Abstract Background Machine learning models have been used to create accurate prediction models for dementia. However, many suffer from overfitting and external validation often results in decreased performance. Pooling data from various sources for model training can improve the generalizability of prediction models. We show here a prediction model for dementia developed on pooled data from the Dementia Risk Prediction Pooling (DRPP) Consortium. Method Data from 11 longitudinal disease cohorts within the DRPP and relevant risk factors (25 in total) were collected and harmonized at baseline and follow‐up exams. An ensemble tree‐based algorithm, LightGBM, was used to create two prediction models for dementia at or before 10 years. The first model contains all variables in the dataset, and the second clinical model excludes the Mini‐Mental State Examination (MMSE) score and APOE genotype. 5‐fold cross‐validation was repeated 1000 times to tune the model hyperparameters (number of leaves, tree depth, learning rate) to yield the greatest area under the curve (AUC). Feature importance of the model was analyzed via individual feature information gain. Analysis was performed in R 4.1.2. Results Among the 55,614 participants from 11 cohorts included in this analysis (Table 1), the first model with all the variables had a cross‐validation AUC of 0.762 (CI: 0.757‐0.767). Counterintuitively, the second model of the without the MMSE and APOE variables yielded an AUC of 0.804 (CI: 0.799‐0.809), which may indicate overfitting in the first model. Feature importance analysis show that age is the most important variable in both models. APOE, MMSE, fasting glucose, and any physical activity are the next most important predictors in the full model (Figure 1). In the second model, fasting glucose, any physical activity, gender, and A1c levels were the next most important predictors (Figure 2). Conclusion By pooling various data sources, we can train machine learning models for dementia risk prediction on more diverse data. Further work is needed to compare the performance of these models with models trained on single data sources via external validation. A pooled dataset also offers an opportunity to understand how model performance will change given shifts in the underlying population.
Introduction: Renal cysts have been generally regarded as benign incidental findings on ultrasound or CT scan. Nevertheless, recent data suggest that such cysts might reflect early stages of end-organ damage. Given the similarities of cerebral and renal vasculatures, in this population-based study we sought to investigate whether individuals with the renal cysts have higher prevalence of cerebrovascular accidents. Methods: Between 2009 and 2014, renal ultrasound was performed on 2,984 participants from the population-based Rotterdam Study (mean age: 71.9 ± 9.0 years, 57% female). Individuals who were found to have renal cysts were categorized into single and multiple cysts carriers. Stroke and transient ischemic attacks (TIA) cases were identified using digital record linkage with general practitioners and medical specialists in the research area. Multivariate logistic regression models were applied to assess the link between single and multiple renal cysts carriers and prevalent stroke and TIAs. Results: Single renal cyst was found in 508 (17.0%) individuals and 197 (6.6%) participants had multiple cysts. Subjects with a single renal cyst compared to no cyst had 1.63 (95% CI: 1.01, 2.64) higher odds of having concomitant stroke. This association was more pronounced in individuals with multiple renal cysts. Subjects with multiple cysts had 2.14 (95% CI: 1.20, 3.82) higher odds of having stroke compared to individuals without cyst. There was no significant association between single or multiple renal cysts and TIAs; odds ratio (OR): 1.06 (95% CI: 0.70, 1.59) and OR: 1.29 (95% CI: 0.74, 2.23), respectively. All these associations were independent of age, sex, systolic and diastolic blood pressure, body mass index, smoking, history of cardiovascular disease, diabetes mellitus and kidney function. Conclusion: In this large population-based study we show that there is a dose dependent association between renal cysts and prevalent stroke. This data questions the general consensus that renal cysts are of no clinical significance as they might signal higher degrees of vascular damages in the brain circulation.
Background There are inconsistent findings on the role of hyperuricemia as an independent risk factor for chronic kidney disease (CKD). Hypertension has been implicated as a factor influencing the association between serum uric acid and CKD. In this population-based study we investigated the association between serum uric acid and decline in renal function and tested whether hypertension moderates this association. Methods We included 2601 subjects aged 55 years and over from the Rotterdam Study. Serum uric acid and estimated glomerular filtration rate (eGFR) were assessed at baseline. After average 6.5 years of follow-up, second eGFR was assessed. CKD was defined as eGFR<60 ml/min/1.73 m2. All associations were corrected for socio-demographic and cardiovascular factors. Results Each unit (mg/dL) increase in serum uric acid was associated with 0.19 ml/min per 1.73 m2 faster annual decline in eGFR. While the association between serum uric acid and incidence of CKD was not significant in our study population (Hazard Ratio: 1.12, 95% confidence interval [CI]: 0.98–1.28), incorporating our results in a meta-analysis with eleven published studies revealed a significant association (Relative Risk: 1.18, 95%CI: 1.15–1.22). In the stratified analyses, we observed that the associations of serum uric acid with eGFR decline and incident CKD were stronger in hypertensive subjects (P for interaction = 0.046 and 0.024, respectively). Conclusions Our findings suggest that hyperuricemia is independently associated with a decline in renal function. Stronger association in hypertensive individuals may indicate that hypertension mediates the association between serum uric acid and CKD.
