determines the prevalence of the diagnoses and the relationships between the input symptoms, the diagnoses, and the biases. The importance of the input symptoms has been distorted largely by the diagnostic criteria.

Defining the vulnerable patient with myeloma-a frailty position paper of the European Myeloma Network

As the treatment landscape continues to evolve towards the application of precision medicine in multiple myeloma (MM), there is a clear need to identify those patients who are at risk of not achieving the maximum benefit whilst exposed to the highest level of toxicity. This group of patients, defined as frail, is an unmet clinical need. However, how we define such a vulnerable group of patients with MM remains to be clarified. An integral aspect of this is to define the physiological age and capacity of patients with MM to deal with the burden of their disease and it's treatment. Such assessments may include not only functional and clinical assessments but also laboratory-based biomarkers of frailty, aging and senescent cellular burden. A need to develop, test and validate clinical screening scores before their adoption into clinical practice is mandated. This position paper from the European Myeloma Network aims to review what is known about defining frailty in MM, and how we can advance this knowledge for the design of clinical trials and ultimately how we deliver treatment in the clinic.

Using a dose-finding benchmark to quantify the loss incurred by dichotomization in Phase II dose-ranging studies

While there is recognition that more informative clinical endpoints can support better decision-making in clinical trials, it remains a common practice to categorize endpoints originally measured on a continuous scale. The primary motivation for this categorization (and most commonly dichotomization) is the simplicity of the analysis. There is, however, a long argument that this simplicity can come at a high cost. Specifically, larger sample sizes are needed to achieve the same level of accuracy when using a dichotomized outcome instead of the original continuous endpoint. The degree of "loss of information" has been studied in the contexts of parallel-group designs and two-stage Phase II trials. Limited attention, however, has been given to the quantification of the associated losses in dose-ranging trials. In this work, we propose an approach to estimate the associated losses in Phase II dose-ranging trials that is free of the actual dose-ranging design used and depends on the clinical setting only. The approach uses the notion of a nonparametric optimal benchmark for dose-finding trials, an evaluation tool that facilitates the assessment of a dose-finding design by providing an upper bound on its performance under a given scenario in terms of the probability of the target dose selection. After demonstrating how the benchmark can be applied to Phase II dose-ranging trials, we use it to quantify the dichotomization losses. Using parameters from real clinical trials in various therapeutic areas, it is found that the ratio of sample sizes needed to obtain the same precision using continuous and binary (dichotomized) endpoints varies between 70% and 75% under the majority of scenarios but can drop to 50% in some cases.

Subcallosal Cingulate Structural Connectivity Differs in Responders and Nonresponders to Electroconvulsive Therapy

BACKGROUND

Subcallosal cingulate (SCC) activity is associated with treatment response in major depressive disorder (MDD). Using electroconvulsive therapy (ECT) as a treatment model in this exploratory study, we addressed whether pretreatment SCC structural connectivity with corticolimbic-striatal circuitry relates to therapeutic outcome and whether these connectivity patterns change with treatment.

METHODS

Diffusion magnetic resonance imaging scans were acquired in 43 patients with MDD (mean [SD] age = 41 [13] years; men/women: 18/25) before and within 1 week of completing an ECT index series and in 31 healthy control subjects scanned twice (mean [SD] age = 38 [11] years; men/women: 17/18). Probabilistic tractography from subject-specific anatomically defined SCC seed regions to the ventral striatum (VS), anterior cingulate cortex (ACC), and bilateral medial prefrontal cortex (mPFC) was used to estimate structural connectivity in the target network.

RESULTS

SCC-mPFC connectivity was lower in responders (>50% symptom improvement) than nonresponders both before (p < .014) (difference 37%-96% left and right hemispheres) and after (p = .023) (difference 100% right hemisphere) treatment. SCC-mPFC connectivity in responders was also decreased compared with control subjects both at baseline (p = .012) and after ECT (p = .006), whereas nonresponders had SCC-right mPFC connectivity similar to that of control subjects. Subjects with MDD also showed decreased SCC-ACC connectivity compared with control subjects (baseline: p < .003, after ECT: p = .001), although SCC-ACC connectivity did not distinguish responders from nonresponders. Bilateral SCC-VS connectivity decreased (11%) with ECT (p = .021) regardless of treatment response.

