Thermal signatures in breast cancer: Deciphering latent biomarkers through deep learning and explainable AI.

Worldwide cancer statistics have shown that breast cancer dominates female cancer incidence and remains a leading cause of death. The Systemic Immune-Inflammation Index (SII) is a new prognostic indicator of systemic inflammation used to assess systemic immune-inflammatory response levels in the human body. It is associated with the prognosis of various diseases, such as malignant tumors, cardiovascular diseases, and autoimmune diseases. Although SII offers valuable information for diagnosing and predicting the risk of female breast cancer (FBC), the association between SII and FBC has not yet been analyzed. Therefore, the relationship between SII and FBC was investigated in this study. Multivariate logistic regression, model fit assessment using Akaike Information Criterion (AIC) and Bayesian Information Criterion (BIC), and smoothing curve fitting were applied to examine the correlation between SII and FBC using data from the National Health and Nutrition Examination Survey (NHANES) 2001-2018. Then the stability of their association was further examined using subgroup analysis and interaction tests among populations. Results showed a positive correlation between SII and FBC in 17,044 participants with age ≥ 20 years. In the fully adjusted model, every 100-unit increase in SII was accompanied by a 3% increased odds of FBC prevalence [OR = 1.03 (95% CI: 1.01, 1.05)]. Individuals in the highest quartile of SII exhibited 44% increased odds of FBC prevalence than those in the lowest quartile [OR = 1.44 (95% CI: 1.11, 1.88)]. Model fitness assessment using AIC and BIC criteria demonstrated that multivariable-adjusted models exhibited better fit compared to unadjusted models for both continuous and categorical SII specifications. Receiver Operating Characteristic (ROC) curve analysis demonstrated that SII exhibited excellent diagnostic capability for breast cancer, with the area under the ROC curve (AUC) of 0.816 (95% CI: 0.801-0.831), comparable to NLR (AUC = 0.816) and neutrophil counts (AUC = 0.815). In disease-specific performance comparison, SII's predictive ability for breast cancer (AUC = 0.816) was slightly superior to that for hypertension (AUC = 0.799), with the difference being statistically significant (P = 0.0407). Our findings confirmed that SII was a promising biomarker associated with FBC prevalence, and it may provide valuable insights into early screening and personalized treatment strategies.

The Conduct and Reporting of Meta-Analyses of Studies of Diagnostic Tests, and a Consideration of ROC Curves: Answers to the January 2010 Journal Club Questions

ObjectiveTo explore the validation of a disease-free survival (DFS) model for predicting disease progression based on the combination of ubiquitin-conjugating enzyme E2 C (UBE2C) levels and clinical indicators in breast cancer patients.MethodsWe enrolled 121 patients with breast cancer, collected their baseline characteristics and follow-up data, and analyzed the UBE2C levels in tumor tissues. We studied the relationship between UBE2C expression in tumor tissues and disease progression events of patients. We used the Kaplan-Meier method for identifying the disease-free survival rate of patients, and the multivariate Cox regression analysis to study the risk factors affecting the prognosis of patients. We sought to develop and validate a model for predicting disease progression.ResultsWe found that the level of expression of UBE2C could effectively distinguish the prognosis of patients. In the Receiver Operating Characteristic (ROC) curve analysis, the Area under the ROC Curve (AUC) = 0.826 (0.714–0.938) indicating that high levels of UBE2C was a high-risk factor for poor prognosis. After evaluating different models using the ROC curve, Concordance index (C-index), calibration curve, Net Reclassification Index (NRI), Integrated Discrimination Improvement Index (IDI), and other methods, we finally developed a model for the expression of Tumor-Node (TN) staging using Ki-67 and UBE2C, which had an AUC=0.870, 95% CI of 0.786–0.953. The traditional TN model had an AUC=0.717, and 95% CI of 0.581–0.853. Decision Curve Analysis (DCA) and Clinical Impact Curve (CIC) analysis indicated that the model had good clinical benefits and it was relatively simple to use.ConclusionWe found that high levels of UBE2C was a high-risk factor for poor prognosis. The use of UBE2C in addition to other breast cancer-related indicators effectively predicted the possible disease progression, thus providing a reliable basis for clinical decision-making.

