Study design features that improve effect sizes in cross-sectional and longitudinal brain-wide association studies

Brain-wide association studies (BWAS) are a fundamental tool in discovering brain-behavior associations. Several recent studies showed that thousands of study participants are required to improve the replicability of BWAS because actual effect sizes are much smaller than those reported in smaller studies. Here, we perform analyses and meta-analyses of a robust effect size index (RESI) using 63 longitudinal and cross-sectional magnetic resonance imaging studies from the Lifespan Brain Chart Consortium (77,695 total scans) to demonstrate that optimizing study design is critical for improving standardized effect sizes and replicability in BWAS. A meta-analysis of brain volume associations with age indicates that BWAS with larger covariate variance have larger effect size estimates and that the longitudinal studies we examined have systematically larger standardized effect sizes than cross-sectional studies. We propose a cross-sectional RESI to adjust for the systematic difference in effect sizes between cross-sectional and longitudinal studies that allows investigators to quantify the benefit of conducting their study longitudinally. Analyzing age effects on global and regional brain measures from the United Kingdom Biobank and the Alzheimer's Disease Neuroimaging Initiative, we show that modifying longitudinal study design through sampling schemes to increase between-subject variability and adding a single additional longitudinal measurement per subject can improve effect sizes. However, evaluating these longitudinal sampling schemes on cognitive, psychopathology, and demographic associations with structural and functional brain outcome measures in the Adolescent Brain and Cognitive Development dataset shows that commonly used longitudinal models can, counterintuitively, reduce effect sizes. We demonstrate that the benefit of conducting longitudinal studies depends on the strengths of the between- and within-subject associations of the brain and non-brain measures. Explicitly modeling between- and within-subject effects avoids conflating the effects and allows optimizing effect sizes for them separately. These findings underscore the importance of considering study design features to improve the replicability of BWAS.

Genetic Variants Associated With Systolic Blood Pressure in Children and Adolescents

Background Genetics, along with lifestyle and behavioral characteristics, play an important role in hypertension in adults. Our aim was to identify genetic variants associated with blood pressure in childhood and adolescence. Methods and Results We conducted a candidate single-nucleotide polymorphism (SNP) analysis and genome-wide association study among 9778 participants aged <18 years in BioVU, the Vanderbilt University Medical Center biobank. The outcome was childhood blood pressure percentile from age 0 to 18 years. For the candidate SNP analysis, a total of 457 previously identified SNPs were examined. Linear regression was used to test the association between genetic variants and median systolic blood pressure (SBP) percentile. Adjusted models included median age, self-reported sex, race, the first 4 principal components of ancestry, and median body mass index Z score. Analyses were conducted in the overall cohort and stratified by age group. A polygenic risk score was calculated for each participant, and the association between polygenic risk score and median SBP percentile in childhood was examined using linear regression. In the overall candidate SNP analysis, 2 SNPs reached significance: rs1018148 (FBN1; P=1.0×10-4) and rs11105354 (ATP2B1; P=1.4×10-4). In the postpuberty age group, 1 SNP reached significance: rs1018148 (FBN1; P=2.2×10-5). In the genome-wide association study of all participants, no SNPs reached genome-wide significance. Higher polygenic risk score was associated with higher SBP percentile (β, 0.35 [95% CI, 0.10-0.60)], and there was a significant interaction with age (P for interaction<0.01). Conclusions These findings suggest that genetic variants play an important role in SBP in childhood and adolescence and provide evidence for age-specific genetic associations with SBP.

