A longitudinal transition imputation model for categorical data applied to a large registry dataset.

Imputation of longitudinal categorical covariates with several waves and many predictors is cumbersome in terms of implausible transitions, colinearity, and overfitting. We designed a simulation study with data obtained from a general practitioners' morbidity registry in Belgium for three waves, with smoking as the longitudinal covariate of interest. We set varying proportions of data on smoking to missing completely at random and missing not at random with proportions of missingness equal to 10%, 30%, 50%, and 70%. This study proposed a 3-stage approach that allows flexibility when imputing time-dependent categorical covariates. First, multiple imputation using fully conditional specification or multiple imputation for the predictor variables was deployed using the wide format such that previous and future information of the same patient was utilized. Second, a joint Markov transition model for initial, forward, backward, and intermittent probabilities was developed for each imputed dataset. Finally, this transition model was used for imputation. We compared the performance of this methodology with an analyses of the complete data and with listwise deletion in terms of bias and root mean square error. Next, we applied this methodology in a clinical case for years 2017 to 2021, where we estimated the effect of several covariates on the pneumococcal vaccination. This methodological framework ensures that the plausibility of transitions is preserved, overfitting and colinearity issues are resolved, and confounders can be utilized. Finally, a companion R package was developed to enable the replication and easy application of this methodology.

AI/ML Models to Predict the Severity of Drug-Induced Liver Injury for Small Molecules.

Drug-induced liver injury (DILI), believed to be a multifactorial toxicity, has been a leading cause of attrition of small molecules during discovery, clinical development, and postmarketing. Identification of DILI risk early reduces the costs and cycle times associated with drug development. In recent years, several groups have reported predictive models that use physicochemical properties or in vitro and in vivo assay endpoints; however, these approaches have not accounted for liver-expressed proteins and drug molecules. To address this gap, we have developed an integrated artificial intelligence/machine learning (AI/ML) model to predict DILI severity for small molecules using a combination of physicochemical properties and off-target interactions predicted in silico. We compiled a data set of 603 diverse compounds from public databases. Among them, 164 were categorized as Most DILI (M-DILI), 245 as Less DILI (L-DILI), and 194 as No DILI (N-DILI) by the FDA. Six machine learning methods were used to create a consensus model for predicting the DILI potential. These methods include k-nearest neighbor (k-NN), support vector machine (SVM), random forest (RF), Naïve Bayes (NB), artificial neural network (ANN), logistic regression (LR), weighted average ensemble learning (WA) and penalized logistic regression (PLR). Among the analyzed ML methods, SVM, RF, LR, WA, and PLR identified M-DILI and N-DILI compounds, achieving a receiver operating characteristic area under the curve of 0.88, sensitivity of 0.73, and specificity of 0.9. Approximately 43 off-targets, along with physicochemical properties (fsp3, log S, basicity, reactive functional groups, and predicted metabolites), were identified as significant factors in distinguishing between M-DILI and N-DILI compounds. The key off-targets that we identified include: PTGS1, PTGS2, SLC22A12, PPARγ, RXRA, CYP2C9, AKR1C3, MGLL, RET, AR, and ABCC4. The present AI/ML computational approach therefore demonstrates that the integration of physicochemical properties and predicted on- and off-target biological interactions can significantly improve DILI predictivity compared to chemical properties alone.

Testing the body specificity hypothesis: A comparative study of the Persian language and the Persian Sign language.

The present study explores the role of language in establishing the lateral space-valence mappings in mind. According to the body specificity hypothesis, regardless of linguistic and cultural conventions, "goodness" in people's minds is associated with the body's dominant side. The current study explores the language's influence on conceptualizing spatial metaphors by comparing two right-handed groups with similar cultural experiences but different language patterns. We used a computer-based task to compare the conceptual mappings between the right/left sides of space and emotional valence in Persian speakers and Persian Sign language users. Our result showed that right-handed Persian speakers strongly relate positive emotions to the right side and negative emotions to the left. This result is predictable by the effects of both linguistic and bodily experiences that are consistent in this case. However, in the case of Persian Sign language users, the bodily and linguistic experiences disagreed. Our finding showed that Sign language participants disregarding their bodily experiences, followed their linguistic patterns.

Use of antibiotics and colorectal cancer risk: a primary care nested case-control study in Belgium.