Introduction and rationale: There is a wide-spread urgent need for an objective reproducible diagnostic tool capable of precisely classifying atypical mammary lesions into cancer vs. benign. Current diagnostic methods in pathology rely on routine and highly subjective histomorphologic and immunophenotypic features. Some authors report a 35% error rate in pathologic interpretation of atypical lesions of the breast (Elmore, 2015). Diagnostic variability can drop to the flip-of-the-coin 50%, even amongst expert breast pathologists (Tozbekian, 2017; Samples, 2017) and is largely attributed to the lack of innovative tools to adequately distinguish biologic differences (Allison, 2016), resulting in diagnoses that are described as “murky” and “arbitrary” (Khoury, 2022). Genomic imprinting, a vital epigenetic regulatory mechanism, is critical in mammalian development and tumorigenesis (Barlow, 2014, Murrell, 2006). Typically, imprinted genes exhibit parent-specific allele silencing through differential methylation. However, in cancer, this regulation often breaks down, resulting in the aberrant activation of the typically silenced allele—a phenomenon known as loss of imprinting (LOI) (Jelinic, 2016). LOI is considered one of the earliest molecular changes in cancer development, occurs early in tumorigenesis, making it ideal for identifying precursor lesions. Objectives: We hypothesize loss of imprinting (LOI) can be developed as a diagnostic, predictive and prognostic novel epigenetic marker (Quantitative Chromogenic Imprinted Gene In-Situ Hybridization, QCIGISH) for breast precancerous lesions and cancer detection. Our objective is to initiate a proof of concept for atypical breast lesions with QCIGISH and develop a diagnostic model. Methods: We retrospectively identified benign, atypical, and carcinoma core needle specimens from year 2023 in the Department of Pathology at Mayo Clinic Florida. Inclusion criteria included female patients, ages 18-99, with archived breast material obtained through standard of care (n=20). All diagnoses were ascertained by a breast pathologist (MKK). Following a thorough literature review, we processed the material for in situ hybridization, and pretreated with RNA scope preparation procedure using probes targeting noncoding intronic regions of nascent RNAs (QCIGISH). Based on high Area Under the Receiver Operating Characteristics Curve (AUROC), 2 genes - HM13 and GRB10 were identified as the most promising candidates for classification with QCIGISH. Upon test completion, the detected gene-expression site appeared as a distinct red dot on the slide under the common bright-field microscope. Next, the gene expression signals were counted and further classified as bi-allelic (BAE), multi-allelic (MAE) and total expression (TE). Results: A decision tree model based on BAE, MAE, and TE metrics showed differential signal expression. High MAE was a strong predictor for malignancy. We achieved 100% sensitivity and specificity with the ground truth, thereby initiating proof of concept. Furthermore, one histomorphologically challenging atypical lesion, for which the positive LOI finding accurately classified as ductal carcinoma in situ (DCIS), was confirmed indeed DCIS by ground truth validation on the excision specimen. Conclusion: We have demonstrated a breakthrough molecular biomarker that is fully objective, reproducible, and practical. First of its kind, the innovative QCIGISH has superior performance than the current standard of care in breast cancer diagnostics allowing for single-cell resolution of epigenetic changes. Next, we seek to undergo full model validation. Citation Format: M. Abdollahi Neisani, R. Guo, T. Cheng, M. K. Komforti. Epigenetic Imprinting as a Novel Precision Biomarker for Early Breast Cancer and Precancer Diagnosis and Detection [abstract]. In: Proceedings of the San Antonio Breast Cancer Symposium 2025; 2025 Dec 9-12; San Antonio, TX. Philadelphia (PA): AACR; Clin Cancer Res 2026;32(4 Suppl):Abstract nr PS2-07-09.

