Machine Learning-Based Prediction of Hemoglobinopathies Using Complete Blood Count Data.

BACKGROUND: Hemoglobinopathies, the most common inherited blood disorder, are frequently underdiagnosed. Early identification of carriers is important for genetic counseling of couples at risk. The aim of this study was to develop and validate a novel machine learning model on a multicenter data set, covering a wide spectrum of hemoglobinopathies based on routine complete blood count (CBC) testing. METHODS: Hemoglobinopathy test results from 10 322 adults were extracted retrospectively from 8 Dutch laboratories. eXtreme Gradient Boosting (XGB) and logistic regression models were developed to differentiate negative from positive hemoglobinopathy cases, using 7 routine CBC parameters. External validation was conducted on a data set from an independent Dutch laboratory, with an additional external validation on a Spanish data set (n = 2629) specifically for differentiating thalassemia from iron deficiency anemia (IDA). RESULTS: The XGB and logistic regression models achieved an area under the receiver operating characteristic (AUROC) of 0.88 and 0.84, respectively, in distinguishing negative from positive hemoglobinopathy cases in the independent external validation set. Subclass analysis showed that the XGB model reached an AUROC of 0.97 for ß-thalassemia, 0.98 for α0-thalassemia, 0.95 for homozygous α+-thalassemia, 0.78 for heterozygous α+-thalassemia, and 0.94 for the structural hemoglobin variants Hemoglobin C, Hemoglobin D, Hemoglobin E. Both models attained AUROCs of 0.95 in differentiating IDA from thalassemia. CONCLUSIONS: Both the XGB and logistic regression model demonstrate high accuracy in predicting a broad range of hemoglobinopathies and are effective in differentiating hemoglobinopathies from IDA. Integration of these models into the laboratory information system facilitates automated hemoglobinopathy detection using routine CBC parameters.

Inconsistency in ferritin reference intervals across laboratories: a major concern for clinical decision making.

OBJECTIVES: Iron deficiency anemia is a significant global health concern, diagnosed by measuring hemoglobin concentrations in combination with plasma ferritin concentration. This study investigated the variability in ferritin reference intervals among laboratories in the Netherlands and examined how this affects the identification of iron-related disorders. METHODS: Ferritin reference intervals from 52 Dutch ISO15189-certified medical laboratories were collected. Ferritin, hemoglobin and mean corpuscular volume data of non-anemic apparently healthy primary care patients, measured by four laboratory platforms (Beckman, Abbott, Siemens, and Roche), were collected (n=397,548). Median ferritin levels were determined per platform, stratified by sex and age. The proportion of ferritin measurements outside of the reference interval was calculated using the reference intervals from the 52 laboratories (using a total of n=1,093,442 ferritin measurements). Lastly, ferritin data from 3,699 patients as captured in general practitioner (GP) data from the PHARMO Data Network were used to assess the variation of abnormal ferritin measurements per GP. RESULTS: Median plasma ferritin concentrations were approximately four times higher in men and twice as high in postmenopausal women compared to premenopausal women. Moreover, there are substantial differences in the median plasma ferritin concentration between the four platforms. However, even among laboratories using the same platform, ferritin reference intervals differ widely. This leads to significant differences in the percentages of measurements classified as abnormal, with the percentage of ferritin measurements below the reference limit in premenopausal women ranging from 11 to 53 %, in postmenopausal women from 3 to 37 %, and in men from 2 to 19 %. The percentage of ferritin measurements above the reference limit in premenopausal women ranged from 0.2 to 11 %, in postmenopausal women from 3 to 36 % and in men from 7 to 32 %. CONCLUSIONS: The lack of harmonization in ferritin measurement and the disagreement in plasma ferritin reference intervals significantly impact the interpretation of the iron status of patients and thereby the number of iron disorder diagnoses made. Standardization or harmonization of the ferritin assays and establishing uniform reference intervals and medical decision limits are essential to reduce the substantial variability in clinical interpretations of ferritin results.

Discrepancies in corrected calcium versus ionised calcium in a geriatric population: an observational study.