OBJECTIVES: To examine the association between the use of oral antibiotics and subsequent colorectal cancer risk. DESIGN: Matched case-control study. SETTING: General practice centres participating in the Integrated Computerised Network database in Flanders, Belgium. PARTICIPANTS: In total, 1705 cases of colorectal cancer diagnosed between 01 January 2010 and 31 December 2015 were matched to 6749 controls by age, sex, comorbidity and general practice centre. PRIMARY OUTCOME MEASURE: The association between the number of prescriptions for oral antibiotics and the incidence of colorectal cancer over a period of 1-10 years, estimated by a conditional logistic regression model. RESULTS: A significantly increased risk of colorectal cancer (OR 1.25, 95% CI 1.10 to 1.44) was found in subjects with one or more prescriptions compared with those with none after correction for diabetes mellitus. No dose-response relationship was found. CONCLUSIONS: This study resulted in a modestly higher risk of having colorectal cancer diagnosed after antibiotic exposure. The main limitation was missing data on known risk factors, in particular smoking behaviour. This study did not allow us to examine the causality of the relationship, indicating the need of further investigation.

Fast and optimal algorithm for case-control matching using registry data: application on the antibiotics use of colorectal cancer patients.

BACKGROUND: In case-control studies most algorithms allow the controls to be sampled several times, which is not always optimal. If many controls are available and adjustment for several covariates is necessary, matching without replacement might increase statistical efficiency. Comparing similar units when having observational data is of utter importance, since confounding and selection bias is present. The aim was twofold, firstly to create a method that accommodates the option that a control is not resampled, and second, to display several scenarios that identify changes of Odds Ratios (ORs) while increasing the balance of the matched sample. METHODS: The algorithm was derived in an iterative way starting from the pre-processing steps to derive the data until its application in a study to investigate the risk of antibiotics on colorectal cancer in the INTEGO registry (Flanders, Belgium). Different scenarios were developed to investigate the fluctuation of ORs using the combination of exact and varying variables with or without replacement of controls. To achieve balance in the population, we introduced the Comorbidity Index (CI) variable, which is the sum of chronic diseases as a means to have comparable units for drawing valid associations. RESULTS: This algorithm is fast and optimal. We simulated data and demonstrated that the run-time of matching even with millions of patients is minimal. Optimal, since the closest controls is always captured (using the appropriate ordering and by creating some auxiliary variables), and in the scenario that a case has only one control, we assure that this control will be matched to this case, thus maximizing the cases to be used in the analysis. In total, 72 different scenarios were displayed indicating the fluctuation of ORs, and revealing patterns, especially a drop when balancing the population. CONCLUSIONS: We created an optimal and computationally efficient algorithm to derive a matched case-control sample with and without replacement of controls. The code and the functions are publicly available as an open source in an R package. Finally, we emphasize the importance of displaying several scenarios and assess the difference of ORs while using an index to balance population in observational data.

Retinal oxygen saturation as a non-invasive estimate for mixed venous oxygen saturation and cardiac output.

PURPOSE: To investigate the correlation between retinal vessel oxygen saturation and mixed venous oxygen saturation (SvO2-mixed ) and cardiac output (CO). METHODS: Retinal arterial (SaO2-retinal ) and venous (SvO2-retinal ) oxygen saturation were measured non-invasively with dual-wavelength retinal oximetry in subjects receiving invasive measurements of SvO2-mixed and CO through right heart catheterization. Correlations were analysed using Spearman's rank correlation coefficients and linear regression models. RESULTS: Fourteen patients (median age 62.7 years, range: 21-77) were included in the analysis. When adjusted for age, SvO2-retinal showed a positive correlation with SvO2-mixed (ß = 0.80, p = 0.003). Retinal arteriovenous oxygen saturation difference was significantly correlated with the inverse of CO (Spearman's ρ = 0.59, p = 0.026). CONCLUSION: This pilot study provides proof of concept for the use of retinal oximetry as a non-invasive tool to assess systemic cardiovascular function.

Improved discrimination between normal-tension and primary open-angle glaucoma with advanced vascular examinations - the Leuven Eye Study.