Acute variceal bleeding (AVB) in patients with cirrhosis is associated with high mortality, ranging from 12 to 20% at 6weeks. The existing prognostic models for AVB lack precision and require further validation. In this prospective study, we aimed to develop and validate a new prognostic model for AVB, and compared it with the existing models. We included 285 patients from March 2017 to November 2017 in the derivation cohort and 238 patients from December 2017 to June 2018 in the validation cohort. Two prognostic models were developed from derivation cohort by logistic regression analysis. Discrimination was assessed using area under the receiver operator characteristic curve (AUROC). The 6-week mortality was 22.1% in derivation cohort and 22.3% in validation cohort, P = 0.866. Model for end-stage liver disease (MELD) [odds ratio (OR) 1.106] and encephalopathy (E) (OR 4.658) in one analysis and Child-Pugh score (OR 1.379) and serum creatinine (OR 1.474) in another analysis were significantly associated with 6-week mortality. MELD-E model (AUROC 0.792) was superior to Child-creatinine model (AUROC) in terms of discrimination. The MELD-E model had highest AUROC; as compared to other models-MELD score (AUROC 0.751, P = 0.036), Child-Pugh score (AUROC 0.737, P = 0.037), D'Amico model (AUROC 0.716, P = 0.014) and Augustin model (AUROC 0.739, P = 0.018) in derivation cohort. In validation cohort, the discriminatory performance of MELD-E model (AUROC 0.805) was higher as compared to other models including MELD score (AUROC 0.771, P = 0.048), Child-Pugh score (AUROC 0.746, P = 0.011), Augustin model (AUROC 0.753, P = 0.039) and D'Amico model (AUROC 0.736, P = 0.021). In cirrhotic patients with AVB, the novel MELD-Encephalopathy model predicts 6weeks mortality with higher accuracy than the existing prognostic models.

Background: Sentinel lymph node (SLN) biopsy accurately stages the axilla, but is time consuming and resource intensive. Nomograms and scoring systems have been developed, based on clinical and pathologic data available before surgery, to attempt to predict the likelihood of lymph node metastasis before surgery. As the management of the axilla in patients with low nodal burden changes, it is also important to predict whether there will be further axillary disease in patients with a positive SLN. Aim: To explore the risk factors for SLN and non-SLN metastasis in Indian women with breast cancer, by analysis of clinical and pathologic data. To assess the validity and clinical utility of two MSKCC nomograms that predicts axillary lymph node status for Western patients. Methods: Clinical data, and pathologic data available from core biopsy, for a consecutive series of women having SLNB was analyzed, and was plotted on two MSKCC nomograms. Univariate analysis was done by χ2 and Fischer exact tests and multivariate analysis was done by logistic regression method. A receiver-operating characteristic (ROC) curve was drawn and predictive accuracy was assessed by calculating the area under the ROC curve (AUC). Results: 34% (89 out of 256) of our patients had SLN positivity. When correlated with SLN metastasis by univariate analysis, LVI (χ2 = 80, P ≤ 0.001), PNI (χ2 = 13.36, P ≤ 0.001), ER+ (χ2 = 6.85, P = 0.009), PR+ (χ2 = 7.1, P = 0.008) and age ( P = 0.03) were significant. However, multivariate analysis showed that age (OR=1.04, P = 0.007) and LVI (OR=0.07, P ≤ 0.001) were identified as independent predictors for SLN metastasis. The area under the ROC curve was 0.772 and it fairly correlated with MSKCC nomogram. Patients with MSKCC scores lower than 38% had a frequency of SLN metastasis of 7.7% (5/65) and this cut-off could be used as a guide for not doing frozen section analysis in this subgroup. Further axillary dissection showed 41% (38 out of 92) had non-sentinel nodes positive. When correlated with non-SLN metastasis by univariate analysis, LVI (χ2 = 8.8, P = 0.003), PNI (χ2 = 6.85, P = 0.009), and extracapsular extension (χ2 = 4.18, P = 0.04) were significant. Number of SLN negative ( P = 0.01), SLN ratio (number of SLN positive/total number of SLN removed) ( P = 0.01) and size of SLN metastasis ( P = 0.002) were significant. However, multivariate analysis showed that only size of SLN metastasis (OR=0.845, P = 0.02) was identified as independent predictor for non-SLN metastasis. The area under the ROC curve was 0.66 and it poorly correlated with MSKCC nomogram. Conclusion: The MSKCC nomogram can provide a fairly accurate prediction of the probability of SLN metastasis, but is not for non-SLN metastasis. An institutional nomogram for non-SLN metastasis, including additional factors such as size of SLN metastasis, may improve prediction.

