Identifying genes associated with disease outcomes using joint sparse canonical correlation analysis-An application in renal clear cell carcinoma.

Somatic changes like copy number aberrations (CNAs) and epigenetic alterations like methylation have pivotal effects on disease outcomes and prognosis in cancer, by regulating gene expressions, that drive critical biological processes. To identify potential biomarkers and molecular targets and understand how they impact disease outcomes, it is important to identify key groups of CNAs, the associated methylation, and the gene expressions they impact, through a joint integrative analysis. Here, we propose a novel analysis pipeline, the joint sparse canonical correlation analysis (jsCCA), an extension of sCCA, to effectively identify an ensemble of CNAs, methylation sites and gene (expression) components in the context of disease endpoints, especially tumor characteristics. Our approach detects potentially orthogonal gene components that are highly correlated with sets of methylation sites which in turn are correlated with sets of CNA sites. It then identifies the genes within these components that are associated with the outcome. Further, we aggregate the effect of each gene expression set on tumor stage by constructing "gene component scores" and test its interaction with traditional risk factors. Analyzing clinical and genomic data on 515 renal clear cell carcinoma (ccRCC) patients from the TCGA-KIRC, we found eight gene components to be associated with methylation sites, regulated by groups of proximally located CNA sites. Association analysis with tumor stage at diagnosis identified a novel association of expression of ASAH1 gene trans-regulated by methylation of several genes including SIX5 and by CNAs in the 10q25 region including TCF7L2. Further analysis to quantify the overall effect of gene sets on tumor stage, revealed that two of the eight gene components have significant interaction with smoking in relation to tumor stage. These gene components represent distinct biological functions including immune function, inflammatory responses, and hypoxia-regulated pathways. Our findings suggest that jsCCA analysis can identify interpretable and important genes, regulatory structures, and clinically consequential pathways. Such methods are warranted for comprehensive analysis of multimodal data especially in cancer genomics.

Experiences and lessons learned from community-engaged recruitment for the South Asian breast cancer study in New Jersey during the COVID-19 pandemic.

BACKGROUND: South Asians are a rapidly growing population in the United States. Breast cancer is a major concern among South Asian American women, who are an understudied population. We established the South Asian Breast Cancer (SABCa) study in New Jersey during early 2020 to gain insights into their breast cancer-related health attitudes. Shortly after we started planning for the study, the COVID-19 disease spread throughout the world. In this paper, we describe our experiences and lessons learned from recruiting study participants by partnering with New Jersey's community organizations during the COVID-19 pandemic. METHODS: We used a cross-sectional design. We contacted 12 community organizations and 7 (58%) disseminated our study information. However, these organizations became considerably busy with pandemic-related needs. Therefore, we had to pivot to alternative recruitment strategies through community radio, Rutgers Cancer Institute of New Jersey's Community Outreach and Engagement Program, and Rutgers Cooperative Extension's community health programs. We recruited participants through these alternative strategies, obtained written informed consent, and collected demographic information using a structured survey. RESULTS: Twenty five women expressed interest in the study, of which 22 (88%) participated. Nine (41%) participants learned about the study through the radio, 5 (23%) through these participants, 1 (4.5%) through a non-radio community organization, and 7 (32%) through community health programs. Two (9%) participants heard about the study from their spouse. All participants were born outside the US, their average age was 52.4 years (range: 39-72 years), and they have lived in the US for an average of 26 years (range: 5-51 years). CONCLUSION: Pivoting to alternative strategies were crucial for successful recruitment. Findings suggest the significant potential of broadcast media for community-based recruitment. Family dynamics and the community's trust in our partners also encouraged participation. Such strategies must be considered when working with understudied populations.

Advancing Social and Environmental Research in Cancer Registries Using Geomasking for Address-Level Data.

Understanding the social and environmental causes of cancer in the United States, particularly in marginalized communities, is a major research priority. Population-based cancer registries are essential for advancing this research, given their nearly complete capture of incident cases within their catchment areas. Most registries limit the release of address-level geocodes linked to cancer outcomes to comply with state health departmental regulations. These policies ensure patient privacy, uphold data confidentiality, and enhance trust in research. However, these restrictions also limit the conduct of high-quality epidemiologic studies on social and environmental factors that may contribute to cancer burden. Geomasking refers to computational algorithms that distort locational data to attain a balance between effectively "masking" the original address location while faithfully maintaining the spatial structure in the data. We propose that the systematic deployment of scalable geomasking algorithms could accelerate research on social and environmental contributions across the cancer continuum by reducing measurement error bias while also protecting privacy. We encourage multidisciplinary teams of registry officials, geospatial analysts, cancer researchers, and others engaged in this form of research to evaluate and apply geomasking procedures based on feasibility of implementation, accuracy, and privacy protection to accelerate population-based research on social and environmental causes of cancer.