BACKGROUND: Calcium can be measured as ionised (Ca-ionised) or albumin-adjusted total calcium (Ca-albumin). Current clinical guidelines predominantly utilise Ca-albumin, despite Ca-ionised being the gold standard. Discrepancies can occur between these measurement modalities and can lead to clinical dilemmas. It remains unclear how large these discrepancies are in older patients. This study investigated the discrepancies between Ca-ionised and Ca-albumin in geriatric patients. METHODS: This is an observational study of all geriatric patients (n = 876) in the Jeroen Bosch Hospital (January 2018 and January 2021) in whom both Ca-ionised and Ca-albumin were measured. Misclassification of calcaemic state (i.e. low, normal or high) was calculated (percentages), the measure of agreement was described using Cohen's Kappa and for the continuous data Pearson's correlation coefficient was used. Relevant categories of age and renal function were considered for effect modification effects and studied by interaction terms in a regression model. RESULTS: In one-third of the measurements, there was a misclassification. Ca-albumin measurements failed to identify 28% of hypocalcaemia. In 3.5%, hypercalcemia based on Ca-albumin was not confirmed by Ca-ionised. The correlation coefficient between Ca-ionised and Ca-albumin was 0.743 (P = 0.01) and measure of agreement by Kappa was 0.213 (P < 0.001). In the oldest old (≥ 85 years) and patients with eGFR <30 ml/min/1.73 m2 ,the agreement by Kappa was lower, with values of 0.192 and 0.104, respectively. CONCLUSION: There is a discrepancy between Ca-albumin and Ca-ionised in one-third of the geriatric patients, leading to clinical dilemmas. In the oldest old and patients with renal dysfunction, this problem is most pronounced.

Albumin determined by bromocresol green leads to erroneous results in routine evaluation of patients with chronic kidney disease.

OBJECTIVES: Measurement of plasma albumin is pivotal for clinical decision-making in patients with chronic kidney disease (CKD). Routinely used methods as bromocresol green (BCG) and bromocresol purple (BCP) can suffer from aselectivity, but the impact of aselectivity on the accuracy of plasma albumin results of CKD-patients is still unknown. Therefore, we evaluated the performance of BCG-, BCP- and JCTLM-endorsed immunological methods in patients with various stages of CKD. METHODS: We evaluated the performance of commonly used albumin methods in patients with CKD stages G1 through G5, the latter divided in two groups based on whether they received hemodialysis treatment. In total, 163 patient plasma samples were measured at 14 laboratories, on six different BCG and BCP-platforms, and four different immunological platforms. The results were compared with an ERM-DA-470k-corrected nephelometric assay. The implications on outcome is evaluated by the proportion of patient results <38 g/L for the diagnosis of protein energy wasting. RESULTS: Albumin results determined with BCP- and immunological methods showed the best agreement with the target value (92.7 and 86.2 %, respectively vs. 66.7 % for BCG, namely due to overestimation). The relative agreement of each method with the target value was platform-dependent, with larger variability in agreement between platforms noted for BCG and immunological methods (3.2-4.6 and 2.6-5.3 %) as opposed to BCP (0.7-1.5 %). The stage of CKD had similar effects on the variability in agreement for the three method-groups (0.6-1.8 % vs. 0.7-1.5 % vs. 0.4-1.6 %). The differences between methods cause discrepancies in clinical decision-making, as structurally fewer patients were diagnosed with protein energy wasting upon using BCG-based albumin results. CONCLUSIONS: Our study shows that BCP is fit for the intended use to measure plasma albumin levels in CKD patients from all stages, including patients on hemodialysis. In contrast, most BCG-based platforms falsely overestimate the plasma albumin concentration.

Validation of the Hem-Col capillary blood collection system for routine laboratory analyses.