e12594 Background: A proportion of triple negative breast cancer (TNBC) patients with pathologic complete response (pCR) will eventually experience disease recurrence. It is crucial to identify the few high-risk patients needing treatment intensification and those at risk of overtreatment. Methods: We prospectively isolated circulating tumor DNA (ctDNA) from pre/post-NAC plasma samples and then analyzed by duplex sequencing of our own designed panel combined driver and resistance-related genes between March 2016 to April 2022 in patients with TNBC. Kaplan-Meier analysis with a Log-rank test and the Cox proportional hazard regression model were used for the univariate and multivariable survival analysis. Survival prediction was assessed using a time-dependent receiver operating characteristic (ROC) curve and area under the ROC curve (AUC). A systematic literature search and meta-analysis was performed to further investigate the prognosis value of TP53 detection in ctDNA on breast cancer patients. Results: Before NAC, 53 of the 95 TNBC patients had detectable ctDNA (55.79%), which decreased after NAC (19/64, 29.69%). The median follow-up was 38.53 months (range 6.03-90.30 months). TP53 (36/95, 37.89%), BRCA2 (9/95, 9.47%) and PI3KCA (7/95, 7.37%) were the most frequently mutated genes pre-NAC. While post-NAC, TP53 (8/64, 12.50%), BRCA2 (5/64, 7.81%) and BRCA1 (4/64, 6.25%) were the most frequently mutated genes. Resistance gene mutations were almost undetectable. Pre/Post-NAC TP53 detection and its conversion status (no clearance after NAC vs. negative from baseline) were independent prognosis factors adjusted by age, tumor size and node status in event free survival (EFS) ( P =0.055, P =0.035, P =0.0008), distant disease-free survival (DDFS) ( P =0.017, P =0.017, P =0.003) and overall survival (OS) ( P =0.016, P =0.066, P =0.023). We made a contrast of the predictive efficiency of traditional clinical variables (age, tumor size and nodal status) combined pCR vs. TP53 conversion status through establishing time-dependent ROC curves. The predictive power were improved in TP53 conversion status combined clinical model in predicting EFS at 1 (AUC=0.82 vs 0.77), 3 (AUC=0.81 vs 0.76) and 5 years (AUC=0.69 vs 0.61), DDFS at 1 (AUC=0.79 vs 0.69), 3 (AUC=0.81 vs 0.76) and 5 years (AUC=0.70 vs 0.60) and OS at 1 (AUC=0.83 vs 0.88), 3 (AUC=0.86 vs 0.79) and 5 years (AUC=0.69 vs 0.64). The meta-analysis including six studies reported data of baseline TP53 detection and survival also demonstrated that plasma pre-NAC TP53 detection in ctDNA was significantly associated with a poor survival (HR = 2.42; P = 0.044). No enough studies were available for prognostic analysis of post-NAC and TP53 conversion status. Conclusions: Customized ctDNA tests are much less valuable than TP53 tests alone. Pre/Post-NAC TP53 detection and its conversion status were independent factor of shorter survival.