Sociodemographic disparities in targeted therapy in ovarian cancer in a national sample.

Background: The treatment landscape for ovarian cancer has changed in recent years with the introduction of targeted therapies to treat patients with advanced disease. We investigated patient demographic and clinical factors associated with use of targeted therapies as a part of the first-line treatment for ovarian cancer. Methods: This study included patients diagnosed with stage I-IV ovarian cancer between 2012 and 2019 from the National Cancer Database. Information on demographic and clinical characteristics were collected and described using frequency and percent across receipt of targeted therapy. Logistic regression was used to compute the odds ratios (ORs) and 95% confidence intervals (CI) associating patient demographic and clinical factors with receipt of targeted therapy. Results: Among 99,286 ovarian cancer patients (mean age 62 years), 4.1% received targeted therapy. The rate of targeted therapy receipt across racial and ethnic groups over the study period was fairly similar; however, non-Hispanic Black women were less likely to receive targeted therapy than their non-Hispanic White counterparts (OR=0.87, 95% CI: 0.76-1.00). Patients who received neoadjuvant chemotherapy were more likely to receive targeted therapy than those who received adjuvant chemotherapy (OR=1.26; 95% CI: 1.15-1.38). Moreover, among patients who received targeted therapy, 28% received neoadjuvant targeted therapy, with non-Hispanic Black women being most likely to receive neoadjuvant targeted therapy (34%) compared with other racial and ethnic groups. Conclusions: We observed differences in receipt of targeted therapy by factors such as age at diagnosis, stage, and comorbidities present at diagnosis, as well as factors related to healthcare access-including neighborhood education level and health insurance status. Approximately 28% of patients received targeted therapy in the neoadjuvant setting, which could negatively impact treatment outcomes and survival due to the increased risk of complications associated with targeted therapies that may delay or prevent surgery. These results warrant further evaluation in a cohort of patients with more comprehensive treatment information.

Neoadjuvant Chemotherapy in Ovarian Cancer: Are There Racial Disparities in Use and Survival?

BACKGROUND: We investigated racial and ethnic disparities in treatment sequence [i.e., neoadjuvant chemotherapy (NACT) plus interval debulking surgery (IDS) versus primary debulking surgery (PDS) plus adjuvant chemotherapy] among patients with ovarian cancer and its contribution to disparities in mortality. METHODS: Study included 37,566 women ages ≥18 years, diagnosed with stage III/IV ovarian cancer from the National Cancer Database (2004-2017). Logistic regression was used to compute ORs and 95% confidence intervals (CI) for racial and ethnic disparities in treatment sequence. Cox proportional hazards regression was used to estimate HRs and 95% CI for racial and ethnic disparities in all-cause mortality. RESULTS: Non-Hispanic Black (NHB) and Asian women were more likely to receive NACT plus IDS relative to PDS plus adjuvant chemotherapy than non-Hispanic White (NHW) women (OR: 1.12; 95% CI: 1.02-1.22 and OR: 1.12; 95% CI: 0.99-1.28, respectively). Compared with NHW women, NHB women had increased hazard of all-cause mortality (HR: 1.14; 95% CI: 1.09-1.20), whereas Asian and Hispanic women had a lower hazard of all-cause mortality (HR: 0.81; 95% CI: 0.74-0.88 and HR: 0.83; 95% CI: 0.77-0.88, respectively), which did not change after accounting for treatment sequence. CONCLUSIONS: NHB women were more likely to receive NACT plus IDS and experience a higher all-cause mortality rates than NHW women. IMPACT: Differences in treatment sequence did not explain racial disparities in all-cause mortality. Further evaluation of racial and ethnic differences in treatment and survival in a cohort of patients with detailed treatment information is warranted.

A model for an undergraduate research experience program in quantitative sciences.