PURPOSE: Vascular factors have been suggested to influence the development and progression of glaucoma. They are thought to be especially relevant for normal-tension glaucoma (NTG) patients. We aim to investigate which vascular factors, including advanced vascular examinations, better describe patients with NTG comparing to those with primary open-angle glaucoma (POAG). METHODS: The Leuven Eye Study database (182 NTG and 202 POAG patients; similar structural and functional damage) was used to compute three multivariate logistic regression models: a conventional model (conventional parameters only, including vascular-related self-reported phenomena, such as migraine or peripheral vasospasm); an advanced vascular model (advanced vascular parameters only: colour Doppler imaging (CDI), retinal oximetry, ocular pulse amplitude and choroidal thickness); and a global model, in which both types of parameters were allowed. Receiver operating characteristic (ROC) curves and corresponding areas under the curve (AUC) were calculated and compared between models. RESULTS: Patients with NTG had a higher resistive index and lower early systolic acceleration (ESA) in their retrobulbar vessels and a smaller arteriovenous retinal oxygen saturation difference. The global model (AUC 0.743) showed a significantly better discriminative ability when compared to either the conventional (AUC 0.687, p = 0.049) or the advanced vascular (AUC 0.677, p = 0.005) models. Also, the conventional and the advanced vascular models showed a similar discriminative ability (p = 0.823). CONCLUSION: Patients with NTG have more signs of vascular dysfunction. Clinical conventional parameters, such as asking simple vascular-related questions, combined with advanced vascular examinations provide information to better understand the value that non-IOP-related factors play in NTG.

On using multiple imputation for exploratory factor analysis of incomplete data.

A simple multiple imputation-based method is proposed to deal with missing data in exploratory factor analysis. Confidence intervals are obtained for the proportion of explained variance. Simulations and real data analysis are used to investigate and illustrate the use and performance of our proposal.

On using multiple imputation for exploratory factor analysis of incomplete data.

A simple multiple imputation-based method is proposed to deal with missing data in exploratory factor analysis. Confidence intervals are obtained for the proportion of explained variance. Simulations and real data analysis are used to investigate and illustrate the use and performance of our proposal.

Non-invasive assessment of cerebral oxygenation: A comparison of retinal and transcranial oximetry.

BACKGROUND: To investigate the correlation between cerebral (SO2-transcranial), retinal arterial (SaO2-retinal) and venous (SvO2-retinal) oxygen saturation as measured by near-infrared spectroscopy (NIRS) and retinal oximetry respectively. METHODS: Paired retinal and cerebral oxygen saturation measurements were performed in healthy volunteers. Arterial and venous retinal oxygen saturation and diameter were measured using a non-invasive spectrophotometric retinal oximeter. Cerebral oxygen saturation was measured using near-infrared spectroscopy. Correlations between SO2-transcranial and retinal oxygen saturation and diameter measurements were assessed using Pearson correlation coefficients. Lin's concordance correlation coefficient (CCC) and Bland-Altman analysis were performed to evaluate the agreement between SO2-transcranial as measured by NIRS and as estimated using a fixed arterial:venous ratio as 0.3 x SaO2-retinal + 0.7 x SvO2-retinal. The individual relative weight of SaO2-retinal and SvO2-retinal to obtain the measured SO2-transcranial was calculated for all subjects. RESULTS: Twenty-one healthy individuals aged 26.4 ± 2.2 years were analyzed. SO2-transcranial was positively correlated with both SaO2-retinal and SvO2-retinal (r = 0.44, p = 0.045 and r = 0.43, p = 0.049 respectively) and negatively correlated with retinal venous diameter (r = -0.51, p = 0.017). Estimated SO2-transcranial based on retinal oximetry showed a tolerance interval of (-13.70 to 14.72) and CCC of 0.46 (95% confidence interval: 0.05 to 0.73) with measured SO2-transcranial. The average relative weights of SaO2-retinal and SvO2-retinal to obtain SO2-transcranial were 0.31 ± 0.11 and 0.69 ± 0.11, respectively. CONCLUSION: This is the first study to show the correlation between retinal and cerebral oxygen saturation, measured by NIRS and retinal oximetry. The average relative weight of arterial and venous retinal oxygen saturation to obtain the measured transcranial oxygen saturation as measured by NIRS, approximates the established arterial:venous ratio of 30:70 closely, but shows substantial inter-individual variation. These findings provide a proof of concept for the role of retinal oximetry in evaluating cerebral oxygenation.

Fast precision estimation in high-dimensional multivariate joint models.

A fast way is proposed based on the multiple outputation idea (Hoffman et al., ; Follmann et al., ) to calculate the precision of parameter estimates for high-dimensional multivariate joint models using a pairwise approach (Fieuws and Verbeke, ; Fieuws et al., ). Simulation results as well as data analysis shows possibly more than 2500 times faster computations using the proposed method. In our real data illustration, the time gain is more than 330 times.