Purpose of study Methods of comparing the accuracy of diagnostic tests are of increasing necessity in biomedical science. When a test result is measured on a continuous scale, an assessment of the performance of the overall value of the test can be made using the Receiver Operating Characteristic (ROC) curve. This curve describes the discrimination ability of a diagnosis test in terms of diseased subjects from non-diseased subjects. The area under the ROC curve (AUC) describes the probability that a randomly chosen diseased subject will have higher probability of having disease than a randomly chosen non-diseased subject. For comparing two or more diagnostic test results, the difference between AUCs is often used. This paper proposes a non-parametric alternative method of comparing two or more correlated area under the curve (AUCs) of diagnostic tests for paired sample data. This method is based on Chi-square test statistic. Methods This paper investigated both parametric and non-parametric methods of comparing the equality of two AUCs and proposed a Chi-square test for the comparison of two or more diagnostic test processes. The proposed method does not require the knowledge of true status of subjects or gold standard in evaluating the accuracy of tests unlike other existing methods. The proposed method is most suitable for paired sample design. It also offers reliable statistical inferences even in small sample problems and circumvent the difficulties of deriving the statistical moments of complex summary statistics as seen in the Delong method. The proposed method provides for further analysis to determine the possible reason for rejecting the null hypothesis of equality of AUCs. Results The proposed method when applied on real data, avoids the lengthy and more difficult procedures of estimating the variances of two AUCs as a way of determining if two AUCs differ significantly. The method is validated using the Cochran Q test and was shown to compare favourably. The proposed method recommended for comparing two or more correlated AUCs when the data is paired. It is simple and does not require prior knowledge of true status of subjects unlike other existing methods. Keywords: Chi-square test, Cochran Q test, cut-off value, area under the curve, receiver operating characteristic, Dichotomous data DOI : 10.7176/JNSR/9-9-06 Publication date :May 31 st 2019

BackgroundIn classification and diagnostic testing, the receiver-operator characteristic (ROC) plot and the area under the ROC curve (AUC) describe how an adjustable threshold causes changes in two types of error: false positives and false negatives. Only part of the ROC curve and AUC are informative however when they are used with imbalanced data. Hence, alternatives to the AUC have been proposed, such as the partial AUC and the area under the precision-recall curve. However, these alternatives cannot be as fully interpreted as the AUC, in part because they ignore some information about actual negatives.MethodsWe derive and propose a new concordant partial AUC and a new partial c statistic for ROC data—as foundational measures and methods to help understand and explain parts of the ROC plot and AUC. Our partial measures are continuous and discrete versions of the same measure, are derived from the AUC and c statistic respectively, are validated as equal to each other, and validated as equal in summation to whole measures where expected. Our partial measures are tested for validity on a classic ROC example from Fawcett, a variation thereof, and two real-life benchmark data sets in breast cancer: the Wisconsin and Ljubljana data sets. Interpretation of an example is then provided.ResultsResults show the expected equalities between our new partial measures and the existing whole measures. The example interpretation illustrates the need for our newly derived partial measures.ConclusionsThe concordant partial area under the ROC curve was proposed and unlike previous partial measure alternatives, it maintains the characteristics of the AUC. The first partial c statistic for ROC plots was also proposed as an unbiased interpretation for part of an ROC curve. The expected equalities among and between our newly derived partial measures and their existing full measure counterparts are confirmed. These measures may be used with any data set but this paper focuses on imbalanced data with low prevalence.Future workFuture work with our proposed measures may: demonstrate their value for imbalanced data with high prevalence, compare them to other measures not based on areas; and combine them with other ROC measures and techniques.