CONCLUSIONS

While SCC-ACC connectivity may be a hallmark of MDD compared with control subjects, lower pretreatment SCC-mPFC connectivity in ECT responders (compared with nonresponders and control subjects) suggests that connectivity in this pathway may serve as a potential biomarker of therapeutic outcome and be relevant for treatment selection.

Composite diagnostic criteria are problematic for linking potentially distinct populations: the case of frailty

Composite diagnostic criteria are common in frailty research. We worry distinct populations may be linked to each other due to complicated criteria. We aim to investigate whether distinct populations might be considered similar based on frailty diagnostic criteria. The Functional Domains Model for frailty diagnosis included four domains: physical, nutritive, cognitive and sensory functioning. Health and Retirement Study participants with two or more deficiencies in the domains were diagnosed frail. The survival distributions were analyzed using discrete-time survival analysis. The distributions of the demographic characteristics and survival across the groups diagnosed with frailty were significantly different (p < 0.05). A deficiency in cognitive functioning was associated with the worst survival pattern compared with a deficiency in the other domains (adjusted p < 0.05). The associations of the domains with mortality were cumulative without interactions. Cognitive functioning had the largest effect size for mortality prediction (Odds ratios, OR = 2.37), larger than that of frailty status (OR = 1.92). The frailty diagnostic criteria may take distinct populations as equal and potentially assign irrelevant interventions to individuals without corresponding conditions. We think it necessary to review the adequacy of composite diagnostic criteria in frailty diagnosis.

Description of Criterion Validity of the Autism Spectrum Rating Scales 6-18 Parent Report: Initial Exploration in a Large Community Sample

Research regarding psychometric properties of autism spectrum disorder questionnaires is lacking. This study explored the criterion validity of the ASRS 6-18 parent report (ASRS-6-18-P) in a large, well-characterized, real-world clinical sample of 422 children ([Formula: see text] autism [AUT] n = 139; non-autism [NOT] n = 283) evaluated with the Autism Diagnostic Observation Schedule-2, a gold-standard measure. Significant mean differences were observed for DSM-5, social, and unusual behaviors. Total, DSM-5, social, and unusual behaviors demonstrated significant correlations with ADOS-2 comparison scores (modules 1-3), but not with module 4 (raw) scores. DSM-5 and Unusual Behaviors demonstrated significant but poor AUCs (0.60). Findings with/without covariates (IQ/age) were overall similar. Sensitivity and specificity could not be optimized. The suggested cutpoint (T-score = 60) demonstrated unacceptably high false positive rates (> 76.33%). While findings suggest limited diagnostic utility of the ASRS-6-18-P, the sample's complex psychiatric presentation and measurement error inherent in cutoff score application should be considered when generalizing results. Further research is recommended.

Subtherapeutic piperacillin concentrations in neurocritical patients

PURPOSE

Increased renal elimination is the leading cause for subtherapeutic concentrations of renally cleared antibiotics and it has been hypothesized that brain damaged patients in the intensive care unit (ICU) are particularly at risk. The objective of this study is to determine the prevalence of subtherapeutic piperacillin concentrations in neurocritical patients and to investigate if having a neurocritical diagnosis is a risk factor for this.

MATERIALS AND METHODS

Single center retrospective analysis of a prospective cohort study of adult ICU patients receiving continuous infusion piperacillin/tazobactam. Patients were categorized as either having a neurocritical diagnosis or not. An unbound piperacillin concentration > 4× the epidemiologic cut-off value (ECOFF) of Pseudomonas aeruginosa was selected as the PKPD target of choice. Multivariable logistic regression was performed to identify risk factors for subtherapeutic piperacillin concentrations.

RESULTS

356 patients had a measured creatinine clearance (mCrCl) and matched piperacillin concentration, 52 of which had a neurocritical diagnosis. Subtherapeutic piperacillin concentrations were reported significantly more frequent in neurocritical patients. In multivariate analysis, the only risk factor identified for subtherapeutic piperacillin concentration was an increasing mCrCl.

CONCLUSION

Subtherapeutic piperacillin concentrations are common in neurocritical patients yet having a neurocritical admission diagnosis was not identified as an independent risk factor.

