Development of a Web-Based System for Exploring Cancer Risk With Long-term Use of Drugs: Logistic Regression Approach

The construction, validation and promotion of the nomogram prognosis prediction model of UCEC, and the experimental verification of the expression and knockdown of the key gene GPX4

Background

Adequate prognostic prediction of Uterine Corpus Endometrial Carcinoma (UCEC) is crucial for informing clinical decision-making. However, there is a scarcity of research on the utilization of a nomogram prognostic evaluation model that incorporates pyroptosis-related genes (PRGs) in UCEC.

Methods

By analyzing data from UCEC patients in the TCGA database, four PRGs associated with prognosis were identified. Subsequently, a "risk score" was developed using these four PRGs and LASSO. Ordinary and web-based dynamic nomogram prognosis prediction models were constructed. The discrimination, calibration, clinical benefit, and promotional value of the selected GPX4 were validated. The expression level of GPX4 in UCEC cell lines was subsequently verified. The effects of GPX4 knock-down on the malignant biological behavior of UCEC cells were assessed.

Results

Four key PRGs and a "risk score" were identified, with the "risk score" calculated as (-0.4323) * GPX4 + (0.2385) * GSDME + (0.0525) * NLRP2 + (-0.3299) * NOD2. The nomogram prognosis prediction model, incorporating the "risk score," "age," and "FIGO stage," demonstrated moderate predictive performance (AUC >0.7), good calibration, and clinical significance for 1, 3, and 5-year survival. The web-based dynamic nomogram demonstrated significant promotional value (https://shibaolu.shinyapps.io/DynamicNomogramForUCEC/). UCEC cells exhibited abnormally elevated expression of GPX4, and the knockdown of GPX4 effectively suppressed malignant biological activities, including proliferation and migration, while inducing apoptosis. The findings from tumorigenic experiments conducted on nude mice further validated the results obtained from cellular experiments.

Conclusion

Following validation, the nomogram prognosis prediction model, which relies on four pivotal PRGs, demonstrated a high degree of accuracy in forecasting the precise probability of prognosis for patients with UCEC. Additionally, the web-based dynamic nomogram exhibited considerable potential for promotion. Notably, the key gene GPX4 exhibited characteristics of a potential oncogene in UCEC, as it facilitated malignant biological behavior and impeded apoptosis.

Pharmaceutical Agents as Potential Drivers in the Development of Early-Onset Colorectal Cancer: Case-Control Study

BACKGROUND

The incidence of early-onset colorectal cancer (EOCRC) rose abruptly in the mid 1990s, is continuing to increase, and has now been noted in many countries. By 2030, 25% of American patients diagnosed with rectal cancer will be 49 years or younger. The large majority of EOCRC cases are not found in patients with germline cancer susceptibility mutations (eg, Lynch syndrome) or inflammatory bowel disease. Thus, environmental or lifestyle factors are suspected drivers. Obesity, sedentary lifestyle, diabetes mellitus, smoking, alcohol, or antibiotics affecting the gut microbiome have been proposed. However, these factors, which have been present since the 1950s, have not yet been conclusively linked to the abrupt increase in EOCRC. The sharp increase suggests the introduction of a new risk factor for young people. We hypothesized that the driver may be an off-target effect of a pharmaceutical agent (ie, one requiring regulatory approval before its use in the general population or an off-label use of a previously approved agent) in a genetically susceptible subgroup of young adults. If a pharmaceutical agent is an EOCRC driving factor, regulatory risk mitigation strategies could be used.

OBJECTIVE

We aimed to evaluate the possibility that pharmaceutical agents serve as risk factors for EOCRC.

METHODS

We conducted a case-control study. Data including demographics, comorbidities, and complete medication dispensing history were obtained from the electronic medical records database of Maccabi Healthcare Services, a state-mandated health provider covering 26% of the Israeli population. The participants included 941 patients with EOCRC (≤50 years of age) diagnosed during 2001-2019 who were density matched at a ratio of 1:10 with 9410 control patients. Patients with inflammatory bowel disease and those with a known inherited cancer susceptibility syndrome were excluded. An advanced machine learning algorithm based on gradient boosted decision trees coupled with Bayesian model optimization and repeated data sampling was used to sort through the very high-dimensional drug dispensing data to identify specific medication groups that were consistently linked with EOCRC while allowing for synergistic or antagonistic interactions between medications. Odds ratios for the identified medication classes were obtained from a conditional logistic regression model.