Perioperative oxidative stress predicts subsequent pain-related outcomes in the 6 months after total knee arthroplasty

ABSTRACT

Total knee arthroplasty (TKA) is effective for pain reduction in most patients, but 15% or more report unsatisfactory long-term pain outcomes. We tested whether oxidative stress (OS) related to extended tourniquet application during TKA and subsequent ischemic reperfusion (IR) contributed to adverse post-TKA pain outcomes. Blood samples were obtained in 91 patients with osteoarthritis (63% female) undergoing TKA before tourniquet placement (T1), 45 minutes after tourniquet inflation (T2), and 15 minutes after tourniquet removal (T3). Plasma levels of F 2 -isoprostanes and isofurans, the most specific measures of in vivo OS, were quantified. Pain intensity and function were assessed at baseline and again at 6 weeks and 6 months after TKA. Results indicated that higher Combined OS (F 2 -isoprostanes + isofurans/2) at T1 baseline and larger increases in Combined OS from T1 to T2 were associated with higher baseline-corrected past 24-hour worst and average pain intensity (numeric rating scale) and higher past week McGill Pain Questionnaire-2 total scores at 6-month follow-up ( P 's < 0.05). Increases in Combined OS from T1 to T3, which should most directly capture OS and IR injury related to tourniquet use, were not associated with short-term or long-term post-TKA pain outcomes. Longer ischemia duration was unexpectedly associated with lower baseline-corrected pain intensity at 6-month follow-up. Combined OS was not linked to functional outcomes at either follow-up. Elevated perioperative OS seems to exert small but significant adverse effects on long-term post-TKA pain outcomes, although this OS seems unrelated to IR injury associated with extended tourniquet use.

Preoperative Predictors of Complex Regional Pain Syndrome Outcomes in the 6 Months Following Total Knee Arthroplasty

This prospective observational study evaluated preoperative predictors of complex regional pain syndrome (CRPS) outcomes in the 6 months following total knee arthroplasty (TKA). Participants were n = 110 osteoarthritis patients (64.5% female) undergoing unilateral TKA with no prior CRPS history. Domains of negative affect (depression, anxiety, catastrophizing), pain (intensity, widespread pain, temporal summation of pain [TSP]), pain interference, sleep disturbance, and pro-inflammatory status (tumor necrosis factor-alpha [TNF-a]) were assessed preoperatively. CRPS outcomes at 6-week and 6-month follow-up included the continuous CRPS Severity Score (CSS) and dichotomous CRPS diagnoses (2012 IASP criteria). At 6 months, 12.7% of participants met CRPS criteria, exhibiting a "warm CRPS" phenotype. Six-week CSS scores were predicted by greater preoperative depression, anxiety, catastrophizing, TSP, pain intensity, sleep disturbance, and TNF-a (P's < .05). Provisional CRPS diagnosis at 6 weeks was predicted by higher preoperative TSP, sleep disturbance, and TNF-a (P's < .05). CSS scores at 6 months were predicted by more widespread and intense preoperative pain, and higher preoperative TSP, pain interference, and TNF-a (P's < .01). CRPS diagnosis at 6 months was predicted only by more widespread and intense pain preoperatively (P's < .05). Risk for CRPS following TKA appears to involve preoperative central sensitization and inflammatory mechanisms. Preoperative negative affect is unlikely to directly influence long-term CRPS risk. PERSPECTIVE: This article identifies preoperative predictors of CRPS features at 6 months following total knee arthroplasty, including more widespread pain and higher pain intensity, temporal summation of pain, pain interference, and tumor necrosis factor-alpha levels. Findings suggest the importance of central sensitization and inflammatory mechanisms in CRPS risk following tissue trauma.

Oxidative stress is associated with characteristic features of the dysfunctional chronic pain phenotype

