The biological subtype of breast cancer influences the selection of systemic therapy. Distinction between luminal A and B cancers depends on consistent assessment of Ki-67, but substantial intra-observer and inter-observer variability exists when immunohistochemistry (IHC) is used. We compared RT-qPCR with IHC in the assessment of Ki-67 and other standard factors used in breast cancer subtyping. RNA was extracted from archival breast tumour tissue of 769 women randomly assigned to the FinHer trial. Cancer ESR1, PGR, ERBB2 and MKI67 mRNA content was quantitated with an RT-qPCR assay. Local pathologists assessed ER, PgR and Ki-67 expression using IHC. HER2 amplification was identified with chromogenic in situ hybridization (CISH) centrally. The results were correlated with distant disease-free survival (DDFS) and overall survival (OS). qPCR-based and IHC-based assessments of ER and PgR showed good concordance. Both low tumour MKI67 mRNA (RT-qPCR) and Ki-67 protein (IHC) levels were prognostic for favourable DDFS [hazard ratio (HR) 0.42, 95 % CI 0.25-0.71, P = 0.001; and HR 0.56, 0.37-0.84, P = 0.005, respectively] and OS. In multivariable analyses, cancer MKI67 mRNA content had independent influence on DDFS (adjusted HR 0.51, 95 % CI 0.29-0.89, P = 0.019) while Ki-67 protein expression had not any influence (P = 0.266) whereas both assessments influenced independently OS. Luminal B patients treated with docetaxel-FEC had more favourable DDFS and OS than those treated with vinorelbine-FEC when the subtype was defined by RT-qPCR (for DDFS, HR 0.52, 95 % CI 0.29-0.94, P = 0.031), but not when defined using IHC. Breast cancer subtypes approximated with RT-qPCR and IHC show good concordance, but cancer MKI67 mRNA content correlated slightly better with DDFS than Ki-67 expression. The findings based on MKI67 mRNA content suggest that patients with luminal B cancer benefit more from docetaxel-FEC than from vinorelbine-FEC.

Aims/hypothesis In this study, we aimed to characterise rapid progressors to type 1 diabetes among children recruited from the general population, on the basis of HLA-conferred disease susceptibility. Methods We monitored 7410 HLA-predisposed children participating in the Finnish Type 1 Diabetes Prediction and Prevention (DIPP) study for the development of beta cell autoimmunity and type 1 diabetes from birth over a median follow-up time of 16.2 years (range 0.9-21.1 years). Islet cell antibodies (ICA) and autoantibodies to insulin (IAA), GAD (GADA) and islet antigen 2 (IA-2A) were assessed as markers of beta cell autoimmunity. Rapid progression was defined as progression to clinical type 1 diabetes within 1.5 years of autoantibody seroconversion. We analysed the association between rapid progression and demographic and autoantibody characteristics as well as genetic markers, including 25 non-HLA SNPs predisposing to type 1 diabetes. Results Altogether, 1550 children (21%) tested positive for at least one diabetes-associated autoantibody in at least two samples, and 248 (16%) of seroconverters progressed to type 1 diabetes by the end of 2015. The median time from seroconversion to diagnosis was 0.51 years in rapid progressors (n = 42, 17%) and 5.4 years in slower progressors. Rapid progression was observed both among young (<5 years) and early pubertal children (> 7 years), resulting in a double-peak distribution of seroconversion age. Compared with slower progressors, rapid progressors had a higher frequency of positivity for multiple (>= 2) autoantibodies and had higher titres of ICA, IAA and IA-2A at seroconversion, and there was a higher prevalence of the secretor genotype in the FUT2 gene among those carrying the high-risk HLA genotype. Compared with autoantibody-positive non-progressors, rapid progressors were younger, were more likely to carry the high-risk HLA genotype and a predisposing SNP in the PTPN22 gene, had higher frequency of ICA, IAA, GADA and IA-2A positivity and multipositivity, and had higher titres of all four autoantibodies at seroconversion. Conclusions/interpretation At seroconversion, individuals with rapid progression to type 1 diabetes were characterised by a younger age, higher autoantibody titres, positivity for multiple autoantibodies and higher prevalence of a FUT2 SNP. The double-peak profile for seroconversion age among the rapid progressors demonstrates for the first time that rapid progression may take place not only in young children but also in children in early puberty. Rapid progressors might benefit from careful clinical follow-up and early preventive measures.