At home collection of capillary blood using Hem-Col tubes (Labonovum) could offer a solution to patients with chronic conditions, who require frequent laboratory analyses. The collection tubes contain a conservation buffer to stabilize analytes for up to 5 days. In this validation study it was investigated whether analytes are measured accurately in Hem-Col tubes 5 days after collection. Forty-six healthy volunteers donated blood via venepuncture as well as capillary blood by finger prick using Hem-Col tubes. The analytes were measured within 2 h for the venepuncture and after 120 h for the Hem-Col method. The results of each analyte were analysed using Passing-Bablok regression analyses. The analytes that met the predefined acceptance criteria were total cholesterol, LDL-cholesterol, thyroid stimulating hormone (TSH) and glycated haemoglobin (HbA1c). HDL-cholesterol, C-reactive protein (CRP), ferritin, bilirubin total, creatinine, gGT and triglycerides met two out of three acceptance criteria. All other analytes did not meet the predefined criteria. The Hem-Col method is suitable for the measurement of total cholesterol, LDL-cholesterol, thyroid stimulating hormone (TSH) and glycated haemoglobin (HbA1c). However, due to this limited set of valid tests and practical limitations, routine application of this novel collection system in daily practice is limited.

Automated prediction of low ferritin concentrations using a machine learning algorithm.

OBJECTIVES: Computational algorithms for the interpretation of laboratory test results can support physicians and specialists in laboratory medicine. The aim of this study was to develop, implement and evaluate a machine learning algorithm that automatically assesses the risk of low body iron storage, reflected by low ferritin plasma levels, in anemic primary care patients using a minimal set of basic laboratory tests, namely complete blood count and C-reactive protein (CRP). METHODS: Laboratory measurements of anemic primary care patients were used to develop and validate a machine learning algorithm. The performance of the algorithm was compared to twelve specialists in laboratory medicine from three large teaching hospitals, who predicted if patients with anemia have low ferritin levels based on laboratory test reports (complete blood count and CRP). In a second round of assessments the algorithm outcome was provided to the specialists in laboratory medicine as a decision support tool. RESULTS: Two separate algorithms to predict low ferritin concentrations were developed based on two different chemistry analyzers, with an area under the curve of the ROC of 0.92 (Siemens) and 0.90 (Roche). The specialists in laboratory medicine were less accurate in predicting low ferritin concentrations compared to the algorithms, even when knowing the output of the algorithms as support tool. Implementation of the algorithm in the laboratory system resulted in one new iron deficiency diagnosis on average per day. CONCLUSIONS: Low ferritin levels in anemic patients can be accurately predicted using a machine learning algorithm based on routine laboratory test results. Moreover, implementation of the algorithm in the laboratory system reduces the number of otherwise unrecognized iron deficiencies.

Renal phospholipidosis and impaired magnesium handling in high-fat-diet-fed mice.

Hypomagnesemia (blood Mg2+ concentration <0.7 mM) is a common electrolyte disorder in patients with type 2 diabetes (T2D), but the etiology remains largely unknown. In patients with T2D, reduced blood Mg2+ levels are associated with an increased decline in renal function, independent of glycemic control and hypertension. To study the underlying mechanism of this phenomenon, we investigated the renal effects of hypomagnesemia in high-fat-diet (HFD)-fed mice. In mice fed a low dietary Mg2+, the HFD resulted in severe hypomagnesemia within 4 wk. Renal or intestinal Mg2+ wasting was not observed after 16 wk on the diets. Despite the absence of urinary or fecal Mg2+ loss, the HFD induced a reduction in the mRNA expression transient receptor potential melastatin type 6 in both the kidney and colon. mRNA expression of distal convoluted tubule (DCT)-specific genes was down-regulated by the LowMg-HFD, indicating atrophy of the DCT. The low dietary Mg2+ resulted in severe HFD-induced proximal tubule phospholipidosis, which was absent in mice on a NormalMg-HFD. This was accompanied by albuminuria, moderate renal damage, and alterations in renal energy metabolism, including enhanced gluconeogenesis and cholesterol synthesis. In conclusion, this study shows that hypomagnesemia is a consequence of diet-induced obesity and insulin resistance. Moreover, hypomagnesemia induces major structural changes in the diabetic kidney, including proximal tubular phospholipidosis, providing a novel mechanism for the increased renal decline in patients with hypomagnesemic T2D.-Kurstjens, S., Smeets, B., Overmars-Bos, C., Dijkman, H. B., den Braanker, D. J. W., de Bel, T., Bindels, R. J. M., Tack, C. J. J., Hoenderop, J. G. J., de Baaij, J. H. F. Renal phospholipidosis and impaired magnesium handling in high-fat-diet-fed mice.