ABSTRACT

The dysfunctional chronic pain (Dysfunctional CP) phenotype is an empirically identifiable CP subtype with unclear pathophysiological mechanisms that cuts across specific medical CP diagnoses. This study tested whether the multidimensional pain and psychosocial features that characterize the dysfunctional CP phenotype are associated broadly with elevated oxidative stress (OS). Measures of pain intensity, bodily extent of pain, catastrophizing cognitions, depression, anxiety, sleep disturbance, pain interference, and function were completed by 84 patients with chronic osteoarthritis before undergoing total knee arthroplasty. Blood samples were obtained at the initiation of surgery before incision or tourniquet placement. Plasma levels of F2-isoprostanes and isofurans, the most highly specific measures of in vivo OS, were quantified using gas chromatography/negative ion chemical ionization mass spectrometry. The results indicated that controlling for differences in age, sex, and body mass index, higher overall OS (mean of isoprostanes and isofurans) was associated with significantly (P < 0.05) greater pain intensity, more widespread pain, greater depressive symptoms and pain catastrophizing, higher pain interference, and lower function. OS measures were not significantly associated with sleep disturbance or anxiety levels (P >0.10). The results build on prior case-control findings suggesting that presence of a CP diagnosis is associated with elevated OS, highlighting that it may specifically be individuals displaying characteristics of the dysfunctional CP phenotype who are characterized by elevated OS. Clinical implications of these findings remain to be determined.

Fitting marginal models in small samples: A simulation study of marginalized multilevel models and generalized estimating equations

In correlated data settings, analysts typically choose between fitting conditional and marginal models, whose parameters come with distinct interpretations, and as such the choice between the two should be made on scientific grounds. For settings where interest lies in marginal-or population-averaged-parameters, the question of how best to estimate those parameters is a statistical one, and analysts have at their disposal two distinct modeling frameworks: generalized estimating equations (GEE) and marginalized multilevel models (MMMs). The two have been contrasted theoretically and in large sample settings, but asymptotic theory provides no guarantees in the small sample settings that are commonplace. In a comprehensive series of simulation studies, we shed light on the relative performance of GEE and MMMs in small-sample settings to help guide analysis decisions in practice. We find that both GEE and MMMs exhibit similar small-sample bias when the correct correlation structure is adopted (ie, when the random effects distribution is correctly specified or moderately misspecified)-but MMMs can be sensitive to misspecification of the correlation structure. When there are a small number of clusters, MMMs only slightly underestimate standard errors (SEs) for within-cluster associations but can severely underestimate SEs for between-cluster associations. By contrast, while GEE severely underestimates SEs, the Mancl and DeRouen correction provides approximately valid inference.

Outcome-dependent sampling in cluster-correlated data settings with application to hospital profiling

Hospital readmission is a key marker of quality of healthcare and an important policy measure, used by the Centers for Medicare and Medicaid Services to determine, in part, reimbursement rates. Currently, analyses of readmissions are based on a logistic-normal generalized linear mixed model that permits estimation of hospital-specific measures while adjusting for case mix differences. Recent moves to identify and address healthcare disparities call for expanding case mix adjustment to include measures of socio-economic status while minimizing additional burden to hospitals associated with collecting data on such measures. Towards resolving this dilemma, we propose that detailed socio-economic data be collected on a subsample of patients via an outcome-dependent sampling scheme, specifically the cluster-stratified case-control design. Estimation and inference, for both the fixed and the random-effects components, are performed via pseudo-maximum-likelihood wherein inverse probability weights are incorporated in the usual integrated likelihood to account for the design. In comprehensive simulations, cluster-stratified case-control sampling proves to be an efficient design whenever interest lies in fixed or random effects of a generalized linear mixed model and covariates are unobserved or expensive to collect. The methods are motivated by and illustrated with an analysis of N = 889661 Medicare beneficiaries hospitalized between 2011 and 2013 with congestive heart failure at one of K = 3116 hospitals. Results highlight that the framework proposed provides a means of mitigating disparities in terms of which hospitals are indicated as being poor performers, relative to a naive analysis that fails to adjust for missing case mix variables.

Development of a multivariable model to predict vulnerability in older American patients hospitalised with cardiovascular disease

OBJECTIVES

To identify vulnerable cardiovascular patients in the hospital using a self-reported function-based screening tool.

PARTICIPANTS

Prospective observational cohort study of 445 individuals aged ≥ 65 years admitted to a university medical centre hospital within the USA with acute coronary syndrome and/or decompensated heart failure.