Background: The proper estimate of the risk of recurrences in early-stage oral tongue squamous cell carcinoma (OTSCC) is mandatory for individual treatment-decision making. However, this remains a challenge even for experienced multidisciplinary centers. Objectives: We compared the performance of four machine learning (ML) algorithms for predicting the risk of locoregional recurrences in patients with OTSCC. These algorithms were Support Vector Machine (SVM), Naive Bayes (NB), Boosted Decision Tree (BDT), and Decision Forest (DF). Materials and methods: The study cohort comprised 311 cases from the five University Hospitals in Finland and A.C. Camargo Cancer Center, Sao Paulo, Brazil. For comparison of the algorithms, we used the harmonic mean of precision and recall called F1 score, specificity, and accuracy values. These algorithms and their corresponding permutation feature importance (PFI) with the input parameters were externally tested on 59 new cases. Furthermore, we compared the performance of the algorithm that showed the highest prediction accuracy with the prognostic significance of depth of invasion (DOI). Results: The results showed that the average specificity of all the algorithms was 71% The SVM showed an accuracy of 68% and F1 score of 0.63, NB an accuracy of 70% and F1 score of 0.64, BDT an accuracy of 81% and F1 score of 0.78, and DF an accuracy of 78% and F1 score of 0.70. Additionally, these algorithms outperformed the DOI-based approach, which gave an accuracy of 63%. With PFI-analysis, there was no significant difference in the overall accuracies of three of the algorithms; PFI-BDT accuracy increased to 83.1%, PFI-DF increased to 80%, PFI-SVM decreased to 64.4%, while PFI-NB accuracy increased significantly to 81.4%. Conclusions: Our findings show that the best classification accuracy was achieved with the boosted decision tree algorithm. Additionally, these algorithms outperformed the DOI-based approach. Furthermore, with few parameters identified in the PFI analysis, ML technique still showed the ability to predict locoregional recurrence. The application of boosted decision tree machine learning algorithm can stratify OTSCC patients and thus aid in their individual treatment planning.

BackgroundMetabolomic changes during pregnancy have been suggested to underlie the etiology of gestational diabetes mellitus (GDM). However, research on metabolites during preconception is lacking. Therefore, this study aimed to investigate distinctive metabolites during the preconception phase between GDM and non-GDM controls in a nested case-control study in Singapore.MethodsWithin a Singapore preconception cohort, we included 33 Chinese pregnant women diagnosed with GDM according to the IADPSG criteria between 24 and 28 weeks of gestation. We then matched them with 33 non-GDM Chinese women by age and pre-pregnancy body mass index (ppBMI) within the same cohort. We performed a non-targeted metabolomics approach using fasting serum samples collected within 12 months prior to conception. We used generalized linear mixed model to identify metabolites associated with GDM at preconception after adjusting for maternal age and ppBMI. After annotation and multiple testing, we explored the additional predictive value of novel signatures of preconception metabolites in terms of GDM diagnosis.ResultsA total of 57 metabolites were significantly associated with GDM, and eight phosphatidylethanolamines were annotated using HMDB. After multiple testing corrections and sensitivity analysis, phosphatidylethanolamines 36:4 (mean difference beta: 0.07; 95% CI: 0.02, 0.11) and 38:6 (beta: 0.06; 0.004, 0.11) remained significantly higher in GDM subjects, compared with non-GDM controls. With all preconception signals of phosphatidylethanolamines in addition to traditional risk factors (e.g., maternal age and ppBMI), the predictive value measured by area under the curve (AUC) increased from 0.620 to 0.843.ConclusionsOur data identified distinctive signatures of GDM-associated preconception phosphatidylethanolamines, which is of potential value to understand the etiology of GDM as early as in the preconception phase. Future studies with larger sample sizes among alternative populations are warranted to validate the associations of these signatures of metabolites and their predictive value in GDM.