Rapid identification of SARS-CoV-2-infected patients at the emergency department using routine testing.

Objectives: The novel coronavirus disease 19 (COVID-19), caused by SARS-CoV-2, spreads rapidly across the world. The exponential increase in the number of cases has resulted in overcrowding of emergency departments (ED). Detection of SARS-CoV-2 is based on an RT-PCR of nasopharyngeal swab material. However, RT-PCR testing is time-consuming and many hospitals deal with a shortage of testing materials. Therefore, we aimed to develop an algorithm to rapidly evaluate an individual's risk of SARS-CoV-2 infection at the ED. Methods: In this multicenter retrospective study, routine laboratory parameters (C-reactive protein, lactate dehydrogenase, ferritin, absolute neutrophil and lymphocyte counts), demographic data and the chest X-ray/CT result from 967 patients entering the ED with respiratory symptoms were collected. Using these parameters, an easy-to-use point-based algorithm, called the corona-score, was developed to discriminate between patients that tested positive for SARS-CoV-2 by RT-PCR and those testing negative. Computational sampling was used to optimize the corona-score. Validation of the model was performed using data from 592 patients. Results: The corona-score model yielded an area under the receiver operating characteristic curve of 0.91 in the validation population. Patients testing negative for SARS-CoV-2 showed a median corona-score of 3 vs. 11 (scale 0-14) in patients testing positive for SARS-CoV-2 (p<0.001). Using cut-off values of 4 and 11 the model has a sensitivity and specificity of 96 and 95%, respectively. Conclusions: The corona-score effectively predicts SARS-CoV-2 RT-PCR outcome based on routine parameters. This algorithm provides the means for medical professionals to rapidly evaluate SARS-CoV-2 infection status of patients presenting at the ED with respiratory symptoms.

Metformin regulates TRPM6, a potential explanation for magnesium imbalance in type 2 diabetes patients.

Metformin therapy is associated with lower serum magnesium (Mg2+) levels in type 2 diabetes patients. The TRPM6 channel determines the fine-tuning of Mg2+ (re)absorption in intestine and kidney. Therefore, we aimed to investigate the short- and long-term effects of metformin on TRPM6. Patch clamp recordings and biotinylation assays were performed upon 1 h of incubation with metformin in TRPM6-transfected HEK293 cells. Additionally, 24 h of treatment of mDCT15 kidney and hCaco-2 colon cells with metformin was applied to measure the effects on endogenous TRPM6 expression by quantitative real-time PCR. To assess Mg2+ absorption, 25Mg2+ uptake measurements were performed using inductively coupled plasma mass spectrometry. Short-term effects of metformin significantly increased TRPM6 activity and its cell surface trafficking. In contrast, long-term effects significantly decreased TRPM6 mRNA expression and 25Mg2+ uptake. Metformin lowered TRPM6 mRNA levels independently of insulin- and AMPK-mediated pathways. Moreover, in type 2 diabetes patients, metformin therapy was associated with lower plasma Mg2+ concentrations and fractional excretion of Mg2+. Thereby, short-term metformin treatment increases TRPM6 activity explained by enhanced cell surface expression. Conversely, long-term metformin treatment results in downregulation of TRPM6 gene expression in intestine and kidney cells. This long-term effect translated in an inverse correlation between metformin and plasma Mg2+ concentration in type 2 diabetes patients.

Increased NEFA levels reduce blood Mg2+ in hypertriacylglycerolaemic states via direct binding of NEFA to Mg2.