Background: HER2+ breast cancers (BC) are clinically and biologically heterogeneous, with approximately half being ER+. Compared to other BC subtypes, HER2+ ER+ tumors display among the lowest rates of pathological complete response (pCR) following neoadjuvant chemotherapy (NACT) +/− anti-HER2 agents (anti-HER2). Yet in spite of this, HER2+/ER+ patients (pts) are typically treated with NACT plus anti-HER2, with the subsequent related toxicity. Currently there is a lack of predictive biomarkers that identify which subgroups of pts will not respond to such therapy. Inactivation of the Retinoblastoma (Rb) signalling pathway is a frequent event in BC. Previously developed gene-signatures of Rb loss-of-function have shown strong prognostic value and prediction of response to NACT. However, none has been extensively studied in the context of HER2+/ER+ BC. We have recently developed a gene-signature of RB-1 loss-of-function (RBsig) that is prognostic in luminal A-like and luminal B-like BC. Here we report the results of a retrospective in-silico study aimed to determine whether low expression of the RBsig in HER2+/ER+ BC correlates with a low pCR rate following NACT +/− anti-HER2. Methods: We performed a PubMed search for clinical trials of NACT +/− anti-HER2 (trastuzumab, lapatinib, or both) in HER2+ BC pts, and selected studies which had available gene expression data, hormone receptors status and pCR information. In-silico analyses of correlation between RBsig expression and pCR were performed using receiver-operating characteristic (ROC) curves and Fisher exact test to assess the prediction performance of the signature score. The threshold RBsig score was set at the 50th percentile of the score distribution. Results: Out of 16 identified studies, 10 fulfilled the inclusion criteria and were included in the analysis (514 pts). Overall, of the 211 HER2+/ER+ BC pts, 49 achieved pCR (23%); the pCR rate following NACT +/− anti-HER2 of pts with RBsig low expression was significantly lower compared to pts with RBsig high expression (16% vs 30%, respectively; Fisher exact test p=0.0098).The area under the ROC curve (AUC) was 0.62 (95% confidence interval (CI) 0.54-0.7, p=0.005). Results were similar for pts receiving NACT alone (94 pts; pCR rate 13% vs 28% in RBsig low vs RBsig high, respectively; Fisher exact test p=0,06; AUC 0.62, 95% CI 0.5-0.74, p=0.043) or combined with anti-HER2 (117 pts; pCR rate 18% vs 33% in RBsig low vs RBsig high, respectively; Fisher exact test p=0,049; AUC 0.61, 95% CI 0.5-0.72, p=0.041). In 303 HER2+/ ER− pts treated with NACT +/− anti-HER2, the pCR rate was 42%. No correlation was found between RBsig expression score and pCR rate in this group (pCR rate 42% vs 43% in RBsig low vs RBsig high, respectively; Fisher exact test p=0.53; AUC 0.5, 95% CI 0.43-0.56, p=0.973). Conclusions: RBsig identifies a subset of HER2+/ER+ pts with a low pCR rate following NACT +/− anti-HER2. We hypothesize that this signature has the potential to identify pts for whom chemotherapy could be avoided in favour of combinations of endocrine therapy and target therapies. Further refinement and validation in an independent dataset is warranted. Citation Format: Risi E, Grilli A, Migliaccio I, Biagioni C, Guarducci C, Bonechi M, Hart CD, Biganzoli L, Bicciato S, Di Leo A, Malorni L. A RB-1 loss-of-function gene-signature (RBsig) predicts resistance to neoadjuvant chemotherapy in HER2+/ER+ breast cancer patients [abstract]. In: Proceedings of the 2016 San Antonio Breast Cancer Symposium; 2016 Dec 6-10; San Antonio, TX. Philadelphia (PA): AACR; Cancer Res 2017;77(4 Suppl):Abstract nr P1-09-13.

To investigate the clinical utility of serial [⁶⁸Ga]Ga-FAPI-04 PET/CT for predicting pathological response and evaluating therapeutic efficacy in breast cancer (BC) patients receiving neoadjuvant chemotherapy (NAC). A total of 64 biopsy-confirmed patients with BC were retrospectively included. Serial [68Ga]Ga-FAPI-04 PET/CT scans were conducted at three time points: prior to NAC (baseline, PET1), after two cycles of NAC (interim, PET2), and before surgery (pre-surgery, PET3). PET/CT parameters derived from the primary BC lesions were recorded before and after NAC. The changes in these parameters were compared between the pathological complete response (pCR) and the non-pCR group. Logistic regression was used to assess the predictive value of [68Ga]Ga-FAPI-04 PET/CT parameters for predicting pCR. Receiver operating characteristic (ROC) curve analysis was employed to determine the optimal cutoff values for predicting pCR. DeLong's test was applied to statistically assess differences in the area under the ROC curves (AUC). Significant reductions in [68Ga]Ga-FAPI-04 PET/CT parameters (ΔSUVmax, ΔSUVmean, ΔFTV, ΔTLF) were observed among all patients, with significantly greater decreases in the pCR group compared to the non-pCR group (all P < 0.001). In the non-pCR group, the total FTV (ΔtFTV1-2 and ΔtFTV1-3) of primary breast lesions and metastatic lymph nodes showed moderate correlation with the residual cancer burden (RCB) score. Multivariate logistic regression identified ΔSUVmean1-2 (P = 0.027), ΔFTV1-2 (P = 0.006), ΔSUVmean1-3 (P = 0.032), and ΔFTV1-3 (P = 0.010) as independent predictors for predicting pCR. On the basis of the ROC curve analysis, ΔSUVmean1-3 (AUC = 0.848) and ΔFTV1-3 (AUC = 0.906) showed slightly higher predictive performance than ΔSUVmean1-2 (AUC = 0.825) and ΔFTV1-2 (AUC = 0.869), respectively, but the difference was not statistically significant (P > 0.05). This study demonstrates that serial [68Ga]Ga-FAPI-04 PET/CT facilitates early prediction of pathological response to neoadjuvant chemotherapy in breast cancer, as well as assessment of therapeutic efficacy. Prospective studies with larger samples are needed.