RESULTS

Out of more than 800 medication classes, we identified several classes that were consistently associated with EOCRC risk across independently trained models. Interactions between medication groups did not seem to substantially affect the risk. In our analysis, drug groups that were consistently positively associated with EOCRC included beta blockers and valerian (Valeriana officinalis). Antibiotics were not consistently associated with EOCRC risk.

CONCLUSIONS

Our analysis suggests that the development of EOCRC may be correlated with prior use of specific medications. Additional analyses should be used to validate the results. The mechanism of action inducing EOCRC by candidate pharmaceutical agents will then need to be determined.

DEVELOPMENT OF A MACHINE LEARNING MODEL FOR PREDICTING 28-DAY MORTALITY OF SEPTIC PATIENTS WITH ATRIAL FIBRILLATION

ABSTRACT

Introduction: Septic patients with atrial fibrillation (AF) are common in the intensive care unit accompanied by high mortality. The early prediction of prognosis of these patients is critical for clinical intervention. This study aimed to develop a model by using machine learning (ML) algorithms to predict the risk of 28-day mortality in septic patients with AF. Methods: In this retrospective cohort study, we extracted septic patients with AF from the Medical Information Mart for Intensive Care III (MIMIC-III) and IV database. Afterward, only MIMIC-IV cohort was randomly divided into training or internal validation set. External validation set was mainly extracted from MIMIC-III database. Propensity score matching was used to reduce the imbalance between the external validation and internal validation data sets. The predictive factors for 28-day mortality were determined by using multivariate logistic regression. Then, we constructed models by using ML algorithms. Multiple metrics were used for evaluation of performance of the models, including the area under the receiver operating characteristic curve, sensitivity, specificity, recall, and accuracy. Results: A total of 5,317 septic patients with AF were enrolled, with 3,845 in the training set, 960 in the internal testing set, and 512 in the external testing set, respectively. Then, we established four prediction models by using ML algorithms. AdaBoost showed moderate performance and had a higher accuracy than the other three models. Compared with other severity scores, the AdaBoost obtained more net benefit. Conclusion: We established the first ML model for predicting the 28-day mortality of septic patients with AF. Compared with conventional scoring systems, the AdaBoost model performed moderately. The model established will have the potential to improve the level of clinical practice.

Chemopreventive Effects of Concomitant or Individual Use of Statins, Aspirin, Metformin, and Angiotensin Drugs: A Study Using Claims Data of 23 Million Individuals

Despite previous studies on statins, aspirin, metformin, and angiotensin-converting-enzyme inhibitors (ACEIs)/angiotensin II receptor blockers (ARBs), little has been studied about all their possible combinations for chemoprevention against cancers. This study aimed to comprehensively analyze the composite chemopreventive effects of all the combinations. In this case-control study, health records were retrieved from claims databases of Taiwan’s Health and Welfare Data Science Center. Eligible cases were matched at a 1:4 ratio with controls for age and sex. Both cases and controls were categorized into 16 exposure groups based on medication use. A total of 601,733 cancer cases were identified. Cancer risks (denoted by adjusted odds ratio; 99% confidence interval) were found to be significantly decreased: overall risk of all cancers in statin-alone (0.864; 0.843, 0.886), aspirin-alone (0.949; 0.939, 0.958), and ACEIs/ARBs (0.982; 0.978, 0.985) users; prostate (0.924; 0.889, 0.962) and female breast (0.967; 0.936, 1.000) cancers in metformin-alone users; gastrointestinal, lung, and liver cancers in aspirin and/or ACEIs/ARBs users; and liver cancer (0.433; 0.398, 0.471) in statin users. In conclusion, the results found no synergistic effect of multiple use of these agents on cancer prevention. Use of two (statins and aspirin, statins and metformin, statins and ACEIs/ARBs, and aspirin and ACEIS/ARBs) showed chemopreventive effects in some combinations, while the use of four, in general, did not.