Abstract Background The aim was to develop prediction rules that may guide early treatment decisions based on baseline clinical predictors of long-term unfavorable outcome in juvenile idiopathic arthritis (JIA). Methods In the Nordic JIA cohort, we assessed baseline disease characteristics as predictors of the following outcomes 8 years after disease onset. Non-achievement of remission off medication according to the preliminary Wallace criteria, functional disability assessed by Childhood Health Assessment Questionnaire (CHAQ) and Physical Summary Score (PhS) of the Child Health Questionnaire, and articular damage assessed by the Juvenile Arthritis Damage Index-Articular (JADI-A). Multivariable models were constructed, and cross-validations were performed by repeated partitioning of the cohort into training sets for developing prediction models and validation sets to test predictive ability. Results The total cohort constituted 423 children. Remission status was available in 410 children: 244 (59.5%) of these did not achieve remission off medication at the final study visit. Functional disability was present in 111/340 (32.7%) children assessed by CHAQ and 40/199 (20.1%) by PhS, and joint damage was found in 29/216 (13.4%). Model performance was acceptable for making predictions of long-term outcome. In validation sets, the area under the curves (AUCs) in the receiver operating characteristic (ROC) curves were 0.78 (IQR 0.72–0.82) for non-achievement of remission off medication, 0.73 (IQR 0.67–0.76) for functional disability assessed by CHAQ, 0.74 (IQR 0.65–0.80) for functional disability assessed by PhS, and 0.73 (IQR 0.63–0.76) for joint damage using JADI-A. Conclusion The feasibility of making long-term predictions of JIA outcome based on early clinical assessment is demonstrated. The prediction models have acceptable precision and require only readily available baseline variables. Further testing in other cohorts is warranted.

BackgroundWe developed multifactorial models for predicting incident dementia and brain pathology in the oldest old using the Vantaa 85+ cohort.MethodsWe included participants without dementia at baseline and at least 2 years of follow-up (N=245) for dementia prediction or with autopsy data (N=163) for pathology. A supervised machine learning method was used for model development, considering sociodemographic, cognitive, clinical, vascular, and lifestyle factors, as well as APOE genotype. Neuropathological assessments included -amyloid, neurofibrillary tangles and neuritic plaques, cerebral amyloid angiopathy (CAA), macro- and microscopic infarcts, -synuclein pathology, hippocampal sclerosis, and TDP-43.ResultsPrediction model performance was evaluated using AUC for 10x10-fold cross-validation. Overall AUCs were 0.73 for dementia, 0.64-0.68 for Alzheimer's disease (AD)- or amyloid-related pathologies, 0.72 for macroinfarcts, and 0.61 for microinfarcts. Predictors for dementia were different from those in previous reports of younger populations; for example, age, sex, and vascular and lifestyle factors were not predictive. Predictors for dementia versus pathology were also different, because cognition and education predicted dementia but not AD- or amyloid-related pathologies. APOE genotype was most consistently present across all models. APOE alleles had a different impact: epsilon 4 did not predict dementia, but it did predict all AD- or amyloid-related pathologies; epsilon 2 predicted dementia, but it was protective against amyloid and neuropathological AD; and epsilon 3 epsilon 3 was protective against dementia, neurofibrillary tangles, and CAA. Very few other factors were predictive of pathology.ConclusionsDifferences between predictors for dementia in younger old versus oldest old populations, as well as for dementia versus pathology, should be considered more carefully in future studies.