Previous studies have reported that microRNAs are implicated in the pathogenesis of diabetic nephropathy (DN). In this study, the underlying molecular mechanisms and diagnostic significance of miR-372-3p were investigated in the process of DN. Cell proliferation and apoptosis were measured using MTT and Annexin V-FITC double staining, respectively. RT-qPCR and western blotting were used to measure the expression levels of mRNA and protein. The diagnostic power of miR-372-3p in plasma for DN was evaluated using the receiver operating characteristics (ROC) curves and the area under the ROC curves (AUC). miR microarray analysis revealed that 126 miRs were significantly differentially expressed in response to high glucose stimulation. Among these miRs, high glucose stimulated miR-372-3p expression at the highest level. In vitro experimental measurements showed that knockdown of miR-372-3p showed the ability to reverse high glucose-induced glomerular endothelial cell apoptosis and impairment of eNOS/NO bioactivity. Mechanistic analysis revealed that fibroblast growth factor-16 (FGF-16) as a direct of miR-372-3p protected against high glucose-induced glomerular endothelial cell dysfunction. ROC analysis revealed that the diagnostic value of miR-372-3p, miR-15a or miR-372-3p combined with miR-15a in type 2 diabetes mellitus patients (AUC = 0.841, p < 0.001; AUC = 0.822, p < 0.001 or AUC = 0.922, p < 0.001) with DN was better than in type 1 diabetes mellitus patients (AUC = 0.805, p < 0.001; AUC = 0.722, p < 0.001 or AUC = 0.865, p < 0.001) with DN. miR-372-3p might be a valuable therapeutic target and diagnostic marker for patients with DN.

Genome-wide association studies in human populations have facilitated the creation of genomic profiles which combine the effects of many associated genetic variants to predict risk of disease. The area under the receiver operator characteristic (ROC) curve is a well established measure for determining the efficacy of tests in correctly classifying diseased and non-diseased individuals. We use quantitative genetics theory to provide insight into the genetic interpretation of the area under the ROC curve (AUC) when the test classifier is a predictor of genetic risk. Even when the proportion of genetic variance explained by the test is 100%, there is a maximum value for AUC that depends on the genetic epidemiology of the disease, i.e. either the sibling recurrence risk or heritability and disease prevalence. We derive an equation relating maximum AUC to heritability and disease prevalence. The expression can be reversed to calculate the proportion of genetic variance explained given AUC, disease prevalence, and heritability. We use published estimates of disease prevalence and sibling recurrence risk for 17 complex genetic diseases to calculate the proportion of genetic variance that a test must explain to achieve AUC = 0.75; this varied from 0.10 to 0.74. We provide a genetic interpretation of AUC for use with predictors of genetic risk based on genomic profiles. We provide a strategy to estimate proportion of genetic variance explained on the liability scale from estimates of AUC, disease prevalence, and heritability (or sibling recurrence risk) available as an online calculator.