METHODS

Participants completed an inperson interview during hospitalisation, which included vulnerable functional status using the Vulnerable Elders Survey (VES-13), sociodemographic, healthcare utilisation practices and clinical patient-specific measures. A multivariable proportional odds logistic regression model examined associations between VES-13 and prior healthcare utilisation, as well as other coincident medical and psychosocial risk factors for poor outcomes in cardiovascular disease.

RESULTS

Vulnerability was highly prevalent (54%) and associated with a higher number of clinic visits, emergency room visits and hospitalisations (all p<0.001). A multivariable analysis demonstrating a 1-point increase in VES-13 (vulnerability) was independently associated with being female (OR 1.55, p=0.030), diagnosis of heart failure (OR 3.11, p<0.001), prior hospitalisations (OR 1.30, p<0.001), low social support (OR 1.42, p=0.007) and depression (p<0.001). A lower VES-13 score (lower vulnerability) was associated with increased health literacy (OR 0.70, p=0.002).

CONCLUSIONS

Vulnerability to functional decline is highly prevalent in hospitalised older cardiovascular patients and was associated with patient risk factors for adverse outcomes and an increased use of healthcare services.

Medication Nonadherence Before Hospitalization for Acute Cardiac Events

Medication nonadherence increases the risk of hospitalization and poor outcomes, particularly among patients with cardiovascular disease. The purpose of this study was to examine characteristics associated with medication nonadherence among adults hospitalized for cardiovascular disease. Patients in the Vanderbilt Inpatient Cohort Study who were admitted for acute coronary syndrome or heart failure completed validated assessments of self-reported medication adherence (the Adherence to Refills and Medications Scale), demographic characteristics, health literacy, numeracy, social support, depressive symptoms, and health competence. We modeled the independent predictors of nonadherence before hospitalization, standardizing estimated effects by each predictor's interquartile range. Among 1,967 patients studied, 70.7% indicated at least some degree of medication nonadherence leading up to their hospitalization. Adherence was significantly lower among patients with lower health literacy (0.18-point change in adherence score per interquartile range change in health literacy), lower numeracy (0.28), lower health competence (0.30), and more depressive symptoms (0.52) and those of younger age, of non-White race, of male gender, or with less social support. Medication nonadherence in the period before hospitalization is more prevalent among patients with lower health literacy, numeracy, or other intervenable psychosocial factors. Addressing these factors in a coordinated care model may reduce hospitalization rates.

Electronic health record design and implementation for pharmacogenomics: a local perspective

PURPOSE

The design of electronic health records to translate genomic medicine into clinical care is crucial to successful introduction of new genomic services, yet there are few published guides to implementation.

METHODS

The design, implemented features, and evolution of a locally developed electronic health record that supports a large pharmacogenomics program at a tertiary-care academic medical center was tracked over a 4-year development period.

RESULTS

Developers and program staff created electronic health record mechanisms for ordering a pharmacogenomics panel in advance of clinical need (preemptive genotyping) and in response to a specific drug indication. Genetic data from panel-based genotyping were sequestered from the electronic health record until drug-gene interactions met evidentiary standards and deemed clinically actionable. A service to translate genotype to predicted drug-response phenotype populated a summary of drug-gene interactions, triggered inpatient and outpatient clinical decision support, updated laboratory records, and created gene results within online personal health records.

CONCLUSION

The design of a locally developed electronic health record supporting pharmacogenomics has generalizable utility. The challenge of representing genomic data in a comprehensible and clinically actionable format is discussed along with reflection on the scalability of the model to larger sets of genomic data.

