A comparison of statistical methods for the study of etiologic heterogeneity

Method for Testing Etiologic Heterogeneity Among Noncompeting Diagnoses, Applied to Impact of Perinatal Exposures on Autism and Attention Deficit Hyperactivity Disorder

BACKGROUND

Testing etiologic heterogeneity, whether a disorder subtype is more or less impacted by a risk factor, is important for understanding causal pathways and optimizing statistical power. The study of mental health disorders especially benefits from strategic subcategorization because these disorders are heterogeneous and frequently co-occur. Existing methods to quantify etiologic heterogeneity are not appropriate for noncompeting events in an open cohort of variable-length follow-up. Thus, we developed a new method.

METHODS

We estimated risks from urban residence, maternal smoking during pregnancy, and parental psychiatric history, with subtypes defined by the presence or absence of a codiagnosis: autism alone, attention deficit hyperactivity disorder (ADHD) alone, and joint diagnoses of autism + ADHD. To calculate the risk of a single diagnosis (e.g., autism alone), we subtracted the risk for autism + ADHD from the risk for autism overall. We tested the equivalency of average risk ratios over time, using a Wald-type test and bootstrapped standard errors.

RESULTS

Urban residence was most strongly linked with autism + ADHD and least with ADHD only; maternal smoking was associated with ADHD only but not autism only; and parental psychiatric history exhibited similar associations with all subgroups.

CONCLUSION

Our method allowed the calculation of appropriate P values to test the strength of association, informing etiologic heterogeneity wherein two of these three risk factors exhibited different impacts across diagnostic subtypes. The method used all available data, avoided neurodevelopmental outcome misclassification, exhibited robust statistical precision, and is applicable to similar heterogeneous complex conditions using common diagnostic data with variable follow-up.

Modification of the Association Between Frequent Aspirin Use and Ovarian Cancer Risk: A Meta-Analysis Using Individual-Level Data From Two Ovarian Cancer Consortia

PURPOSE

Frequent aspirin use has been associated with reduced ovarian cancer risk, but no study has comprehensively assessed for effect modification. We leveraged harmonized, individual-level data from 17 studies to examine the association between frequent aspirin use and ovarian cancer risk, overall and across subgroups of women with other ovarian cancer risk factors.

METHODS

Nine cohort studies from the Ovarian Cancer Cohort Consortium (n = 2,600 cases) and eight case-control studies from the Ovarian Cancer Association Consortium (n = 5,726 cases) were included. We used Cox regression and logistic regression to assess study-specific associations between frequent aspirin use (≥ 6 days/week) and ovarian cancer risk and combined study-specific estimates using random-effects meta-analysis. We conducted analyses within subgroups defined by individual ovarian cancer risk factors (endometriosis, obesity, family history of breast/ovarian cancer, nulliparity, oral contraceptive use, and tubal ligation) and by number of risk factors (0, 1, and ≥ 2).

RESULTS

Overall, frequent aspirin use was associated with a 13% reduction in ovarian cancer risk (95% CI, 6 to 20), with no significant heterogeneity by study design (P = .48) or histotype (P = .60). Although no association was observed among women with endometriosis, consistent risk reductions were observed among all other subgroups defined by ovarian cancer risk factors (relative risks ranging from 0.79 to 0.93, all P-heterogeneity > .05), including women with ≥ 2 risk factors (relative risk, 0.81; 95% CI, 0.73 to 0.90).

CONCLUSION

This study, the largest to-date on aspirin use and ovarian cancer, provides evidence that frequent aspirin use is associated with lower ovarian cancer risk regardless of the presence of most other ovarian cancer risk factors. Risk reductions were also observed among women with multiple risk factors, providing proof of principle that chemoprevention programs with frequent aspirin use could target higher-risk subgroups.