A novel approach to determine two optimal cut-points of a continuous predictor with a U-shaped relationship to hazard ratio in survival data: simulation and application

BACKGROUND

In clinical and epidemiological researches, continuous predictors are often discretized into categorical variables for classification of patients. When the relationship between a continuous predictor and log relative hazards is U-shaped in survival data, there is a lack of a satisfying solution to find optimal cut-points to discretize the continuous predictor. In this study, we propose a novel approach named optimal equal-HR method to discretize a continuous variable that has a U-shaped relationship with log relative hazards in survival data.

METHODS

The main idea of the optimal equal-HR method is to find two optimal cut-points that have equal log relative hazard values and result in Cox models with minimum AIC value. An R package 'CutpointsOEHR' has been developed for easy implementation of the optimal equal-HR method. A Monte Carlo simulation study was carried out to investigate the performance of the optimal equal-HR method. In the simulation process, different censoring proportions, baseline hazard functions and asymmetry levels of U-shaped relationships were chosen. To compare the optimal equal-HR method with other common approaches, the predictive performance of Cox models with variables discretized by different cut-points was assessed.

RESULTS

Simulation results showed that in asymmetric U-shape scenarios the optimal equal-HR method had better performance than the median split method, the upper and lower quantiles method, and the minimum p-value method regarding discrimination ability and overall performance of Cox models. The optimal equal-HR method was applied to a real dataset of small cell lung cancer. The real data example demonstrated that the optimal equal-HR method could provide clinical meaningful cut-points and had good predictive performance in Cox models.

CONCLUSIONS

In general, the optimal equal-HR method is recommended to discretize a continuous predictor with right-censored outcomes if the predictor has an asymmetric U-shaped relationship with log relative hazards based on Cox regression models.

Handling ties in continuous outcomes for confounder adjustment with rank-ordered logit and its application to ordinal outcomes

The rank-ordered logit (rologit) model was recently introduced as a robust approach for analysing continuous outcomes, with the linear exposure effect estimated by scaling the rank-based log-odds estimate. Here we extend the application of the rologit model to continuous outcomes with ties and ordinal outcomes treated as imperfectly-observed continuous outcomes. By identifying the functional relationship between survival times and continuous outcomes, we explicitly establish the equivalence between the rologit and Cox models to justify the use of the Breslow, Efron and perturbation methods in the analysis of continuous outcomes with ties. Using simulation, we found all three methods perform well with few ties. Although an increasing extent of ties increased the bias of the log-odds and linear effect estimates and resulted in reduced power, which was somewhat worse when the model was mis-specified, the perturbation method maintained a type I error around 5%, while the Efron method became conservative with heavy ties but outperformed Breslow. In general, the perturbation method had the highest power, followed by the Efron and then the Breslow method. We applied our approach to three real-life datasets, demonstrating a seamless analytical workflow that uses stratification for confounder adjustment in studies of continuous and ordinal outcomes.

Spurious interaction as a result of categorization

Background

It is common in applied epidemiological and clinical research to convert continuous variables into categorical variables by grouping values into categories. Such categorized variables are then often used as exposure variables in some regression model. There are numerous statistical arguments why this practice should be avoided, and in this paper we present yet another such argument.

Methods

We show that categorization may lead to spurious interaction in multiple regression models. We give precise analytical expressions for when this may happen in the linear regression model with normally distributed exposure variables, and we show by simulations that the analytical results are valid also for other distributions. Further, we give an interpretation of the results in terms of a measurement error problem.

Results

We show that, in the case of a linear model with two normally distributed exposure variables, both categorized at the same cut point, a spurious interaction will be induced unless the two variables are categorized at the median or they are uncorrelated. In simulations with exposure variables following other distributions, we confirm this general effect of categorization, but we also show that the effect of the choice of cut point varies over different distributions.

Conclusion

Categorization of continuous exposure variables leads to a number of problems, among them spurious interaction effects. Hence, this practice should be avoided and other methods should be considered.

integRATE: a desirability-based data integration framework for the prioritization of candidate genes across heterogeneous omics and its application to preterm birth

BACKGROUND

The integration of high-quality, genome-wide analyses offers a robust approach to elucidating genetic factors involved in complex human diseases. Even though several methods exist to integrate heterogeneous omics data, most biologists still manually select candidate genes by examining the intersection of lists of candidates stemming from analyses of different types of omics data that have been generated by imposing hard (strict) thresholds on quantitative variables, such as P-values and fold changes, increasing the chance of missing potentially important candidates.