Objectives: Maternal serum inhibin-A , pregnancy associated plasma protein-A (PAPP-A) and PAPP-A2 together with placental growth factor (PlGF), maternal risk factors and uterine arter y pulsatility inde x (UtA PI) were analysed to study thei r ability to predict pre-eclampsia (PE). Study design: Serial serum samples for the nested case-control study were collected prospectively at 12-14, 18-20 and 26-28 weeks of gestation from 11 women who later developed early-onset PE (EO PE , diagnosis < 34 + 0 weeks of gestation), 34 women who developed late-onset PE (LO PE , diagnosis 2 34 + 0 weeks) and 89 controls. Main outcome measures: Gestational age-adjusted multiples of the median (MoM) values were calculated for biomarker concentrations. Multivariate regression analyses were performed to combine first trimester bio-markers, previously reported results on PlGF, maternal risk factors and UtA PI. Area under cu r v e (AUC) values and 95% confidence intervals (CIs) for the prediction of PE and its subtypes were calculated . Results: A high first trimester inhibin-A predicted PE (AUC 0.618, 95%CI, 0.513-0.724), whereas PAPP-A and PlGF predicted only EO PE (0.701, 0.562-0.840 and 0.798, 0.686-0.909, respectively). At 26-28 weeks PAPP-A2 and inhibin-A predicted a l l PE subtypes. In the multivariate setting inhibin-A combined with maternal pre-pregnancy body mass index, prior PE and mean UtA PI predicted PE (0.811,0.726-0.896) and LO PE (0.824, 0.733-0.914). Conclusions: At first trimester inhibin-A show potential ability to predict not only EO PE but also LO PE whereas PlGF and PAPP-A predict only EO PE. At late second trimester inhibin-A and PAPP-A2 might be usef u l for short-term prediction of PE.

The fullPIERS model is a risk prediction model developed to predict adverse maternal outcomes within 48 h for women admitted with pre-eclampsia. External validation of the model is required before implementation for clinical use. We assessed the temporal and external validity of the fullPIERS model in high income settings using five cohorts collected between 2003 and 2016, from tertiary hospitals in Canada, the United States of America, Finland and the United Kingdom. The cohorts were grouped into three datasets for assessing the primary external, and temporal validity, and broader transportability of the model. The predicted risks of developing an adverse maternal outcome were calculated using the model equation and model performance was evaluated based on discrimination, calibration, and stratification. Our study included a total of 2429 women, with an adverse maternal outcome rate of 6.7%, 6.6%, and 7.0% in the primary external, temporal, and combined (broader) validation cohorts, respectively. The model had good discrimination in all datasets: 0.81 (95%CI 0.75-0.86), 0.82 (95%CI 0.76-0.87), and 0.75 (95%CI 0.71-0.80) for the primary external, temporal, and broader validation datasets, respectively. Calibration was best for the temporal cohort but poor in the broader validation dataset The likelihood ratios estimated to rule in adverse maternal outcomes were high at a cut-off of >= 30% in all datasets. The fullPIERS model is temporally and externally valid and will be useful in the management of women with pre-eclampsia in high income settings although model recalibration is required to improve performance, specifically in the broader healthcare settings.

Background: Models to predict disease course and long-term outcome based on clinical characteristics at disease onset may guide early treatment strategies in juvenile idiopathic arthritis (JIA). Before a prediction model can be recommended for use in clinical practice, it needs to be validated in a different cohort than the one used for building the model. The aim of the current study was to validate the predictive performance of the Canadian prediction model developed by Guzman et al. and the Nordic model derived from Rypdal et al. to predict severe disease course and non-achievement of remission in Nordic patients with JIA. Methods: The Canadian and Nordic multivariable logistic regression models were evaluated in the Nordic JIA cohort for prediction of non-achievement of remission, and the data-driven outcome denoted severe disease course. A total of 440 patients in the Nordic cohort with a baseline visit and an 8-year visit were included. The Canadian prediction model was first externally validated exactly as published. Both the Nordic and Canadian models were subsequently evaluated with repeated fine-tuning of model coefficients in training sets and testing in disjoint validation sets. The predictive performances of the models were assessed with receiver operating characteristic curves and C-indices. A model with a C-index above 0.7 was considered useful for clinical prediction. Results: The Canadian prediction model had excellent predictive ability and was comparable in performance to the Nordic model in predicting severe disease course in the Nordic JIA cohort. The Canadian model yielded a C-index of 0.85 (IQR 0.83-0.87) for prediction of severe disease course and a C-index of 0.66 (0.63-0.68) for prediction of non-achievement of remission when applied directly. The median C-indices after fine-tuning were 0.85 (0.80-0.89) and 0.69 (0.65-0.73), respectively. Internal validation of the Nordic model for prediction of severe disease course resulted in a median C-index of 0.90 (0.86-0.92). Conclusions: External validation of the Canadian model and internal validation of the Nordic model with severe disease course as outcome confirm their predictive abilities. Our findings suggest that predicting long-term remission is more challenging than predicting severe disease course.