High protein intake in early childhood is associated with obesity, suggesting possible adverse effects on other cardiometabolic outcomes. However, studies in adults have suggested beneficial effects of protein intake on blood pressure (BP) and lipid profile. Whether dietary protein intake is associated with cardiovascular and metabolic health in children is unclear. Therefore, we aimed to systematically review the evidence on the associations of protein intake with BP, insulin sensitivity and blood lipids in children. We searched the databases Medline, Embase, Cochrane Central and PubMed for interventional and observational studies in healthy children up to the age of 18 years, in which associations of total, animal and/or vegetable protein intake with one or more of the following outcomes were reported: BP; measures of insulin sensitivity; cholesterol levels; or TAG levels. In the search, we identified 6636 abstracts, of which fifty-six studies met all selection criteria. In general, the quality of the included studies was low. Most studies were cross-sectional, and many did not control for potential confounders. No overall associations were observed between protein intake and insulin sensitivity or blood lipids. A few studies suggested an inverse association between dietary protein intake and BP, but evidence was inconclusive. Only four studies examined the effects of vegetable or animal protein intake, but with inconsistent results. In conclusion, the literature, to date provides insufficient evidence for effects of protein intake on BP, insulin sensitivity or blood lipids in children. Future studies could be improved by adequately adjusting for key confounders such as energy intake and obesity.
Introduction: Leukocyte telomere length (LTL) is an important aging biomarker implicated in the pathogenesis of age-related conditions. We conducted a large-scale proteomics study to characterize the proteomic signatures of LTL and its genetic determinants measured by polygenic risk score (PRS). Hypothesis: We hypothesize that LTL and LTL PRS are associated with blood proteomic signatures linked to aging-related conditions. Methods: We used TelSeq to estimate LTL based on whole genome sequencing (WGS) data measured in samples collected in ARIC visits 1, 2, and 3 (i.e., the baseline for this study). TelSeq counted reads containing at least 12 TTAGGG repeats. This count was then normalized to the number of reads with GC content between 48% and 52%. Only samples with read lengths of 150 or 151 were included in subsequent analyses, and LTL estimates were inverse normalized within read length group and WGS group before being merged. We derived a PRS for LTL based on previously published genome-wide association study (GWAS) summary statistics for 150 significant autosomal SNPs and ARIC GWAS data. We then analyzed LTL or LTL PRS as a predictor and plasma levels of proteins as outcomes, measured at ARIC visit 2 (for the LTL PRS analysis) and visit 3 (for the LTL analysis) using SOMAscan v4 (~5,000 proteins). Race-specific linear regressions were performed to evaluate the association between LTL or LTL PRS and levels of each protein adjusting for potential confounders at baseline. We used a Bonferroni correction to account for multiple testing. The analysis in Whites served as a discovery (p≤1x10 -5 ) and that in Blacks as a replication. Results: The final sample sizes were as follows: LTL PRS analysis (n=7,587 Whites and 2,094 Blacks) and LTL analysis (n=5,014 Whites and 884 Blacks). In Whites, LTL PRS was significantly associated with five proteins (p≤1x10 -5 ) that also showed the same direction of association in Blacks: THPO, GP1Bα, PEAR1, KDR, and GP5, with the association for KDR reaching nominal significance in Blacks (p=0.02). For LTL as the predictor, GP1Bα, GDF15 and TXNDC5 were significant in Whites, with consistent direction of associations in Blacks (p>0.05). These proteins are linked to hemostasis (all), endothelial proliferation (KDR), heart disease and cancer (GDF15), and cancer and rheumatoid arthritis (TXNDC5) in the literature. Conclusion: This large-scale proteomic analysis reveals blood proteomic signatures for LTL PRS and LTL, which will likely improve the understanding of biological pathways and clinical conditions implicated by LTL.
To investigate the association of kidney function with white matter microstructural integrity.We included 2,726 participants with a mean age of 56.6 years (45% men) from the population-based Rotterdam Study. Albumin-to-creatinine ratio, and estimated glomerular filtration rate (eGFR), using serum cystatin C (eGFRcys) and creatinine (eGFRcr), were measured to evaluate kidney function. Diffusion-MRI was used to assess microstructural integrity of the normal-appearing white matter. Multiple linear regression models, adjusted for macrostructural MRI markers and cardiovascular risk factors, were used to model the association of kidney function with white matter microstructure.Participants had average eGFRcr of 86.1 mL/min/1.73 m(2), average eGFRcys of 86.2 mL/min/1.73 m(2), and median albumin-to-creatinine ratio of 3.4 mg/g. Lower eGFRcys was associated with worse global white matter microstructural integrity, reflected as lower fractional anisotropy (standardized difference per SD: -0.053, 95% confidence interval [CI]: -0.092, -0.014) and higher mean diffusivity (0.036, 95% CI: 0.001, 0.070). Similarly, higher albumin-to-creatinine ratio was associated with lower fractional anisotropy (-0.044, 95% CI: -0.078, -0.011). There was no linear association between eGFRcr and white matter integrity. Subgroup analyses showed attenuation of the associations after excluding subjects with hypertension. The associations with global diffusion tensor imaging measures did not seem to be driven by particular tracts, but rather spread across multiple tracts in various brain regions.Reduced kidney function is associated with worse white matter microstructural integrity. Our findings highlight the importance for clinicians to consider concomitant macro- and microstructural changes of the brain in patients with impaired kidney function.
'%3e%3cpath fill='%23012169' d='M0 0v30h60V0z'/%3e%3cpath stroke='%23fff' stroke-width='6' d='m0 0 60 30m0-30L0 30'/%3e%3cpath stroke='%23C8102E' stroke-width='4' d='m0 0 60 30m0-30L0 30' clip-path='url(%23b)'/%3e%3cpath stroke='%23fff' stroke-width='10' d='M30 0v30M0 15h60'/%3e%3cpath stroke='%23C8102E' stroke-width='6' d='M30 0v30M0 15h60'/%3e%3c/g%3e%3c/svg%3e)