We developed a summer research experience program within a freestanding comprehensive cancer center to cultivate undergraduate students with an interest in and an aptitude for quantitative sciences focused on oncology. This unconventional location for an undergraduate program is an ideal setting for interdisciplinary training in the intersection of oncology, statistics, and epidemiology. This paper describes the development and implementation of a hands-on research experience program in this unique environment. Core components of the program include faculty-mentored projects, instructional programs to improve research skills and domain knowledge, and professional development activities. We discuss key considerations such as effective partnership between research and administrative units, recruiting students, and identifying faculty mentors with quantitative projects. We describe evaluation approaches and discuss post-program outcomes and lessons learned. In its initial two years, the program successfully improved students' perception of competence gained in research skills and statistical knowledge across several knowledge domains. The majority of students also went on to pursue graduate degrees in a quantitative field or work in oncology-centric academic research roles. Our research-based training model can be adapted by a variety of organizations motivated to develop a summer research experience program in quantitative sciences for undergraduate students.

Factors Associated with Surgery Among South Asian American and Non-Hispanic White Women with Breast Cancer.

South Asian American (SA) women are diagnosed with more aggressive breast cancer than non-Hispanic White (NHW) women. Understanding the factors associated with the types of surgery received by these women sheds light on disease management in these culturally distinct populations. We used data on age at diagnosis, stage, grade, estrogen and progesterone receptors, and surgery from 4,590 SA and 429,030 NHW breast cancer cases in the Surveillance, Epidemiology and End Results (SEER) program. We used logistic regression with surgery as the binary outcome (subcutaneous, total, or radical mastectomy (STRM) versus partial mastectomy, no, unknown or other (PNUM)) and included additive effects of all the variables and interactions of age, stage, grade, and estrogen and progesterone receptors with race/ethnicity. Type I error of 5% was used to assess statistical significance of the effects. SA were significantly more likely than NHW cases to receive STRM relative to PNUM surgery among women diagnosed at or after age 50 years and having localized stage disease (Odds Ratio (OR) = 1.27, 95% Confidence Interval (CI) = 1.06 - 1.52). Further, SA were significantly less likely than NHW cases to receive STRM relative to PNUM surgery among those diagnosed before age 50 years and having regional or distant stage disease (OR = 0.75, 95% CI = 0.59 - 0.95 for age at diagnosis < 40 years; OR = 0.77, 95% CI = 0.62 - 0.95 for age at diagnosis 40-49 years). The type of surgery received by SA and NHW women differ according to age at diagnosis and disease stage.

Breast cancer among Asian Indian and Pakistani Americans: A surveillance, epidemiology and end results-based study.

Breast cancer incidence is increasing among Asian Indian and Pakistani women living in the United States. We examined the characteristics of breast cancer in Asian Indian and Pakistani American (AIPA) and non-Hispanic white (NHW) women using data from the surveillance, epidemiology and end results (SEER) program. Breast cancer incidence rates were estimated via segmented Poisson regression using data between 1990 and 2014 from SEER 9 registries, including New Jersey and California. Disease characteristics, treatment and survival information between 2000 and 2016 for 4900 AIPA and 482 250 NHW cases diagnosed after age 18 were obtained from SEER 18 registries and compared using descriptive analyses and multivariable competing risk proportional hazards regression. Breast cancer incidence was lower in AIPA than NHW women, increased with age and the rate of increase declined after age of 46 years. AIPA women were diagnosed at significantly younger age (mean (SD) = 54.5 (13.3) years) than NHW women (mean (SD) = 62 (14) years, P < .0001) and were more likely than NHW cases (P < .0001) to have regional or distant stage, higher grade, estrogen receptor-negative, progesterone receptor-negative, triple-negative or human epidermal growth factor receptor 2-enriched tumors, subcutaneous or total mastectomy, and lower cumulative incidence of death due to breast cancer (hazard ratio = 0.79, 95% CI: 0.72-0.86, P < .0001). AIPA had shorter median follow-up (52 months) than NHW cases (77 months). Breast cancer in AIPA women has unique characteristics that need to be further studied along with a comprehensive evaluation of their follow-up patterns.

Estimating Additive Interaction Effect in Stratified Two-Phase Case-Control Design.