AIMS/HYPOTHESIS: The blood triacylglycerol level is one of the main determinants of blood Mg2+ concentration in individuals with type 2 diabetes. Hypomagnesaemia (blood Mg2+ concentration <0.7 mmol/l) has serious consequences as it increases the risk of developing type 2 diabetes and accelerates progression of the disease. This study aimed to determine the mechanism by which triacylglycerol levels affect blood Mg2+ concentrations. METHODS: Using samples from 285 overweight individuals (BMI >27 kg/m2) who participated in the 300-Obesity study (an observational cross-sectional cohort study, as part of the Human Functional Genetics Projects), we investigated the association between serum Mg2+ with laboratory variables, including an extensive lipid profile. In a separate set of studies, hyperlipidaemia was induced in mice and in healthy humans via an oral lipid load, and blood Mg2+, triacylglycerol and NEFA concentrations were measured using colourimetric assays. In vitro, NEFAs harvested from albumin were added in increasing concentrations to several Mg2+-containing solutions to study the direct interaction between Mg2+ and NEFAs. RESULTS: In the cohort of overweight individuals, serum Mg2+ levels were inversely correlated with triacylglycerols incorporated in large VLDL particles (r = -0.159, p ≤ 0.01). After lipid loading, we observed a postprandial increase in plasma triacylglycerol and NEFA levels and a reciprocal reduction in blood Mg2+ concentration both in mice (Δ plasma Mg2+ -0.31 mmol/l at 4 h post oral gavage) and in healthy humans (Δ plasma Mg2+ -0.07 mmol/l at 6 h post lipid intake). Further, in vitro experiments revealed that the decrease in plasma Mg2+ may be explained by direct binding of Mg2+ to NEFAs. Moreover, Mg2+ was found to bind to albumin in a NEFA-dependent manner, evidenced by the fact that Mg2+ did not bind to fatty-acid-free albumin. The NEFA-dependent reduction in the free Mg2+ concentration was not affected by the presence of physiological concentrations of other cations. CONCLUSIONS/INTERPRETATION: This study shows that elevated NEFA and triacylglycerol levels directly reduce blood Mg2+ levels, in part explaining the high prevalence of hypomagnesaemia in metabolic disorders. We show that blood NEFA level affects the free Mg2+ concentration, and therefore, our data challenge how the fractional excretion of Mg2+ is calculated and interpreted in the clinic.

Magnesium deficiency prevents high-fat-diet-induced obesity in mice.

AIMS/HYPOTHESIS: Hypomagnesaemia (blood Mg2+ <0.7 mmol/l) is a common phenomenon in individuals with type 2 diabetes. However, it remains unknown how a low blood Mg2+ concentration affects lipid and energy metabolism. Therefore, the importance of Mg2+ in obesity and type 2 diabetes has been largely neglected to date. This study aims to determine the effects of hypomagnesaemia on energy homeostasis and lipid metabolism. METHODS: Mice (n = 12/group) were fed either a low-fat diet (LFD) or a high-fat diet (HFD) (10% or 60% of total energy) in combination with a normal- or low-Mg2+ content (0.21% or 0.03% wt/wt) for 17 weeks. Metabolic cages were used to investigate food intake, energy expenditure and respiration. Blood and tissues were taken to study metabolic parameters and mRNA expression profiles, respectively. RESULTS: We show that low dietary Mg2+ intake ameliorates HFD-induced obesity in mice (47.00 ± 1.53 g vs 38.62 ± 1.51 g in mice given a normal Mg2+-HFD and low Mg2+-HFD, respectively, p < 0.05). Consequently, fasting serum glucose levels decreased and insulin sensitivity improved in low Mg2+-HFD-fed mice. Moreover, HFD-induced liver steatosis was absent in the low Mg2+ group. In hypomagnesaemic HFD-fed mice, mRNA expression of key lipolysis genes was increased in epididymal white adipose tissue (eWAT), corresponding to reduced lipid storage and high blood lipid levels. Low Mg2+-HFD-fed mice had increased brown adipose tissue (BAT) Ucp1 mRNA expression and a higher body temperature. No difference was observed in energy expenditure between the two HFD groups. CONCLUSIONS/INTERPRETATION: Mg2+-deficiency abrogates HFD-induced obesity in mice through enhanced eWAT lipolysis and BAT activity.

Serum magnesium and the risk of prediabetes: a population-based cohort study.