A Generalized Integration Approach to Association Analysis with Multi-category Outcome: An Application to a Tumor Sequencing Study of Colorectal Cancer and Smoking

Cancer is a heterogeneous disease, and rapid progress in sequencing and -omics technologies has enabled researchers to characterize tumors comprehensively. This has stimulated an intensive interest in studying how risk factors are associated with various tumor heterogeneous features. The Cancer Prevention Study-II (CPS-II) cohort is one of the largest prospective studies, particularly valuable for elucidating associations between cancer and risk factors. In this paper, we investigate the association of smoking with novel colorectal tumor markers obtained from targeted sequencing. However, due to cost and logistic difficulties, only a limited number of tumors can be assayed, which limits our capability for studying these associations. Meanwhile, there are extensive studies for assessing the association of smoking with overall cancer risk and established colorectal tumor markers. Importantly, such summary information is readily available from the literature. By linking this summary information to parameters of interest with proper constraints, we develop a generalized integration approach for polytomous logistic regression model with outcome characterized by tumor features. The proposed approach gains the efficiency through maximizing the joint likelihood of individual-level tumor data and external summary information under the constraints that narrow the parameter searching space. We apply the proposed method to the CPS-II data and identify the association of smoking with colorectal cancer risk differing by the mutational status of APC and RNF43 genes, neither of which is identified by the conventional analysis of CPS-II individual data only. These results help better understand the role of smoking in the etiology of colorectal cancer.

Reflection on modern methods: causal inference considerations for heterogeneous disease etiology

Molecular pathological epidemiology research provides information about pathogenic mechanisms. A common study goal is to evaluate whether the effects of risk factors on disease incidence vary between different disease subtypes. A popular approach to carrying out this type of research is to implement a multinomial regression in which each of the non-zero values corresponds to a bona fide disease subtype. Then, heterogeneity in the exposure effects across subtypes is examined by comparing the coefficients of the exposure between the different subtypes. In this paper, we explain why this common method potentially cannot recover causal effects, even when all confounders are measured, due to a particular type of selection bias. This bias can be explained by recognizing that the multinomial regression is equivalent to a series of logistic regressions; each compares cases of a certain subtype to the controls. We further explain how this bias arises using directed acyclic graphs and we demonstrate the potential magnitude of the bias by analysis of a hypothetical data set and by a simulation study.

Validity of a method for identifying disease subtypes that are etiologically heterogeneous

A focus of cancer epidemiologic research has become the identification of risk factors that influence specific subtypes of disease, a phenomenon known as etiologic heterogeneity. In previous work we developed a novel strategy to cluster tumor markers and identify disease subtypes that differ maximally with respect to known risk factors for use in the context of case-control studies. The method relies on the premise that unsupervised k-means clustering will find candidate solutions that are closely aligned with the sought-after etiologically distinct clusters, which may not be true in the presence of clusters of tumor markers that are not related to risk of disease. In this article, we investigate in detail the ability of the method to identify the "true" clusters in the presence of clusters that are unrelated to risk factors, what we term "counterfeit" clusters. We find that our method works when the strength of structure is larger in the clusters that truly represent etiologic heterogeneity than in the counterfeit clusters, but when this condition is not met, or when there are many tumor markers that simply represent noise, the method will not find the correct solution without first performing variable selection to identify the tumor markers most strongly related to the risk factors. We illustrate the results using data from a breast cancer case-control study.

Risk factors for melanoma by anatomical site: an evaluation of aetiological heterogeneity

BACKGROUND

Melanoma aetiology has been proposed to have two pathways, which are determined by naevi and type of sun exposure and related to the anatomical site where melanoma develops.

OBJECTIVES

We examined associations with melanoma by anatomical site for a comprehensive set of risk factors including pigmentary and naevus phenotypes, ultraviolet radiation exposure and polygenic risk.

METHODS

We analysed harmonized data from 2617 people with incident first invasive melanoma and 975 healthy controls recruited through two population-based case-control studies in Australia and the UK. Questionnaire data were collected by interview using a single protocol, and pathway-specific polygenic risk scores were derived from DNA samples. We estimated adjusted odds ratios using unconditional logistic regression that compared melanoma cases at each anatomical site with all controls.

