This paper addresses patient heterogeneity associated with prediction problems in biomedical applications. We propose a systematic hypothesis testing approach to determine the existence of patient subgroup structure and the number of subgroups in patient population if subgroups exist. A mixture of generalized linear models is considered to model the relationship between the disease outcome and patient characteristics and clinical factors, including targeted biomarker profiles. We construct a test statistic based on expectation maximization (EM) algorithm and derive its asymptotic distribution under the null hypothesis. An important computational advantage of the test is that the involved parameter estimates under the complex alternative hypothesis can be obtained through a small number of EM iterations, rather than optimizing the objective function. We demonstrate the finite sample performance of the proposed test in terms of type-I error rate and power, using extensive simulation studies. The applicability of the proposed method is illustrated through an application to a multicenter prostate cancer study.

There is an increasing need to construct a risk-prediction scoring system for survival data and identify important risk factors (e.g., biomarkers) for patient screening and treatment recommendation. However, most existing methodologies either rely on strong model assumptions (e.g., proportional hazards) or only handle binary outcomes. In this article, we propose a flexible method that simultaneously selects important risk factors and identifies the optimal linear combination of risk factors by maximizing a pseudo-likelihood function based on the time-dependent area under the receiver operating characteristic curve. Our method is particularly useful for risk evaluation and recommendation of optimal subsequent treatments. We show that the proposed method has desirable theoretical properties, including asymptotic normality and the oracle property after variable selection. Numerical performance is evaluated on several simulation data sets and an application to hepatocellular carcinoma data.

There is an increasing need to construct a risk-prediction scoring system for survival data and identify important risk factors (e.g., biomarkers) for patient screening and treatment recommendation. However, most existing methodologies either rely on strong model assumptions (e.g., proportional hazards) or only handle binary outcomes. In this article, we propose a flexible method that simultaneously selects important risk factors and identifies the optimal linear combination of risk factors by maximizing a pseudo-likelihood function based on the time-dependent area under the receiver operating characteristic curve. Our method is particularly useful for risk evaluation and recommendation of optimal subsequent treatments. We show that the proposed method has desirable theoretical properties, including asymptotic normality and the oracle property after variable selection. Numerical performance is evaluated on several simulation data sets and an application to hepatocellular carcinoma data.

Research progress in diseases of the exocrine pancreas [chronic pancreatitis (CP), pancreatogenic diabetes mellitus, and pancreatic cancer] has been hampered by the disorders' heterogeneity, the limitations of previous small cross-sectional studies, the inability to safely obtain pancreatic tissue for study, and the lack of structured epidemiology tools, genetic testing, and biomarker development. Mechanism-based research of these diseases has suffered from the lack of systematically collected clinical measures in longitudinal cohort studies linked with biospecimens. Given the increasing incidence and prevalence of CP and its association to the development of pancreatic cancer, its complications, high mortality rate, and associated health care cost, the National Institute for Diabetes and Digestive and Kidney Diseases and the National Cancer Institute funded the Consortium for the Study of Chronic Pancreatitis, Diabetes and Pancreatic Cancer to identify research gaps and foster multidisciplinary collaborations to better diagnose, characterize, and manage CP and its sequelae. The CPDPC structure, governance, and research objectives are described in this article. Studies undertaken by the CPDPC are described in other articles in this journal's issue.

Purpose

This study evaluated the discriminatory power of salivary transcriptomic and proteomic biomarkers in distinguishing oral squamous cell carcinoma cases from controls and potentially malignant oral disorders (PMOD).

Experimental design

A total of 180 samples (60 OSCC patients, 60 controls, and 60 PMOD patients) were used in the study. Seven transcriptomic markers (IL8, IL1β, SAT1, OAZ1, DUSP1, S100P, and H3F3A) were measured using qPCR, and two proteomic markers (IL8 and IL1β) were evaluated by ELISA.

