A machine learning-based algorithm to identify U-500R insulin candidates among adults with type 2 diabetes mellitus in US retrospective databases

OBJECTIVE

To develop a machine learning-based predictive algorithm to identify patients with type 2 diabetes mellitus (T2DM) who are candidates for initiation of U-500R insulin (U-500R).

METHODS

A retrospective cohort of patients with T2DM was used from a large US administrative claims and electronic health records (EHR) database affiliated with Optum. Predictor variables derived from the data were used to identify appropriate supervised machine learning models including least absolute shrinkage and selection operator (LASSO) and extreme gradient boosted (XGBoost) methods. Predictive performance was assessed using precision-recall (PR) and receiver operating characteristic (ROC) area under the curve (AUC). The clinical interpretation of the final model was supported by fitting the final set of variables from the LASSO and XGBoost models to a traditional logistic regression model. Model choice was determined by comparing Akaike Information Criterion (AIC), residual deviances, and scaled Brier scores.

RESULTS

Among 81,242 patients who met the study eligibility criteria, 577 initiated U-500R and were assigned to the positive class. Predictors of U-500R initiation included overweight/obesity, neuropathy, HbA1c ≥9% and 8%-9%, BUN 23.8 to <112 mg/dl, ALT 35.9-2056.2 U/L, no radiological chest exams, no GFR labs, and gait/mobility abnormalities. The best performing model was the LASSO model with an ROC AUC of 0.776 on the hold-out test set.

CONCLUSION

This study successfully developed and validated a machine learning-based algorithm to identify U-500R candidates among patients with T2DM. This may help health care providers and decision-makers to understand important characteristics of patients who could use U-500R therapies which in turn could support policies and guidelines for optimal patient management.

The severity assessment of Parkinson's disease based on plasma inflammatory factors and third ventricle width by transcranial sonography

BACKGROUND

Predicting Parkinson's disease (PD) can provide patients with targeted therapies. However, disease severity can be roughly evaluated in clinical practice based on the patient's symptoms and signs.

OBJECTIVE

The current study attempted to explore the factors linked with PD severity and construct a predictive model.

METHOD

The PD patients and healthy controls were recruited from our study center while recording their basic demographic information. The serum inflammatory markers levels, such as Cystatin C (Cys C), C-reactive protein (CRP), RANTES (regulated on activation, normal T cell expressed and secreted), Interleukin-10 (IL-10), and Interleukin-6 (IL-6) were determined for all the participants. PD patients were categorized into early and mid-advanced groups based on the Hoehn and Yahr (H-Y) scale and evaluated using PD-related scales. LASSO logistic regression analysis (Model C) helped select variables based on clinical scale evaluations, serum inflammatory factor levels, and transcranial sonography measurements. The optimal harmonious model coefficient λ was determined via 10-fold cross-validation. Moreover, Model C was compared with multivariate (Model A) and stepwise (Model B) logistic regression. The area under the curve (AUC) of a receiver operator characteristic (ROC), brier score, calibration curve, and decision curve analysis (DCA) helped determine the discrimination and calibration of the predictive model, followed by configuring a forest plot and column chart.

RESULTS

The study included 113 healthy individuals and 102 PD patients, with 26 early and 76 mid-advanced patients. Univariate analysis of variance screened out statistically significant differences among inflammatory markers Cys C and RANTES. The average Cys C level in the mid-advanced stage was significantly higher than in the early stage (p < 0.001) but not for RANTES (p = 0.740). The LASSO logistic regression model (λ.1se = 0.061) associated with UPDRS-I, UPDRS-II, UPDRS-III, HAMA, PDQ-39, and Cys C as the included independent variables revealed that the Model C discrimination and calibration (AUC = 0.968, Brier = 0.049) were superior to Model A (AUC = 0.926, Brier = 0.079) and Model B (AUC = 0.929, Brier = 0.071) models.

CONCLUSION

The study results show multiple factors are linked with PD assessment. Moreover, the inflammatory marker Cys C and transcranial sonography measurement could objectively predict PD symptom severity, helping doctors monitor PD evolution in patients while targeting interventions.

Investigating unique genes of five molecular subtypes of breast cancer using penalized logistic regression

Background

Breast cancer (BC) is the most common cancer and the fifth cause of death in women worldwide. Exploring unique genes for cancers has been interesting.

Patients and Methods

This study aimed to explore unique genes of five molecular subtypes of BC in women using penalized logistic regression models. For this purpose, microarray data of five independent GEO data sets were combined. This combination includes genetic information of 324 women with BC and 12 healthy women. Least absolute shrinkage and selection operator (LASSO) logistic regression and adaptive LASSO logistic regression were used to extract unique genes. The biological process of extracted genes was evaluated in an open-source GOnet web application. R software version 3.6.0 with the glmnet package was used for fitting the models.