METHODS

To better facilitate the unbiased integration of heterogeneous omics data collected from diverse platforms and samples, we propose a desirability function framework for identifying candidate genes with strong evidence across data types as targets for follow-up functional analysis. Our approach is targeted towards disease systems with sparse, heterogeneous omics data, so we tested it on one such pathology: spontaneous preterm birth (sPTB).

RESULTS

We developed the software integRATE, which uses desirability functions to rank genes both within and across studies, identifying well-supported candidate genes according to the cumulative weight of biological evidence rather than based on imposition of hard thresholds of key variables. Integrating 10 sPTB omics studies identified both genes in pathways previously suspected to be involved in sPTB as well as novel genes never before linked to this syndrome. integRATE is available as an R package on GitHub ( https://github.com/haleyeidem/integRATE ).

CONCLUSIONS

Desirability-based data integration is a solution most applicable in biological research areas where omics data is especially heterogeneous and sparse, allowing for the prioritization of candidate genes that can be used to inform more targeted downstream functional analyses.

Variations in Hippocampal White Matter Diffusivity Differentiate Response to Electroconvulsive Therapy in Major Depression

BACKGROUND

Electroconvulsive therapy (ECT) is an effective treatment for severe depression and is shown to increase hippocampal volume and modulate hippocampal functional connectivity. Whether variations in hippocampal structural connectivity occur with ECT and relate to clinical response is unknown.

METHODS

Patients with major depression (n = 36, 20 women, age 41.49 ± 13.57 years) underwent diffusion magnetic resonance imaging at baseline and after ECT. Control subjects (n = 32, 17 women, age 39.34 ± 12.27 years) underwent scanning twice. Functionally defined seeds in the left and right anterior hippocampus and probabilistic tractography were used to extract tract volume and diffusion metrics (fractional anisotropy and axial, radial, and mean diffusivity). Statistical analyses determined effects of ECT and time-by-response group interactions (>50% change in symptoms before and after ECT defined response). Differences between baseline measures across diagnostic groups and in association with treatment outcome were also examined.

RESULTS

Significant effects of ECT (all p < .01) and time-by-response group interactions (all p < .04) were observed for axial, radial, and mean diffusivity for right, but not left, hippocampal pathways. Follow-up analyses showed that ECT-related changes occurred in responders only (all p < .01) as well as in relation to change in mood examined continuously (all p < .004). Baseline measures did not relate to symptom change or differ between patients and control subjects. All measures remained stable across time in control subjects. No significant effects were observed for fractional anisotropy and volume.

CONCLUSIONS

Structural connectivity of hippocampal neural circuits changed with ECT and distinguished treatment responders. The findings suggested neurotrophic, glial, or inflammatory response mechanisms affecting axonal integrity.

Index or illusion: The case of frailty indices in the Health and Retirement Study

Introduction

Frailty is a geriatric syndrome that has been defined differently with various indices. Without a uniform definition, it remains unclear how to interpret and compare different frailty indices (FIs). With the advances in index mining, we find it necessary to review the implicit assumptions about the creation of FIs. We are concerned the processing of frailty data may introduce measurement error and bias. We aim to review the assumptions, interpretability and predictive power of FIs regarding mortality.

Methods

Three FIs, the Functional Domains Model proposed by Strawbridge et al. (1998), the Burden Model by Rockwood et al. (2007) and the Biologic Syndrome Model by Fried et al. (2004), were directly compared using the data from the Health and Retirement Study (HRS), a longitudinal study since 1996 mainly following up Americans aged 50 years and over. The FIs were reproduced according to Cigolle et al. (2009) and interpreted with their input variables through forward-stepwise regression. Biases were the residuals of the FIs that could not be explained by own input variables. Any four of the input variables were used to create alternative indices. Discrete-time survival analysis was conducted to compare the predictive power of FIs, input variables and alternative indices on mortality.

Results

We found frailty a syndrome not unique to the elderly. The FIs were produced with different degrees of bias. The FIs could not be fully interpreted with the theory-based input variables. The bias induced by the Biological Syndrome Model better predicted mortality than frailty status. A complicated FI, the Burden Model, could be simplified. The input variables better predicted mortality than the FIs. The continuous FIs predicted mortality better than the frailty statuses. At least 6865 alternative indices better predicted mortality than the FIs.