Results

Among 7 transcriptomic markers, transcript level of DUSP1 was significantly lower in OSCC patients than in controls and PMOD patients. Between the proteomic markers, the protein concentration of IL8 and IL1β was significantly higher in OSCC patients than controls and dysplasia patients. Univariate fractional polynomial (FP) models revealed that salivary IL8 protein (IL8p) has the highest AUC value between OSCC patients and controls (0.74) and between OSCC and PMOD patients (0.72). Applying a 2-marker FP model, salivary IL8p combined with IL1β gave the best AUC value for discrimination between OSCC patients and controls, as well as the IL8p combined with H3F3A mRNA, which gave the best AUC value for discrimination between OSCC and PMOD patients. Multivariate models analysis combining salivary analytes and risk factor exposure related to oral carcinogenesis formed the best combinatory variables for differentiation between OSCC versus PMOL (AUC = 0.80), OSCC versus controls (AUC = 0.87), and PMOD versus controls (AUC = 0.78).

Conclusions

The combination of transcriptomic and proteomic salivary markers is of great value for oral cancer detection and differentiation from PMOD patients and controls. Clin Cancer Res; 22(13); 3340-7. ©2016 AACR.

The diagnostic work-up for Sjögrens syndrome is challenging and complex, including testing for serum autoantibodies to SSA/Ro and a labial salivary gland biopsy. Furthermore, the diagnosis is often delayed. In this study, we tested the hypothesis that anti-SSA/Ro autoantibodies are detectable in the saliva of patients with primary Sjögrens syndrome (pSS) because the disease affects the salivary glands, and these autoantibodies display greater discriminatory performance in saliva than in serum. SSA/Ro-52 Ags were used to develop what is, to our knowledge, a novel quantitative electrochemical-based immunoassay: the electric field-induced release and measurement (EFIRM) platform. The clinical utility was determined by measuring salivary anti-SSA/Ro-52 autoantibodies in patients with pSS and sicca (n = 34), patients without pSS with sicca (n = 35), and healthy subjects (n = 41). The statistical analysis of discrimination included the area under the receiver operating characteristic curve. Salivary anti-SSA/Ro-52 autoantibodies were measured in 94% (32 of 34) of patients with pSS with 85% (29 of 34) seropositivity. Four of the five seronegative patients with pSS had EFIRM-measurable anti-SSA/Ro-52 autoantibodies in saliva. Additionally, 60% (21 of 35) of the seronegative patients without pSS who had sicca had EFIRM-detectable SSA/Ro-52 autoantibodies in saliva, indicating the onset of autoimmune disease. Two of the 41 healthy control subjects had EFIRM-detectable SSA/Ro-52 autoantibodies in their saliva. Salivary SSA/Ro-52 autoantibodies significantly discriminated patients with pSS or patients with the initial stage of autoimmune disease from healthy subjects with an area under the receiver operating characteristic curve of 0.91. Our findings suggest that the proposed saliva SSA/Ro-52 immunoassay improves early and accurate diagnosis of seronegative patients with pSS and patients with early-onset autoimmune disease.

Rationale

Constitutive activation of the epidermal growth factor receptor (EGFR) is prevalent in epithelial cancers, particularly in non-small cell lung carcinoma (NSCLC). Mutations identified in EGFR predict the sensitivity to EGFR-targeted therapy. Detection of these mutations is mainly based on tissue biopsy, which is invasive, expensive, and time consuming.

Objectives

Noninvasive, real-time, inexpensive detection and monitoring of EGFR mutations in patients with NSCLC is highly desirable.

Methods

We developed a novel core technology, electric field-induced release and measurement (EFIRM), which relies on a multiplexible electrochemical sensor that can detect EGFR mutations directly in bodily fluids.

Measurements and main results

We established EFIRM for the detection of the EGFR mutations in vitro and correlated the results with tumor size from xenografted mice. In clinical application, we demonstrated that EFIRM could detect EGFR mutations in the saliva and plasma of 22 patients with NSCLC. Finally, a blinded test was performed on saliva samples from 40 patients with NSCLC. The receiver operating characteristic analysis indicated that EFIRM detected the exon 19 deletion with an area under the curve of 0.94 and the L858R mutation with an area under the curve of 0.96.

Conclusions

Our data indicate that EFIRM is effective, accurate, rapid, user-friendly, and cost effective for the detection of EGFR mutations in the saliva of patients with NSCLC. We termed this saliva-based EGFR mutation detection (SABER).