Results

Totally, 119 genes were extracted among 15 pairwise comparisons. Seventeen genes (14%) showed overlap between comparative groups. According to GO enrichment analysis, the biological process of extracted genes was enriched in negative and positive regulation biological processes, and molecular function tracking revealed that most genes are involved in kinase and transferring activities. On the other hand, we identified unique genes for each comparative group and the subsequent pathways for them. However, a significant pathway was not identified for genes in normal-like versus ERBB2 and luminal A, basal versus control, and lumina B versus luminal A groups.

Conclusion

Most genes selected by LASSO logistic regression and adaptive LASSO logistic regression identified unique genes and related pathways for comparative subgroups of BC, which would be useful to comprehend the molecular differences between subgroups that would be considered for further research and therapeutic approaches in the future.

A Nomogram for Predicting Prognosis of Advanced Schistosomiasis japonica in Dongzhi County-A Case Study

BACKGROUNDS

Advanced schistosomiasis is the late stage of schistosomiasis, seriously jeopardizing the quality of life or lifetime of infected people. This study aimed to develop a nomogram for predicting mortality of patients with advanced schistosomiasis japonica, taking Dongzhi County of China as a case study.

METHOD

Data of patients with advanced schistosomiasis japonica were collected from Dongzhi Schistosomiasis Hospital from January 2019 to July 2022. Data of patients were randomly divided into a training set and validation set with a ratio of 7:3. Candidate variables, including survival outcomes, demographics, clinical features, laboratory examinations, and ultrasound examinations, were analyzed and selected by LASSO logistic regression for the nomogram. The performance of the nomogram was assessed by concordance index (C-index), sensitivity, specificity, positive predictive value (PPV) and negative predictive value (NPV). The calibration of the nomogram was evaluated by the calibration plots, while clinical benefit was evaluated by decision curve and clinical impact curve analysis.

RESULTS

A total of 628 patients were included in the final analysis. Atrophy of the right liver, creatinine, ascites level III, N-terminal procollagen III peptide, and high-density lipoprotein were selected as parameters for the nomogram model. The C-index, sensitivity, specificity, PPV, and NPV of the nomogram were 0.97 (95% [CI]: [0.95-0.99]), 0.78 (95% [CI]: [0.64-0.87]), 0.97 (95% [CI]: [0.94-0.98]), 0.78 (95% [CI]: [0.64-0.87]), 0.97 (95% [CI]: [0.94-0.98]) in the training set; and 0.98 (95% [CI]: [0.94-0.99]), 0.86 (95% [CI]: [0.64-0.96]), 0.97 (95% [CI]: [0.93-0.99]), 0.79 (95% [CI]: [0.57-0.92]), 0.98 (95% [CI]: [0.94-0.99]) in the validation set, respectively. The calibration curves showed that the model fitted well between the prediction and actual observation in both the training set and validation set. The decision and the clinical impact curves showed that the nomogram had good clinical use for discriminating patients with high risk of death.

CONCLUSIONS

A nomogram was developed to predict prognosis of advanced schistosomiasis. It could guide clinical staff or policy makers to formulate intervention strategies or efficiently allocate resources against advanced schistosomiasis.

Establishment and validation of a clinical model for diagnosing solitary pulmonary nodules

OBJECTIVES

The main purpose of this study was to develop and validate a clinical model for estimating the risk of malignancy in solitary pulmonary nodules (SPNs).

METHODS

A total of 672 patients with SPNs were retrospectively reviewed. The least absolute shrinkage and selection operator algorithm was applied for variable selection. A regression model was then constructed with the identified predictors. The discrimination, calibration, and clinical validity of the model were evaluated by the area under the receiver-operating-characteristic curve (AUC), calibration curve, and decision curve analysis (DCA).

RESULTS

Ten predictors, including gender, age, nodule type, diameter, lobulation sign, calcification, vascular convergence sign, mediastinal lymphadenectasis, the natural logarithm of carcinoembryonic antigen, and combination of cytokeratin 19 fragment 21-1, were incorporated into the model. The prediction model demonstrated valuable prediction performance with an AUC of 0.836 (95% CI: 0.777-0.896), outperforming the Mayo (0.747, p = 0.024) and PKUPH (0.749, p = 0.018) models. The model was well-calibrated according to the calibration curves. The DCA indicated the nomogram was clinically useful over a wide range of threshold probabilities.

CONCLUSION

This study proposed a clinical model for estimating the risk of malignancy in SPNs, which may assist clinicians in identifying the pulmonary nodules that require invasive procedures and avoid the occurrence of overtreatment.