BACKGROUND AND AIMS: There is considerable interest in epidemiology to estimate an additive interaction effect between two risk factors in case-control studies. An additive interaction is defined as the differential reduction in absolute risk associated with one factor between different levels of the other factor. A stratified two-phase case-control design is commonly used in epidemiology to reduce the cost of assembling covariates. It is crucial to obtain valid estimates of the model parameters by accounting for the underlying stratification scheme to obtain accurate and precise estimates of additive interaction effects. The aim of this paper is to examine the properties of different methods for estimating model parameters and additive interaction effects under a stratified two-phase case-control design. METHODS: Using simulations, we investigate the properties of three existing methods, namely stratum-specific offset, inverse-probability weighting, and multiple imputation for estimating model parameters and additive interaction effects. We also illustrate these properties using data from two published epidemiology studies. RESULTS: Simulation studies show that the multiple imputation method performs well when both the true and analysis models are additive (i.e., does not include multiplicative interaction terms) but does not provide a discernible advantage over the offset method when the analysis models are non-additive (i.e., includes multiplicative interaction terms). The offset method exhibits the best overall properties when the analysis model contains multiplicative interaction effects. CONCLUSION: When estimating additive interaction between risk factors in stratified two-phase case-control studies, we recommend estimating model parameters using multiple imputation when the analysis model is additive, and we recommend the offset method when the analysis model is non-additive.

A reconstructed melanoma data set for evaluating differential treatment benefit according to biomarker subgroups.

The data presented in this article are related to the research article entitled "Measuring differential treatment benefit across marker specific subgroups: the choice of outcome scale" (Satagopan and Iasonos, 2015) [1]. These data were digitally reconstructed from figures published in Larkin et al. (2015) [2]. This article describes the steps to digitally reconstruct patient-level data on time-to-event outcome and treatment and biomarker groups using published Kaplan-Meier survival curves. The reconstructed data set and the corresponding computer programs are made publicly available to enable further statistical methodology research.

Measuring differential treatment benefit across marker specific subgroups: The choice of outcome scale.

Clinical and epidemiological studies of anticancer therapies increasingly seek to identify predictive biomarkers to obtain insights into variation in treatment benefit. For time to event endpoints, a predictive biomarker is typically assessed using the interaction between the biomarker and treatment in a proportional hazards model. Interactions are contrasts of summaries of outcomes and depend upon the choice of the outcome scale. In this paper, we investigate interaction contrasts under three scales - the natural logarithm of hazard ratio, the natural logarithm of survival probability, and survival probability at a pre-specified time. We illustrate that we can have a non-zero interaction on survival or logarithm of survival probability scales even when there is no interaction on the logarithm of hazard ratio scale. Since survival probabilities have clinically useful interpretation and are easier to convey to patients than hazard ratios, we recommend evaluating a predictive biomarker using survival probabilities. We provide empirical illustration of the three scales of interaction for evaluating a predictive biomarker using reconstructed data from a published melanoma study.

Statistical interactions and Bayes estimation of log odds in case-control studies.

This paper is concerned with the estimation of the logarithm of disease odds (log odds) when evaluating two risk factors, whether or not interactions are present. Statisticians define interaction as a departure from an additive model on a certain scale of measurement of the outcome. Certain interactions, known as removable interactions, may be eliminated by fitting an additive model under an invertible transformation of the outcome. This can potentially provide more precise estimates of log odds than fitting a model with interaction terms. In practice, we may also encounter nonremovable interactions. The model must then include interaction terms, regardless of the choice of the scale of the outcome. However, in practical settings, we do not know at the outset whether an interaction exists, and if so whether it is removable or nonremovable. Rather than trying to decide on significance levels to test for the existence of removable and nonremovable interactions, we develop a Bayes estimator based on a squared error loss function. We demonstrate the favorable bias-variance trade-offs of our approach using simulations, and provide empirical illustrations using data from three published endometrial cancer case-control studies. The methods are implemented in an R program, and available freely at http://www.mskcc.org/biostatistics/~satagopj .

The study of nevi in children: Principles learned and implications for melanoma diagnosis.

Melanocytic nevi are a strong phenotypic marker of cutaneous melanoma risk. Changes in nevi during childhood and adolescence make these prime periods for studying nevogenesis. Insights gained by the study of nevi in childhood have implications for melanoma detection in both adults and children. A more comprehensive understanding of the morphologic characteristics of nevi in different anatomic locations, in association with the patient's age and pigmentary phenotype may aid in the identification of melanomas. When monitoring melanocytic lesions over time, it is essential to differentiate normal from abnormal change. This review summarizes the rapidly expanding body of literature relevant to nevus phenotype, particularly in the context of our experience with the Study of Nevi in Children (SONIC) Project.

Quantifying Treatment Benefit in Molecular Subgroups to Assess a Predictive Biomarker.