RESULTS

When cases were compared with control participants, there were stronger associations for many naevi vs. no naevi for melanomas on the trunk, and upper and lower limbs than on the head and neck (P-heterogeneity < 0·001). Very fair skin (vs. olive/brown skin) was more weakly related to melanoma on the trunk than to melanomas at other sites (P-heterogeneity = 0·04). There was no significant difference by anatomical site for polygenic risk. Increased weekday sun exposure was positively associated with melanoma on the head and neck but not on other sites.

CONCLUSIONS

We found evidence of aetiological heterogeneity for melanoma, supporting the dual pathway hypothesis. These findings enhance understanding of risk factors for melanoma and can guide prevention and skin examination education and practices.

A mixed-model approach for powerful testing of genetic associations with cancer risk incorporating tumor characteristics

Cancers are routinely classified into subtypes according to various features, including histopathological characteristics and molecular markers. Previous genome-wide association studies have reported heterogeneous associations between loci and cancer subtypes. However, it is not evident what is the optimal modeling strategy for handling correlated tumor features, missing data, and increased degrees-of-freedom in the underlying tests of associations. We propose to test for genetic associations using a mixed-effect two-stage polytomous model score test (MTOP). In the first stage, a standard polytomous model is used to specify all possible subtypes defined by the cross-classification of the tumor characteristics. In the second stage, the subtype-specific case-control odds ratios are specified using a more parsimonious model based on the case-control odds ratio for a baseline subtype, and the case-case parameters associated with tumor markers. Further, to reduce the degrees-of-freedom, we specify case-case parameters for additional exploratory markers using a random-effect model. We use the Expectation-Maximization algorithm to account for missing data on tumor markers. Through simulations across a range of realistic scenarios and data from the Polish Breast Cancer Study (PBCS), we show MTOP outperforms alternative methods for identifying heterogeneous associations between risk loci and tumor subtypes. The proposed methods have been implemented in a user-friendly and high-speed R statistical package called TOP (https://github.com/andrewhaoyu/TOP).

Acute sleep deprivation and culpable motor vehicle crash involvement

Study Objectives

To quantify the relationship between acute sleep deprivation and culpable involvement in motor vehicle crashes.

Methods

Participants were 6845 drivers involved in a representative sample of crashes investigated by the US Department of Transportation in years 2005-2007. A modified case-control study design was used to compare self-reported hours of sleep in the 24 hr before crashing between drivers deemed culpable versus nonculpable. Analyses controlled for fatigue-related, driver-related, and environmental factors. Specific errors that led to crashes were also examined.

Results

Drivers who reported having slept for 6, 5, 4, and less than 4 hr in the 24 hr before crashing had 1.3 (95% confidence interval [CI] = 1.04 to 1.7), 1.9 (1.1 to 3.2), 2.9 (1.4 to 6.2), and 15.1 (4.2 to 54.4) times the odds, respectively, of having been culpable for their crashes, compared with drivers who reported 7-9 hr of sleep. Drivers who had slept less than 4 hr had 3.4 (95% CI = 2.1 to 5.6) times the increase in odds of culpable involvement in single-vehicle crashes compared with multiple-vehicle crashes. Recent change in sleep schedule, typically feeling drowsy upon waking, and driving for 3+ hr were also associated with culpability (all p ≤ 0.013). Assuming nonculpable drivers comprised a representative sample of all drivers present where crashes occurred, these odds ratios approximate incidence rate ratios for culpable crash involvement per unit of time driving.

Conclusions

Driving after having slept less than 7 hr in a 24 hr period is associated with elevated risk of culpable crash involvement. Risk is greatest for drivers who have slept less than 4 hr and is manifested disproportionately in single-vehicle crashes.