The Comprehensive Machine Learning Analytics for Heart Failure

Background: Early detection of heart failure is the basis for better medical treatment and prognosis. Over the last decades, both prevalence and incidence rates of heart failure have increased worldwide, resulting in a significant global public health issue. However, an early diagnosis is not an easy task because symptoms of heart failure are usually non-specific. Therefore, this study aims to develop a risk prediction model for incident heart failure through a machine learning-based predictive model. Although African Americans have a higher risk of incident heart failure among all populations, few studies have developed a heart failure risk prediction model for African Americans. Methods: This research implemented the Least Absolute Shrinkage and Selection Operator (LASSO) logistic regression, support vector machine, random forest, and Extreme Gradient Boosting (XGBoost) to establish the Jackson Heart Study's predictive model. In the analysis of real data, missing data are problematic when building a predictive model. Here, we evaluate predictors' inclusion with various missing rates and different missing imputation strategies to discover the optimal analytics. Results: According to hundreds of models that we examined, the best predictive model was the XGBoost that included variables with a missing rate of less than 30 percent, and we imputed missing values by non-parametric random forest imputation. The optimal XGBoost machine demonstrated an Area Under Curve (AUC) of 0.8409 to predict heart failure for the Jackson Heart Study. Conclusion: This research identifies variations of diabetes medication as the most crucial risk factor for heart failure compared to the complete cases approach that failed to discover this phenomenon.

Shorter androgen receptor polyQ alleles protect against life-threatening COVID-19 disease in European males

Background

While SARS-CoV-2 similarly infects men and women, COVID-19 outcome is less favorable in men. Variability in COVID-19 severity may be explained by differences in the host genome.

Methods

We compared poly-amino acids variability from WES data in severely affected COVID-19 patients versus SARS-CoV-2 PCR-positive oligo-asymptomatic subjects.

Findings

Shorter polyQ alleles (≤22) in the androgen receptor (AR) conferred protection against severe outcome in COVID-19 in the first tested cohort (both males and females) of 638 Italian subjects. The association between long polyQ alleles (≥23) and severe clinical outcome (p = 0.024) was also validated in an independent cohort of Spanish men <60 years of age (p = 0.014). Testosterone was higher in subjects with AR long-polyQ, possibly indicating receptor resistance (p = 0.042 Mann-Whitney U test). Inappropriately low serum testosterone level among carriers of the long-polyQ alleles (p = 0.0004 Mann-Whitney U test) predicted the need for intensive care in COVID-19 infected men. In agreement with the known anti-inflammatory action of testosterone, patients with long-polyQ and age ≥60 years had increased levels of CRP (p = 0.018, not accounting for multiple testing).

Interpretation

We identify the first genetic polymorphism that appears to predispose some men to develop more severe disease. Failure of the endocrine feedback to overcome AR signaling defects by increasing testosterone levels during the infection leads to the polyQ tract becoming dominant to serum testosterone levels for the clinical outcome. These results may contribute to designing reliable clinical and public health measures and provide a rationale to test testosterone as adjuvant therapy in men with COVID-19 expressing long AR polyQ repeats.

Funding

MIUR project “Dipartimenti di Eccellenza 2018-2020” to Department of Medical Biotechnologies University of Siena, Italy (Italian D.L. n.18 March 17, 2020) and “Bando Ricerca COVID-19 Toscana” project to Azienda Ospedaliero-Universitaria Senese. Private donors for COVID-19 research and charity funds from Intesa San Paolo.

Pancreatic ductal adenocarcinoma can be detected by analysis of volatile organic compounds (VOCs) in alveolar air. BMC Cancer. 2018;18:529. [PMID: 29728093 PMCID: PMC5935919 DOI: 10.1186/s12885-018-4452-0]

Background

In the last decade many studies showed that the exhaled breath of subjects suffering from several pathological conditions has a peculiar volatile organic compound (VOC) profile. The objective of the present work was to analyse the VOCs in alveolar air to build a diagnostic tool able to identify the presence of pancreatic ductal adenocarcinoma in patients with histologically confirmed disease.

Methods

The concentration of 92 compounds was measured in the end-tidal breath of 65 cases and 102 controls. VOCs were measured with an ion-molecule reaction mass spectrometry. To distinguish between subjects with pancreatic adenocarcinomas and controls, an iterated Least Absolute Shrinkage and Selection Operator multivariate Logistic Regression model was elaborated.

Results

The final predictive model, based on 10 VOCs, significantly and independently associated with the outcome had a sensitivity and specificity of 100 and 84% respectively, and an area under the ROC curve of 0.99. For further validation, the model was run on 50 other subjects: 24 cases and 26 controls; 23 patients with histological diagnosis of pancreatic adenocarcinomas and 25 controls were correctly identified by the model.

Conclusions

Pancreatic cancer is able to alter the concentration of some molecules in the blood and hence of VOCs in the alveolar air in equilibrium. The detection and statistical rendering of alveolar VOC composition can be useful for the clinical diagnostic approach of pancreatic neoplasms with excellent sensitivity and specificity.

Electronic supplementary material

The online version of this article (10.1186/s12885-018-4452-0) contains supplementary material, which is available to authorized users.