Operational implementation of prospective genotyping for personalized medicine: the design of the Vanderbilt PREDICT project

The promise of "personalized medicine" guided by an understanding of each individual's genome has been fostered by increasingly powerful and economical methods to acquire clinically relevant information. We describe the operational implementation of prospective genotyping linked to an advanced clinical decision-support system to guide individualized health care in a large academic health center. This approach to personalized medicine entails engagement between patient and health-care provider, identification of relevant genetic variations for implementation, assay reliability, point-of-care decision support, and necessary institutional investments. In one year, approximately 3,000 patients, most of whom were scheduled for cardiac catheterization, were genotyped on a multiplexed platform that included genotyping for CYP2C19 variants that modulate response to the widely used antiplatelet drug clopidogrel. These data are deposited into the electronic medical record (EMR), and point-of-care decision support is deployed when clopidogrel is prescribed for those with variant genotypes. The establishment of programs such as this is a first step toward implementing and evaluating strategies for personalized medicine.

A two-stage strategy to accommodate general patterns of confounding in the design of observational studies

Accommodating general patterns of confounding in sample size/power calculations for observational studies is extremely challenging, both technically and scientifically. While employing previously implemented sample size/power tools is appealing, they typically ignore important aspects of the design/data structure. In this paper, we show that sample size/power calculations that ignore confounding can be much more unreliable than is conventionally thought; using real data from the US state of North Carolina, naive calculations yield sample size estimates that are half those obtained when confounding is appropriately acknowledged. Unfortunately, eliciting realistic design parameters for confounding mechanisms is difficult. To overcome this, we propose a novel two-stage strategy for observational study design that can accommodate arbitrary patterns of confounding. At the first stage, researchers establish bounds for power that facilitate the decision of whether or not to initiate the study. At the second stage, internal pilot data are used to estimate key scientific inputs that can be used to obtain realistic sample size/power. Our results indicate that the strategy is effective at replicating gold standard calculations based on knowing the true confounding mechanism. Finally, we show that consideration of the nature of confounding is a crucial aspect of the elicitation process; depending on whether the confounder is positively or negatively associated with the exposure of interest and outcome, naive power calculations can either under or overestimate the required sample size. Throughout, simulation is advocated as the only general means to obtain realistic estimates of statistical power; we describe, and provide in an R package, a simple algorithm for estimating power for a case-control study.

Identifying unpredicted drug benefit through query of patient experiential knowledge: a proof of concept web-based system

Information necessary to recognize unexpected drug efficacy is not routinely collected. Once a drug is approved, opportunities for understanding these phenomena are usually lost within clinical care. We propose that patients are willing to provide a wide range of experiential knowledge about the effects of therapies that is seldom solicited. Experience with various drug therapies might be solicited directly from patients in both structured and unstructured formats. Although the signal to noise ratio is expected to be low, these data, if organized in a constructive manner, could provide a useful hypothesis generation resource for areas of further pharmacologic inquiry. A pilot study was conducted for 18 months; 1,065 individuals using the MyHealthAtVanderbilt.com patient portal clicked on a research link to find more information about the study; 375 completed the survey (response rate of 37%). Of those, 218 patients reported that they were currently taking at least one prescription. Statistical methods applied detected known associations between drugs and their intended effects. This validated the type and frequency of effects being reported by patients and provided evidence for the potential for using patient-supplied information to generate hypotheses related to unexpected positive benefits associated with medications. Improved data filtering and mining methods will be needed to expand this concept.

What can paper-based clinical information systems tell us about the design of computerized clinical information systems (CIS) in the ICU?

BACKGROUND

Screen designs in computerized clinical information systems (CIS) have been modeled on their paper predecessors. However, limited understanding about how paper forms support clinical work means that we risk repeating old mistakes and creating new opportunities for error and inefficiency as illustrated by problems associated with computerized provider order entry systems.

PURPOSE

This study was designed to elucidate principles underlying a successful ICU paper-based CIS. The research was guided by two exploratory hypotheses: (1) paper-based artefacts (charts, notes, equipment, order forms) are used differently by nurses, doctors and other healthcare professionals in different (formal and informal) conversation contexts and (2) different artefacts support different decision processes that are distributed across role-based conversations.