An increased interest has been expressed in finding predictive biomarkers that can guide treatment options for both mutation carriers and noncarriers. The statistical assessment of variation in treatment benefit (TB) according to the biomarker carrier status plays an important role in evaluating predictive biomarkers. For time-to-event endpoints, the hazard ratio (HR) for interaction between treatment and a biomarker from a proportional hazards regression model is commonly used as a measure of variation in TB. Although this can be easily obtained using available statistical software packages, the interpretation of HR is not straightforward. In this article, we propose different summary measures of variation in TB on the scale of survival probabilities for evaluating a predictive biomarker. The proposed summary measures can be easily interpreted as quantifying differential in TB in terms of relative risk or excess absolute risk due to treatment in carriers versus noncarriers. We illustrate the use and interpretation of the proposed measures with data from completed clinical trials. We encourage clinical practitioners to interpret variation in TB in terms of measures based on survival probabilities, particularly in terms of excess absolute risk, as opposed to HR. Clin Cancer Res; 22(9); 2114-20. ©2016 AACR.

Bayes and empirical Bayes methods for reduced rank regression models in matched case-control studies.

Matched case-control studies are popular designs used in epidemiology for assessing the effects of exposures on binary traits. Modern studies increasingly enjoy the ability to examine a large number of exposures in a comprehensive manner. However, several risk factors often tend to be related in a nontrivial way, undermining efforts to identify the risk factors using standard analytic methods due to inflated type-I errors and possible masking of effects. Epidemiologists often use data reduction techniques by grouping the prognostic factors using a thematic approach, with themes deriving from biological considerations. We propose shrinkage-type estimators based on Bayesian penalization methods to estimate the effects of the risk factors using these themes. The properties of the estimators are examined using extensive simulations. The methodology is illustrated using data from a matched case-control study of polychlorinated biphenyls in relation to the etiology of non-Hodgkin's lymphoma.

Statistical Interactions from a Growth Curve Perspective.

Logistic regression is widely used to evaluate the association between risk factors and a binary outcome. The logistic curve is symmetric around its point of inflection. Alternative families of curves, such as the additive Gompertz or Guerrero-Johnson models, have been proposed in various scenarios due to their asymmetry: disease risk may initially increase rapidly and be followed by a longer period where the rate of growth slowly decreases. When modeling binary outcomes in relation to risk factors, an additive logistic model may not provide a good fit to the data. Suppose the outcome and an additive function of the risk factors are indeed related through an asymmetric function, but we model the relationship using a logistic function. We illustrate - both from a mathematical framework and through a simulation-based evaluation - that higher-order terms, such as pairwise interactions and quadratic terms, may be required in a logistic regression model to obtain a good fit to the data. Importantly, as significant higher-order terms may be a manifestation of model misspecification, these terms should be cautiously interpreted; a more pragmatic approach is to develop contrasts of disease risk coming from a good fitting model. We illustrate these concepts in 2 cohort studies examining early death for late-stage colorectal and pancreatic cancer cases, and 2 case-control studies investigating NAT2 acetylation, smoking, and advanced colorectal adenoma and bladder cancer.

Prognostic and Predictive Values and Statistical Interactions in the Era of Targeted Treatment.

The current era of targeted treatment has accelerated the interest in studying gene-treatment, gene-gene, and gene-environment interactions using statistical models in the health sciences. Interactions are incorporated into models as product terms of risk factors. The statistical significance of interactions is traditionally examined using a likelihood ratio test (LRT). Epidemiological and clinical studies also evaluate interactions in order to understand the prognostic and predictive values of genetic factors. However, it is not clear how different types and magnitudes of interaction effects are related to prognostic and predictive values. The contribution of interaction to prognostic values can be examined via improvements in the area under the receiver operating characteristic curve due to the inclusion of interaction terms in the model (ΔAUC). We develop a resampling based approach to test the significance of this improvement and show that it is equivalent to LRT. Predictive values provide insights into whether carriers of genetic factors benefit from specific treatment or preventive interventions relative to noncarriers, under some definition of treatment benefit. However, there is no unique definition of the term treatment benefit. We show that ΔAUC and relative excess risk due to interaction (RERI) measure predictive values under two specific definitions of treatment benefit. We investigate the properties of LRT, ΔAUC, and RERI using simulations. We illustrate these approaches using published melanoma data to understand the benefits of possible intervention on sun exposure in relation to the MC1R gene. The goal is to evaluate possible interventions on sun exposure in relation to MC1R.