Risk prediction models for postoperative mortality after intra-abdominal surgery have typically been developed using preoperative variables. It is unclear if intraoperative data add significant value to these risk prediction models. With IRB approval, an institutional retrospective cohort of intra-abdominal surgery patients in the 2005 to 2015 American College of Surgeons National Surgical Quality Improvement Program was identified. Intraoperative data were obtained from the electronic health record. The primary outcome was 30-day mortality. We evaluated the performance of machine learning algorithms to predict 30-day mortality using: 1) baseline variables and 2) baseline + intraoperative variables. Algorithms evaluated were: 1) logistic regression with elastic net selection, 2) random forest (RF), 3) gradient boosting machine (GBM), 4) support vector machine (SVM), and 5) convolutional neural networks (CNNs). Model performance was evaluated using the area under the receiver operator characteristic curve (AUROC). The sample was randomly divided into a training/testing split with 80%/20% probabilities. Repeated 10-fold cross-validation identified the optimal model hyperparameters in the training dataset for each model, which were then applied to the entire training dataset to train the model. Trained models were applied to the test cohort to evaluate model performance. Statistical significance was evaluated using P < .05. The training and testing cohorts contained 4322 and 1079 patients, respectively, with 62 (1.4%) and 15 (1.4%) experiencing 30-day mortality, respectively. When using only baseline variables to predict mortality, all algorithms except SVM (area under the receiver operator characteristic curve [AUROC], 0.83 [95% confidence interval {CI}, 0.69-0.97]) had AUROC >0.9: GBM (AUROC, 0.96 [0.94-1.0]), RF (AUROC, 0.96 [0.92-1.0]), CNN (AUROC, 0.96 [0.92-0.99]), and logistic regression (AUROC, 0.95 [0.91-0.99]). AUROC significantly increased with intraoperative variables with CNN (AUROC, 0.97 [0.96-0.99]; P = .047 versus baseline), but there was no improvement with GBM (AUROC, 0.97 [0.95-0.99]; P = .3 versus baseline), RF (AUROC, 0.96 [0.93-1.0]; P = .5 versus baseline), and logistic regression (AUROC, 0.94 [0.90-0.99]; P = .6 versus baseline). Postoperative mortality is predicted with excellent discrimination in intra-abdominal surgery patients using only preoperative variables in various machine learning algorithms. The addition of intraoperative data to preoperative data also resulted in models with excellent discrimination, but model performance did not improve.

e15026 Background: Heterogeneity of 18F-fluorodeoxyglucose (FDG) uptake is a promising marker for predicting response to treatment. This study aimed to evaluate the ability of pretreatment positron emission tomography/computed tomography (PET/CT) 18F-FDG-based heterogeneity to predict the response to pyrotinib in patients with human epidermal growth factor receptor 2 (HER2)-positive metastatic breast cancer (MBC). Methods: Patients with MBC in the Fudan University Shanghai Cancer Center who underwent whole-body 18F-FDG PET/CT before the initiation of pyrotinib was included. The intertumoral and intratumoral heterogeneity indexes (HI-inter and HI-intra, respectively), maximum standardized uptake value (SUVmax), total lesion glycolysis (TLG), and metabolic tumor volume (MTV) on the baseline PET/CT were assessed. Progression-free survival (PFS) was estimated by the Kaplan-Meier method and compared by log-rank test. Time-dependent receiver operating characteristic (ROC) curve analysis was performed, and the predictive accuracies of all markers were evaluated by plotting the cumulative area under the ROC curve (AUC) over time. Results: A total of 22 patients were included in this study. The median PFS of patients with a high HI-intra (&gt; 1.9) was 6.6 months, whereas that of patients with a low HI-intra was 13.4 months (p = 0.044). The HI-inter was able to discriminate patients as well as the coefficient of variance. Univariate analysis showed that patients with a higher HI-inter tended to have worse PFS (10.6 months vs. 13.4 months, p = 0.067). Higher SUVmax and TLG were also associated with worse PFS. ROC curve analysis confirmed the predictive value of the HI-inter and HI-intra. TLG had the highest accuracy in predicting PFS (AUC = 0.87), followed by HI-inter (AUC = 0.86), SUVmax (AUC = 0.85), HI-intra (AUC = 0.80), mean standardized uptake value (AUC = 0.63), and MTV (AUC = 0.60). Conclusions: Intratumoral and intertumoral heterogeneities in metastatic lesions on pretreatment 18F-FDG PET/CT could predict response to pyrotinib treatment in patients with HER2-positive breast cancer, which could provide information to guide treatment decisions.