Conclusion

FIs have been used as outcome in clinical trials and need to be reviewed for adequacy based on our findings. The three FIs are not closely linked to the theories because of bias introduced by data manipulation and excessive numbers of input variables. We are developing new algorithms to develop and validate innovative indices.

Four simple ways to increase power without increasing the sample size

Underpowered experiments have three problems: true effects are harder to detect, the true effects that are detected tend to have inflated effect sizes and as power decreases so does the probability that a statistically significant result represents a true effect. Many biology experiments are underpowered and recent calls to change the traditional 0.05 significance threshold to a more stringent value of 0.005 will further reduce the power of the average experiment. Increasing power by increasing the sample size is often the only option considered, but more samples increases costs, makes the experiment harder to conduct and is contrary to the 3Rs principles for animal research. We show how the design of an experiment and some analytical decisions can have a surprisingly large effect on power.

A stratification approach using logit-based models for confounder adjustment in the study of continuous outcomes

The control of confounding is an area of extensive epidemiological research, especially in the field of causal inference for observational studies. Matched cohort and case-control study designs are commonly implemented to control for confounding effects without specifying the functional form of the relationship between the outcome and confounders. This paper extends the commonly used regression models in matched designs for binary and survival outcomes (i.e. conditional logistic and stratified Cox proportional hazards) to studies of continuous outcomes through a novel interpretation and application of logit-based regression models from the econometrics and marketing research literature. We compare the performance of the maximum likelihood estimators using simulated data and propose a heuristic argument for obtaining the residuals for model diagnostics. We illustrate our proposed approach with two real data applications. Our simulation studies demonstrate that our stratification approach is robust to model misspecification and that the distribution of the estimated residuals provides a useful diagnostic when the strata are of moderate size. In our applications to real data, we demonstrate that parity and menopausal status are associated with percent mammographic density, and that the mean level and variability of inpatient blood glucose readings vary between medical and surgical wards within a national tertiary hospital. Our work highlights how the same class of regression models, available in most statistical software, can be used to adjust for confounding in the study of binary, time-to-event and continuous outcomes.

The risks of biomarker-based epidemiology: Associations of circulating calcium levels with age, mortality, and frailty vary substantially across populations

Recent studies have shown contradictory associations between calcium levels and health outcomes. We suspected these conflicting results were the consequence of more general issues with how biomarkers are analyzed in epidemiological studies, particularly in the context of aging. To demonstrate the risks of typical analyses, we used three longitudinal aging cohort studies and their demographic subsets to analyze how calcium levels change with age and predict risk of mortality and frailty. We show that calcium levels either increase or decrease with age depending on the population, and positively or negatively predict frailty depending on the population and analysis; both age and frailty results showed substantial heterogeneity. Mortality analyses revealed few significant associations but were likely underpowered. Variation in population composition (demographics, diseases, diet, etc.) leads to contradictory findings in the literature for calcium and likely for other biomarkers. Epidemiological studies of biomarkers are particularly sensitive to population composition both because biomarkers generally have non-linear and often non-monotonic relationships with other key variables, notably age and health outcomes, and because there is strong interdependence among biomarkers, which are integrated into complex regulatory networks. Consequently, most biomarkers have multiple physiological roles and are implicated in multiple pathologies. We argue that epidemiological studies of aging using biomarkers must account for these factors, and suggest methods to do this.

Ranking, selecting, and prioritising genes with desirability functions

In functional genomics experiments, researchers often select genes to follow-up or validate from a long list of differentially expressed genes. Typically, sharp thresholds are used to bin genes into groups such as significant/non-significant or fold change above/below a cut-off value, and ad hoc criteria are also used such as favouring well-known genes. Binning, however, is inefficient and does not take the uncertainty of the measurements into account. Furthermore, p-values, fold-changes, and other outcomes are treated as equally important, and relevant genes may be overlooked with such an approach. Desirability functions are proposed as a way to integrate multiple selection criteria for ranking, selecting, and prioritising genes. These functions map any variable to a continuous 0–1 scale, where one is maximally desirable and zero is unacceptable. Multiple selection criteria are then combined to provide an overall desirability that is used to rank genes. In addition to p-values and fold-changes, further experimental results and information contained in databases can be easily included as criteria. The approach is demonstrated with a breast cancer microarray data set. The functions and an example data set can be found in the desiR package on CRAN (https://cran.r-project.org/web/packages/desiR/) and the development version is available on GitHub (https://github.com/stanlazic/desiR).