Integration of microbiology, molecular pathology, and epidemiology: a new paradigm to explore the pathogenesis of microbiome-driven neoplasms

Molecular pathological epidemiology (MPE) is an integrative transdisciplinary field that addresses heterogeneous effects of exogenous and endogenous factors (collectively termed 'exposures'), including microorganisms, on disease occurrence and consequences, utilising molecular pathological signatures of the disease. In parallel with the paradigm of precision medicine, findings from MPE research can provide aetiological insights into tailored strategies of disease prevention and treatment. Due to the availability of molecular pathological tests on tumours, the MPE approach has been utilised predominantly in research on cancers including breast, lung, prostate, and colorectal carcinomas. Mounting evidence indicates that the microbiome (inclusive of viruses, bacteria, fungi, and parasites) plays an important role in a variety of human diseases including neoplasms. An alteration of the microbiome may be not only a cause of neoplasia but also an informative biomarker that indicates or mediates the association of an epidemiological exposure with health conditions and outcomes. To adequately educate and train investigators in this emerging area, we herein propose the integration of microbiology into the MPE model (termed 'microbiology-MPE'), which could improve our understanding of the complex interactions of environment, tumour cells, the immune system, and microbes in the tumour microenvironment during the carcinogenic process. Using this approach, we can examine how lifestyle factors, dietary patterns, medications, environmental exposures, and germline genetics influence cancer development and progression through impacting the microbial communities in the human body. Further integration of other disciplines (e.g. pharmacology, immunology, nutrition) into microbiology-MPE would expand this developing research frontier. With the advent of high-throughput next-generation sequencing technologies, researchers now have increasing access to large-scale metagenomics as well as other omics data (e.g. genomics, epigenomics, proteomics, and metabolomics) in population-based research. The integrative field of microbiology-MPE will open new opportunities for personalised medicine and public health. Copyright © 2019 Pathological Society of Great Britain and Ireland. Published by John Wiley & Sons, Ltd.

Integrative analysis of exogenous, endogenous, tumour and immune factors for precision medicine

Immunotherapy strategies targeting immune checkpoints such as the CTLA4 and CD274 (programmed cell death 1 ligand 1, PD-L1)/PDCD1 (programmed cell death 1, PD-1) T-cell coreceptor pathways are revolutionising oncology. The approval of pembrolizumab use for solid tumours with high-level microsatellite instability or mismatch repair deficiency by the US Food and Drug Administration highlights promise of precision immuno-oncology. However, despite evidence indicating influences of exogenous and endogenous factors such as diet, nutrients, alcohol, smoking, obesity, lifestyle, environmental exposures and microbiome on tumour-immune interactions, integrative analyses of those factors and immunity lag behind. Immune cell analyses in the tumour microenvironment have not adequately been integrated into large-scale studies. Addressing this gap, the transdisciplinary field of molecular pathological epidemiology (MPE) offers research frameworks to integrate tumour immunology into population health sciences, and link the exposures and germline genetics (eg, HLA genotypes) to tumour and immune characteristics. Multilevel research using bioinformatics, in vivo pathology and omics (genomics, epigenomics, transcriptomics, proteomics and metabolomics) technologies is possible with use of tissue, peripheral blood circulating cells, cell-free plasma, stool, sputum, urine and other body fluids. This immunology-MPE model can synergise with experimental immunology, microbiology and systems biology. GI neoplasms represent exemplary diseases for the immunology-MPE model, given rich microbiota and immune tissues of intestines, and the well-established carcinogenic role of intestinal inflammation. Proof-of-principle studies on colorectal cancer provided insights into immunomodulating effects of aspirin, vitamin D, inflammatory diets and omega-3 polyunsaturated fatty acids. The integrated immunology-MPE model can contribute to better understanding of environment-tumour-immune interactions, and effective immunoprevention and immunotherapy strategies for precision medicine.