Association between Statin Use and Gastric Cancer: A Nested Case-Control Study Using a National Health Screening Cohort in Korea

Concerns about the hazards of statins on the development and mortality of stomach cancers remain controversial. Here, we investigated the likelihood of incident gastric cancers and related mortality depending on statin exposure, statin type, and the duration of use. This nested case-control-designed study was composed of 8798 patients who were diagnosed with gastric cancer and matched with 35,192 controls at a 1:4 ratio based on propensity scores of age, sex, residential area, and income from the Korean National Health Insurance Service-Health Screening Cohort database (2002-2015). Propensity score overlap weighting was adjusted to balance the baseline covariates. Overlap propensity score-weighted logistic regression analyses were assessed to determine associations of the prior use of statins (any statin, hydrophilic statins vs. lipophilic statins) with incident gastric cancer and its mortality depending on the medication duration (<180 days, 180-545 days, and >545 days) after adjusting for multiple covariates. After adjustment, the use of any statin, hydrophilic statins, or lipophilic statins showed significant associations with lower odds for incident stomach cancer when used for a short-term period (180-545 days) (OR = 0.88, 95% CI = 0.81-0.86, p = 0.002; OR = 0.78, 95% CI = 0.66-0.92, p = 0.004; and OR = 0.91, 95% CI = 0.84-0.99, p = 0.039, respectively) compared to the control group. Hydrophilic statin use for 180-545 days was associated with 53% lower overall mortality (OR = 0.47; 95% CI = 0.29-0.77). In subgroup analyses, beneficial effects on both cancer development and mortality persisted in patients ≥65 years old, patients with normal blood pressure, and patients with high fasting glucose levels. There were no such associations with long-term statin use (>545 days). Thus, the current nationwide cohort study suggests that prior short-term statin use may have anti-gastric cancer benefits in elderly patients with hyperglycemia.

DeepDRG: Performance of Artificial Intelligence Model for Real-Time Prediction of Diagnosis-Related Groups

Nowadays, the use of diagnosis-related groups (DRGs) has been increased to claim reimbursement for inpatient care. The overall benefits of using DRGs depend upon the accuracy of clinical coding to obtain reasonable reimbursement. However, the selection of appropriate codes is always challenging and requires professional expertise. The rate of incorrect DRGs is always high due to the heavy workload, poor quality of documentation, and lack of computer assistance. We therefore developed deep learning (DL) models to predict the primary diagnosis for appropriate reimbursement and improving hospital performance. A dataset consisting of 81,486 patients with 128,105 episodes was used for model training and testing. Patients' age, sex, drugs, diseases, laboratory tests, procedures, and operation history were used as inputs to our multiclass prediction model. Gated recurrent unit (GRU) and artificial neural network (ANN) models were developed to predict 200 primary diagnoses. The performance of the DL models was measured by the area under the receiver operating curve, precision, recall, and F1 score. Of the two DL models, the GRU method, had the best performance in predicting the primary diagnosis (AUC: 0.99, precision: 83.2%, and recall: 66.0%). However, the performance of ANN model for DRGs prediction achieved AUC of 0.99 with a precision of 0.82 and recall of 0.57. The findings of our study show that DL algorithms, especially GRU, can be used to develop DRGs prediction models for identifying primary diagnosis accurately. DeepDRGs would help to claim appropriate financial incentives, enable proper utilization of medical resources, and improve hospital performance.

Predicting Hepatocellular Carcinoma With Minimal Features From Electronic Health Records: Development of a Deep Learning Model

BACKGROUND

Hepatocellular carcinoma (HCC), usually known as hepatoma, is the third leading cause of cancer mortality globally. Early detection of HCC helps in its treatment and increases survival rates.

OBJECTIVE

The aim of this study is to develop a deep learning model, using the trend and severity of each medical event from the electronic health record to accurately predict the patients who will be diagnosed with HCC in 1 year.

METHODS

Patients with HCC were screened out from the National Health Insurance Research Database of Taiwan between 1999 and 2013. To be included, the patients with HCC had to register as patients with cancer in the catastrophic illness file and had to be diagnosed as a patient with HCC in an inpatient admission. The control cases (non-HCC patients) were randomly sampled from the same database. We used age, gender, diagnosis code, drug code, and time information as the input variables of a convolution neural network model to predict those patients with HCC. We also inspected the highly weighted variables in the model and compared them to their odds ratio at HCC to understand how the predictive model works.