AIMS/HYPOTHESIS: Previous studies have found an association between serum magnesium and incident diabetes; however, this association may be due to reverse causation, whereby diabetes may induce urinary magnesium loss. In contrast, in prediabetes (defined as impaired fasting glucose), serum glucose levels are below the threshold for urinary magnesium wasting and, hence, unlikely to influence serum magnesium levels. Thus, to study the directionality of the association between serum magnesium levels and diabetes, we investigated its association with prediabetes. We also investigated whether magnesium-regulating genes influence diabetes risk through serum magnesium levels. Additionally, we quantified the effect of insulin resistance in the association between serum magnesium levels and diabetes risk. METHODS: Within the population-based Rotterdam Study, we used Cox models, adjusted for age, sex, lifestyle factors, comorbidities, kidney function, serum levels of electrolytes and diuretic use, to study the association between serum magnesium and prediabetes/diabetes. In addition, we performed two mediation analyses: (1) to study if common genetic variation in eight magnesium-regulating genes influence diabetes risk through serum magnesium levels; and (2) to quantify the proportion of the effect of serum magnesium levels on diabetes that is mediated through insulin resistance (quantified by HOMA-IR). RESULTS: A total of 8555 participants (mean age, 64.7 years; median follow-up, 5.7 years) with normal glucose levels (mean ± SD: 5.46 ± 0.58 mmol/l) at baseline were included. A 0.1 mmol/l decrease in serum magnesium level was associated with an increase in diabetes risk (HR 1.18 [95% CI 1.04, 1.33]), confirming findings from previous studies. Of interest, a similar association was found between serum magnesium levels and prediabetes risk (HR 1.12 [95% CI 1.01, 1.25]). Genetic variation in CLDN19, CNNM2, FXYD2, SLC41A2, and TRPM6 significantly influenced diabetes risk (p < 0.05), and for CNNM2, FXYD2, SLC41A2 and TRPM6 this risk was completely mediated by serum magnesium levels. We found that 29.1% of the effect of serum magnesium levels on diabetes was mediated through insulin resistance, whereas for prediabetes 13.4% was mediated through insulin resistance. CONCLUSIONS/INTERPRETATION: Low serum magnesium levels are associated with an increased risk of prediabetes and this increased risk is similar to that of diabetes. Furthermore, common variants in magnesium-regulating genes modify diabetes risk through serum magnesium levels. Both findings support a potential causal role of magnesium in the development of diabetes, where the hypothesised pathway is partly mediated through insulin resistance.

Magnesium increases insulin-dependent glucose uptake in adipocytes.

Background: Type 2 diabetes (T2D) is characterized by a decreased insulin sensitivity. Magnesium (Mg2+) deficiency is common in people with T2D. However, the molecular consequences of low Mg2+ levels on insulin sensitivity and glucose handling have not been determined in adipocytes. The aim of this study is to determine the role of Mg2+ in the insulin-dependent glucose uptake. Methods: First, the association of low plasma Mg2+ with markers of insulin resistance was assessed in a cohort of 395 people with T2D. Secondly, the molecular role of Mg2+ in insulin-dependent glucose uptake was studied by incubating 3T3-L1 adipocytes with 0 or 1 mmol/L Mg2+ for 24 hours followed by insulin stimulation. Radioactive-glucose labelling, enzymatic assays, immunocytochemistry and live microscopy imaging were used to analyze the insulin receptor phosphoinositide 3-kinases/Akt pathway. Energy metabolism was assessed by the Seahorse Extracellular Flux Analyzer. Results: In people with T2D, plasma Mg2+ concentration was inversely associated with markers of insulin resistance; i.e., the lower Mg2+, the more insulin resistant. In Mg2+-deficient adipocytes, insulin-dependent glucose uptake was decreased by approximately 50% compared to control Mg2+condition. Insulin receptor phosphorylation Tyr1150/1151 and PIP3 mass were not decreased in Mg2+-deficient adipocytes. Live imaging microscopy of adipocytes transduced with an Akt sensor (FoxO1-Clover) demonstrated that FoxO1 translocation from the nucleus to the cytosol was reduced, indicting less Akt activation in Mg2+-deficient adipocytes. Immunocytochemistry using a Lectin membrane marker and at the membrane located Myc epitope-tagged glucose transporter 4 (GLUT4) demonstrated that GLUT4 translocation was diminished in insulin-stimulated Mg2+-deficient adipocytes compared to control conditions. Energy metabolism in Mg2+ deficient adipocytes was characterized by decreased glycolysis, upon insulin stimulation. Conclusions: Mg2+ increases insulin-dependent glucose uptake in adipocytes and suggests that Mg2+ deficiency may contribute to insulin resistance in people with T2D.