Ten Essential Practices for Developing or Reforming a Biostatistics Core for a NCI Designated Cancer Center

There are 69 National Cancer Institute (NCI) designated Cancer Centers (CCs) in the United States. Biostatistical collaboration is pivotal in cancer research, and support for a cancer biostatistics shared resource facility (C-BSRF) is included in the award. Although the services and staff needed in a C-BSRF have been outlined in general terms and best practices for biostatistical consultations and collaboration in an academic health center have been agreed upon, implementing these practices in the demanding setting of cancer centers interested in pursuing or maintaining NCI designation remains challenging. We surveyed all C-BSRF websites to assess their organizational charts, governance, size, services provided, and financial models and have identified 10 essential practices for the development of a successful C-BSRF. Here, we share our success with, and barriers to, implementation of these practices. Showcasing development plans for these essential practices resulted in an NCI score of "Excellent to Outstanding" for our C-BSRF in 2015, and performance metrics in 2016-2017 demonstrated notable improvement since our original Cancer Center Support Grant (CCSG) application in 2014. We believe that the essential practices described here can be adapted and adjusted, as needed, for CCs of various sizes and with different types of cancer research programs.

Integration of pharmacology, molecular pathology, and population data science to support precision gastrointestinal oncology

Precision medicine has a goal of customizing disease prevention and treatment strategies. Under the precision medicine paradigm, each patient has unique pathologic processes resulting from cellular genomic, epigenomic, proteomic, and metabolomic alterations, which are influenced by pharmacological, environmental, microbial, dietary, and lifestyle factors. Hence, to realize the promise of precision medicine, multi-level research methods that can comprehensively analyze many of these variables are needed. In order to address this gap, the integrative field of molecular pathology and population data science (i.e., molecular pathological epidemiology) has been developed to enable such multi-level analyses, especially in gastrointestinal cancer research. Further integration of pharmacology can improve our understanding of drug effects, and inform decision-making of drug use at both the individual and population levels. Such integrative research demonstrated potential benefits of aspirin in colorectal carcinoma with PIK3CA mutations, providing the basis for new clinical trials. Evidence also suggests that HPGD (15-PDGH) expression levels in normal colon and the germline rs6983267 polymorphism that relates to tumor CTNNB1 (β-catenin)/WNT signaling status may predict the efficacy of aspirin for cancer chemoprevention. As immune checkpoint blockade targeting the CD274 (PD-L1)/PDCD1 (PD-1) pathway for microsatellite instability-high (or mismatch repair-deficient) metastatic gastrointestinal or other tumors has become standard of care, potential modifying effects of dietary, lifestyle, microbial, and environmental factors on immunotherapy need to be studied to further optimize treatment strategies. With its broad applicability, our integrative approach can provide insights into the interactive role of medications, exposures, and molecular pathology, and guide the development of precision medicine.

The clinical trajectory of emerging bipolar disorder among the high-risk offspring of bipolar parents: current understanding and future considerations

BACKGROUND

Relatively little is known about the onset of bipolar disorder, yet the early illness course is already associated with significant morbidity and mortality. Therefore, characterizing the bipolar illness trajectory is key to risk prediction and early intervention advancement.

MAIN BODY

In this narrative review, we discuss key findings from prospective longitudinal studies of the high-risk offspring of bipolar parents and related meta-analyses that inform us about the clinical trajectory of emerging bipolar disorder. Challenges such as phenotypic and etiologic heterogeneity and the non-specificity of early symptoms and syndromes are highlighted. Implications of the findings for both research and clinical practice are discussed.

CONCLUSION

Bipolar disorder in young people at familial risk does not typically onset with a hypomanic or manic episode. Rather the first activated episode is often preceded by years of impairing psychopathological states that vary over development and across emerging bipolar subtype. Taking heterogeneity into account and adopting a more comprehensive approach to diagnosis seems necessary to advance earlier identification and our understanding of the onset of bipolar disorder.