RESULTS

We included 47,945 individuals, 9553 of whom were patients with HCC. The area under the receiver operating curve (AUROC) of the model for predicting HCC risk 1 year in advance was 0.94 (95% CI 0.937-0.943), with a sensitivity of 0.869 and a specificity 0.865. The AUROC for predicting HCC patients 7 days, 6 months, 1 year, 2 years, and 3 years early were 0.96, 0.94, 0.94, 0.91, and 0.91, respectively.

CONCLUSIONS

The findings of this study show that the convolutional neural network model has immense potential to predict the risk of HCC 1 year in advance with minimal features available in the electronic health records.

Risk of cancer in long-term levothyroxine users: Retrospective population-based study

Levothyroxine is a widely prescribed medication for the treatment of an underactive thyroid. The relationship between levothyroxine use and cancer risk is largely underdetermined. To investigate the magnitude of the possible association between levothyroxine use and cancer risk, this retrospective case‐control study was conducted using Taiwan’s Health and Welfare Data Science Center database. Cases were defined as all patients who were aged ≥20 years and had a first‐time diagnosis for cancer at any site for the period between 2001 and 2011. Multivariable conditional logistic regression models were used to calculate an adjusted odds ratio (AOR) to reduce potential confounding factors. A total of 601 733 cases and 2 406 932 controls were included in the current study. Levothyroxine users showed a 50% higher risk of cancer at any site (AOR: 1.50, 95% CI: 1.46‐1.54; P < .0001) compared with non–users. Significant increased risks were also observed for brain cancer (AOR: 1.90, 95% CI: 1.48‐2.44; P < .0001), skin cancer (AOR: 1.42, 95% CI: 1.17‐1.72; P < .0001), pancreatic cancer (AOR: 1.27, 95% CI: 1.01‐1.60; P = .03), and female breast cancer (AOR: 1.24, 95% CI: 1.15‐1.33; P < .0001). Our study results showed that levothyroxine use was significantly associated with an increased risk of cancer, particularly brain, skin, pancreatic, and female breast cancers. Levothyroxine remains a highly effective therapy for hypothyroidism; therefore, physicians should carefully consider levothyroxine therapy and monitor patients’ condition to avoid negative outcomes. Additional studies are needed to confirm these findings and to evaluate the potential biological mechanisms.

Development and performance assessment of novel machine learning models to predict pneumonia after liver transplantation

Background

Pneumonia is the most frequently encountered postoperative pulmonary complications (PPC) after orthotopic liver transplantation (OLT), which cause high morbidity and mortality rates. We aimed to develop a model to predict postoperative pneumonia in OLT patients using machine learning (ML) methods.

Methods

Data of 786 adult patients underwent OLT at the Third Affiliated Hospital of Sun Yat-sen University from January 2015 to September 2019 was retrospectively extracted from electronic medical records and randomly subdivided into a training set and a testing set. With the training set, six ML models including logistic regression (LR), support vector machine (SVM), random forest (RF), adaptive boosting (AdaBoost), extreme gradient boosting (XGBoost) and gradient boosting machine (GBM) were developed. These models were assessed by the area under curve (AUC) of receiver operating characteristic on the testing set. The related risk factors and outcomes of pneumonia were also probed based on the chosen model.

Results

591 OLT patients were eventually included and 253 (42.81%) were diagnosed with postoperative pneumonia, which was associated with increased postoperative hospitalization and mortality (P < 0.05). Among the six ML models, XGBoost model performed best. The AUC of XGBoost model on the testing set was 0.734 (sensitivity: 52.6%; specificity: 77.5%). Pneumonia was notably associated with 14 items features: INR, HCT, PLT, ALB, ALT, FIB, WBC, PT, serum Na⁺, TBIL, anesthesia time, preoperative length of stay, total fluid transfusion and operation time.

Conclusion

Our study firstly demonstrated that the XGBoost model with 14 common variables might predict postoperative pneumonia in OLT patients.

Supplementary Information

The online version contains supplementary material available at 10.1186/s12931-021-01690-3.