Lower plasma calcium associated with COVID-19, but not with disease severity: a two-centre retrospective cohort study.

BACKGROUND: Previous studies indicate hypocalcaemia as a potential diagnostic and prognostic marker of corona-virus disease 2019 (COVID-19). Our aim was to investigate these relations in more detail in a large test cohort and an independent validation cohort. METHODS: We retrospectively included 2792 COVID-19 suspected patients that presented to the emergency department (ED) of two hospitals. Plasma calcium and ionized plasma calcium levels were compared between COVID-19 positive and negative patients, and between severe and non-severe COVID-19 patients using univariate and multivariate analyses in the first hospital (N = 1363). Severe COVID-19 was defined as intensive care unit (ICU) admission or death within 28 d after admission. The results were validated by repeating the same analyses in the second hospital (N = 1429). RESULTS: A total of 693 (24.8%) of the enrolled patients were COVID-19 positive, of whom 238 (34.3%) had severe COVID-19. In both hospitals, COVID-19 positive patients had lower plasma calcium levels than COVID-19 negative patients, regardless of correction for albumin, in univariate and multivariate analysis (Δ0.06-0.13 mmol/L, p < .001). Ionized plasma calcium concentrations, with and without correction for pH, were also lower in COVID-19 positive patients in multivariate analyses (Δ0.02-0.05 mmol/L, N = 567, p < .001). However, we did not find a significant association between COVID-19 disease severity and plasma calcium in multivariate analyses. CONCLUSIONS: Plasma calcium concentrations were lower in COVID-19 positive than COVID-19 negative patients but we found no association with disease severity in multivariate analyses. Further understanding of plasma calcium perturbation may facilitate the development of new preventive and therapeutic modalities for the current pandemic.

Clinical prediction models for mortality in patients with covid-19: external validation and individual participant data meta-analysis.

OBJECTIVE: To externally validate various prognostic models and scoring rules for predicting short term mortality in patients admitted to hospital for covid-19. DESIGN: Two stage individual participant data meta-analysis. SETTING: Secondary and tertiary care. PARTICIPANTS: 46 914 patients across 18 countries, admitted to a hospital with polymerase chain reaction confirmed covid-19 from November 2019 to April 2021. DATA SOURCES: Multiple (clustered) cohorts in Brazil, Belgium, China, Czech Republic, Egypt, France, Iran, Israel, Italy, Mexico, Netherlands, Portugal, Russia, Saudi Arabia, Spain, Sweden, United Kingdom, and United States previously identified by a living systematic review of covid-19 prediction models published in The BMJ, and through PROSPERO, reference checking, and expert knowledge. MODEL SELECTION AND ELIGIBILITY CRITERIA: Prognostic models identified by the living systematic review and through contacting experts. A priori models were excluded that had a high risk of bias in the participant domain of PROBAST (prediction model study risk of bias assessment tool) or for which the applicability was deemed poor. METHODS: Eight prognostic models with diverse predictors were identified and validated. A two stage individual participant data meta-analysis was performed of the estimated model concordance (C) statistic, calibration slope, calibration-in-the-large, and observed to expected ratio (O:E) across the included clusters. MAIN OUTCOME MEASURES: 30 day mortality or in-hospital mortality. RESULTS: Datasets included 27 clusters from 18 different countries and contained data on 46 914patients. The pooled estimates ranged from 0.67 to 0.80 (C statistic), 0.22 to 1.22 (calibration slope), and 0.18 to 2.59 (O:E ratio) and were prone to substantial between study heterogeneity. The 4C Mortality Score by Knight et al (pooled C statistic 0.80, 95% confidence interval 0.75 to 0.84, 95% prediction interval 0.72 to 0.86) and clinical model by Wang et al (0.77, 0.73 to 0.80, 0.63 to 0.87) had the highest discriminative ability. On average, 29% fewer deaths were observed than predicted by the 4C Mortality Score (pooled O:E 0.71, 95% confidence interval 0.45 to 1.11, 95% prediction interval 0.21 to 2.39), 35% fewer than predicted by the Wang clinical model (0.65, 0.52 to 0.82, 0.23 to 1.89), and 4% fewer than predicted by Xie et al's model (0.96, 0.59 to 1.55, 0.21 to 4.28). CONCLUSION: The prognostic value of the included models varied greatly between the data sources. Although the Knight 4C Mortality Score and Wang clinical model appeared most promising, recalibration (intercept and slope updates) is needed before implementation in routine care.