Purpose

We have investigated the potential of metabolomics to discover blood-based biomarkers relevant to lung cancer screening and early detection. An untargeted metabolomics approach was applied to identify biomarker candidates using prediagnostic sera from the Beta-Carotene and Retinol Efficacy Trial (CARET) study.

Patients and methods

A liquid chromatography/mass spectrometry hydrophilic interaction method designed to profile a wide range of metabolites was applied to prediagnostic serum samples from CARET participants (current or former heavy smokers), consisting of 100 patients who subsequently developed non-small-cell lung cancer (NSCLC) and 199 matched controls. A separate aliquot was used to quantify levels of pro-surfactant protein B (pro-SFTPB), a previously established protein biomarker for NSCLC. On the basis of the results from the discovery set, blinded validation of a metabolite, identified as N(1),N(12)-diacetylspermine (DAS), and pro-SFTPB was performed using an independent set of CARET prediagnostic sera from 108 patients with NSCLC and 216 matched controls.

Results

Serum DAS was elevated by 1.9-fold, demonstrating significant specificity and sensitivity in the discovery set for samples collected up to 6 months before diagnosis of NSCLC. In addition, DAS significantly complemented performance of pro-SFTPB in both the discovery and validations sets, with a combined area under the curve in the validation set of 0.808 (P < .001 v pro-SFTPB).

Conclusion

DAS is a novel serum metabolite with significant performance in prediagnostic NSCLC and has additive performance with pro-SFTPB.

Previous studies have suggested occurrence of altered serum glycan profiles in patients with lung cancer. Here, we aimed to determine the predictive value of serum glycans to distinguish non-small cell lung cancer (NSCLC) cases from controls in prediagnostic samples using a previously validated predictive protein marker pro-SFTPB, as anchor. Blinded prediagnostic serum samples were obtained from the Carotene and Retinol Efficacy Trial (CARET), and included a discovery set of 100 NSCLC cases and 199 healthy controls. A second test set consisted of 108 cases and 216 controls. Cases and controls were matched for age at baseline (5-year groups), sex, smoking status (current vs. former), study enrollment cohort, and date of blood draw. Serum glycan profiles were determined by mass spectrometry. Twelve glycan variables were identified to have significant discriminatory power between cases and controls in the discovery set (AUC > 0.6). Of these, four were confirmed in the independent validation set. A combination marker yielded AUCs of 0.74 and 0.64 in the discovery and test set, respectively. Four glycan variables exhibited significant incremental value when combined with pro-SFTPB compared with pro-SFTPB alone with AUCs of 0.73, 0.72, 0.72, and 0.72 in the test set, indicating that serum glycan signatures have relevance to risk assessment for NSCLC.

Mutations identified in the epidermal growth factor receptor (EGFR) predict sensitivity to EGFR-targeted therapy for non-small cell lung carcinoma (NSCLC). We previously reported that Electric Field-Induced Release and Measurement (EFIRM)-based liquid biopsy could detect EGFR ctDNA with >94% concordance with tissue-based genotyping. A side-by-side comparison of concordance of EFIRM and droplet digital PCR (ddPCR) for the detection of the two front-line actionable EFGR mutations was performed with paired plasma and saliva samples from 13 NSCLC patients. Deep sequencing analysis based on single-strand DNA library preparation was employed to determine the size distributions of EGFR L858R ctDNA in plasma and saliva samples. EFIRM detected both EGFR mutations with 100% sensitivity in both plasma and saliva samples, whereas ddPCR detected EGFR mutations with sensitivities of 84.6% and 15.4%, respectively. In saliva samples, the majority of EGFR L858R ctDNA fragments detected were <80 bp. Deep sequencing analysis of ctDNA enriched for the EGFR L858R mutation revealed the significant presence of EGFR L858R ctDNA as ultra-short circulating tumor DNA (usctDNA) with the size of 40-60 bp in patient plasma and saliva. Most of usctDNAs are not amplifiable with the current ddPCR assay. Further examination using cell lines and patient biofluids revealed that the majority of usctDNAs were predominately localized in the exosomal fraction. Our study revealed the abundant existence of EGFR ctDNA in the plasma and saliva of NSCLC patients is usctDNA. usctDNA is a novel type of targets for liquid biopsy that can be efficiently detected by EFIRM technology.