Predictors and long-term reproducibility of urinary phthalate metabolites in middle-aged men and women living in urban Shanghai.

Phthalate esters are man-made chemicals commonly used as plasticizers and solvents, and humans may be exposed through ingestion, inhalation, and dermal absorption. Little is known about predictors of phthalate exposure, particularly in Asian countries. Because phthalates are rapidly metabolized and excreted from the body following exposure, it is important to evaluate whether phthalate metabolites measured at a single point in time can reliably rank exposures to phthalates over a period of time. We examined the concentrations and predictors of phthalate metabolite concentrations among 50 middle-aged women and 50 men from two Shanghai cohorts, enrolled in 1997-2000 and 2002-2006, respectively. We assessed the reproducibility of urinary concentrations of phthalate metabolites in three spot samples per participant taken several years apart (mean interval between first and third sample was 7.5 years [women] or 2.9 years [men]), using Spearman's rank correlation coefficients and intra-class correlation coefficients. We detected ten phthalate metabolites in at least 50% of individuals for two or more samples. Participant sex, age, menopausal status, education, income, body mass index, consumption of bottled water, recent intake of medication, and time of day of collection of the urine sample were associated with concentrations of certain phthalate metabolites. The reproducibility of an individual's urinary concentration of phthalate metabolites across several years was low, with all intra-class correlation coefficients and most Spearman rank correlation coefficients ≤0.3. Only mono(2-ethylhexyl) phthalate, a metabolite of di(2-ethylhexyl) phthalate, had a Spearman rank correlation coefficient ≥0.4 among men, suggesting moderate reproducibility. These findings suggest that a single spot urine sample is not sufficient to rank exposures to phthalates over several years in an adult urban Chinese population.

Sunburn, sun exposure, and sun sensitivity in the Study of Nevi in Children.

PURPOSE: To examine the joint effect of sun exposure and sunburn on nevus counts (on the natural logarithm scale; log nevi) and the role of sun sensitivity. METHODS: We describe an analysis of cross-sectional data from 443 children enrolled in the prospective Study of Nevi in Children. To evaluate the joint effect, we partitioned the sum of squares because of interaction between sunburn and sun exposure into orthogonal components representing (1) monotonic increase in log nevi with increasing sun exposure (rate of increase of log nevi depends on sunburn), and (2) nonmonotonic pattern. RESULTS: In unadjusted analyses, there was a marginally significant monotonic pattern of interaction (P = .08). In adjusted analyses, sun exposure was associated with higher log nevi among those without sunburn (P < .001), but not among those with sunburn (P = .14). Sunburn was independently associated with log nevi (P = .02), even though sun sensitivity explained 29% (95% confidence interval: 2%-56%, P = .04) of its effect. Children with high sun sensitivity and sunburn had more nevi, regardless of sun exposure. CONCLUSIONS: A program of increasing sun protection in early childhood as a strategy for reducing nevi, when applied to the general population, may not equally benefit everyone.

Blocking and randomization to improve molecular biomarker discovery.

Randomization and blocking have the potential to prevent the negative impacts of nonbiologic effects on molecular biomarker discovery. Their use in practice, however, has been scarce. To demonstrate the logistic feasibility and scientific benefits of randomization and blocking, we conducted a microRNA study of endometrial tumors (n = 96) and ovarian tumors (n = 96) using a blocked randomization design to control for nonbiologic effects; we profiled the same set of tumors for a second time using no blocking or randomization. We assessed empirical evidence of differential expression in the two studies. We performed simulations through virtual rehybridizations to further evaluate the effects of blocking and randomization. There was moderate and asymmetric differential expression (351/3,523, 10%) between endometrial and ovarian tumors in the randomized dataset. Nonbiologic effects were observed in the nonrandomized dataset, and 1,934 markers (55%) were called differentially expressed. Among them, 185 were deemed differentially expressed (185/351, 53%) and 1,749 not differentially expressed (1,749/3,172, 55%) in the randomized dataset. In simulations, when randomization was applied to all samples at once or within batches of samples balanced in tumor groups, blocking improved the true-positive rate from 0.95 to 0.97 and the false-positive rate from 0.02 to 0.002; when sample batches were unbalanced, randomization was associated with the true-positive rate (0.92) and the false-positive rate (0.10) regardless of blocking. Normalization improved the detection of true-positive markers but still retained sizeable false-positive markers. Randomization and blocking should be used in practice to more fully reap the benefits of genomics technologies.