Deep learning with robustness to missing data: A novel approach to the detection of COVID-19.

In the context of the current global pandemic and the limitations of the RT-PCR test, we propose a novel deep learning architecture, DFCN (Denoising Fully Connected Network). Since medical facilities around the world differ enormously in what laboratory tests or chest imaging may be available, DFCN is designed to be robust to missing input data. An ablation study extensively evaluates the performance benefits of the DFCN as well as its robustness to missing inputs. Data from 1088 patients with confirmed RT-PCR results are obtained from two independent medical facilities. The data includes results from 27 laboratory tests and a chest x-ray scored by a deep learning model. Training and test datasets are taken from different medical facilities. Data is made publicly available. The performance of DFCN in predicting the RT-PCR result is compared with 3 related architectures as well as a Random Forest baseline. All models are trained with varying levels of masked input data to encourage robustness to missing inputs. Missing data is simulated at test time by masking inputs randomly. DFCN outperforms all other models with statistical significance using random subsets of input data with 2-27 available inputs. When all 28 inputs are available DFCN obtains an AUC of 0.924, higher than any other model. Furthermore, with clinically meaningful subsets of parameters consisting of just 6 and 7 inputs respectively, DFCN achieves higher AUCs than any other model, with values of 0.909 and 0.919.

Performance of commercially-available cholesterol self-tests.

BACKGROUND: Hypercholesterolemia (plasma cholesterol concentration ≥5.2 mmol/L) is a risk factor for cardiovascular disease and stroke. Many different cholesterol self-tests are readily available at general stores, pharmacies and web shops. However, there is limited information on their analytical and diagnostic performance. METHODS: We included 62 adult patients who required a lipid panel measurement (cholesterol, high-density lipoprotein (HDL), triglycerides and LDLcalc) for routine care. The performance of five different cholesterol self-tests, three quantitative meters (Roche Accutrend Plus, Mission 3-in-1 and Qucare) and two semi-quantitative strip tests (Veroval and Mylan MyTest), was assessed according to the manufacturers' protocol. RESULTS: The average plasma cholesterol concentration was 5.2 ± 1.2 mmol/L. The mean absolute relative difference (MARD) of the five cholesterol self-tests ranged from 6 ± 5% (Accutrend Plus) to 20 ± 12% (Mylan Mytest). The Accutrend Plus cholesterol meter showed the best diagnostic performance with a 92% sensitivity and 89% specificity. The Qucare and Mission 3-in-1 are able to measure HDL concentrations and can thus provide a cholesterol:HDL ratio. The Passing-Bablok regression analyses for the ratio showed poor performance in both self-tests (Mission 3-in-1: y = 1.62x-1.20; Qucare: y = 0.61x + 1.75). The Accutrend Plus is unable to measure the plasma high-density lipoprotein concentration.Conclusions/interpretation: The Accutrend Plus cholesterol meter (Roche) had excellent diagnostic and analytic performance. However, several of the commercially-available self-tests had considerably poor accuracy and diagnostic performance and therefore do not meet the required qualifications, potentially leading to erroneous results. Better regulation, standardization and harmonization of cholesterol self-tests is warranted.