METHOD

All conversations undertaken at the bedsides of five patients were recorded with any supporting artefacts for five days per patient. Data was coded according to conversational role-holders, clinical decision process, conversational context and artefacts. 2133 data points were analyzed using Poisson logistic regression analyses.

RESULTS

Results show significant interactions between artefacts used during different professional conversations in different contexts (chi(2)((df=16))=55.8, p<0.0001). The interaction between artefacts used during different professional conversations for different clinical decision processes was not statistically significant although all two-way interactions were statistically significant.

CONCLUSIONS

Paper-based CIS have evolved to support complex interdisciplinary decision processes. The translation of two design principles - support interdisciplinary perspectives and integrate decision processes - from paper to computerized CIS may minimize the risks associated with computerization.

Adjustment for selection bias in observational studies with application to the analysis of autopsy data

BACKGROUND

The interpretation of neuropathological studies of dementia and Alzheimer's disease is complicated by potential selection mechanisms that can drive whether or not a study participant is observed to undergo autopsy. Notwithstanding this, there appears to have been little emphasis placed on potential selection bias in published reports from population-based neuropathological studies of dementia.

METHODS

We provide an overview of methodological issues relating to the identification of and adjustment for selection bias. When information is available on factors that govern selection, inverse-probability weighting provides an analytic approach to adjust for selection bias. The weights help alleviate bias by serving to bridge differences between the population from which the observed data may be viewed as a representative sample and the target population, identified as being of scientific interest.

RESULTS

We illustrate the methods with data obtained from the Adult Changes in Thought study. Adjustment for potential selection bias yields substantially strengthened association between neuropathological measurements and risk of dementia.

CONCLUSIONS

Armed with analytic techniques to adjust for selection bias and to ensure generalizability of results from population-based neuropathological studies, researchers should consider incorporating information related to selection into their data collection schemes.

Exposure and measurement contributions to estimates of acute air pollution effects

Air pollution health effect studies are intended to estimate the effect of a pollutant on a health outcome. The definition of this effect depends upon the study design, disease model parameterization, and the type of analysis. Further limitations are imposed by the nature of exposure and our ability to measure it. We define a plausible exposure model for air pollutants that are relatively nonreactive and discuss how exposure varies. We discuss plausible disease models and show how their parameterizations are affected by different exposure partitions and by different study designs. We then discuss a measurement model conditional on ambient concentrations and incorporate this into the disease model. We use simulation studies to show the impact of a range of exposure model assumptions on estimation of the health effect in the ecologic time series design. This design only uses information from the time-varying ambient source exposure. When ambient and nonambient sources are independent, exposure variation due to nonambient source exposures behaves like Berkson measurement error and does not bias the effect estimates. Variation in the population attenuation of ambient concentrations over time does bias the estimates with the bias being either positive or negative depending upon the association of this parameter with ambient pollution. It is not realistic to substitute measured average personal exposures into time series studies because so much of the variation in personal exposures comes from nonambient sources that do not contribute information in the time series design. We conclude that general statements about the implications of measurement error need to be conditioned on the health effect study design and the health effect parameter to be estimated.

Identifying UMLS concepts from ECG Impressions using KnowledgeMap

Electrocardiogram (ECG) impressions represent a wealth of medical information for potential decision support and drug-effect discovery. Much of this information is inaccessible to automated methods in the free-text portion of the ECG report. We studied the application of the KnowledgeMap concept identifier (KMCI) to map Unified Medical Language System (UMLS) concepts from ECG impressions. ECGs were processed by KMCI and the results scored for accuracy by multiple raters. Reviewers also recorded unidentified concepts through the scoring interface. Overall, KMCI correctly identified 1059 out of 1171 concepts for a recall of 0.90. Precision, indicating the proportion of ECG concepts correctly identified, was 0.94. KMCI was particularly effective at identifying ECG rhythms (330/333), perfusion changes (65/66), and noncardiac medical concepts (11/11). In conclusion, KMCI is an effective method for mapping ECG impressions to UMLS concepts.