From microscope to micropixels: A rapid review of artificial intelligence for the peripheral blood film.

Artificial intelligence (AI) and its application in classification of blood cells in the peripheral blood film is an evolving field in haematology. We performed a rapid review of the literature on AI and peripheral blood films, evaluating the condition studied, image datasets, machine learning models, training set size, testing set size and accuracy. A total of 283 studies were identified, encompassing 6 broad domains: malaria (n = 95), leukemia (n = 81), leukocytes (n = 72), mixed (n = 25), erythrocytes (n = 15) or Myelodysplastic syndrome (MDS) (n = 1). These publications have demonstrated high self-reported mean accuracy rates across various studies (95.5% for malaria, 96.0% for leukemia, 94.4% for leukocytes, 95.2% for mixed studies and 91.2% for erythrocytes), with an overall mean accuracy of 95.1%. Despite the high accuracy, the challenges toward real world translational usage of these AI trained models include the need for well-validated multicentre data, data standardisation, and studies on less common cell types and non-malarial blood-borne parasites.

Commutability assessment of human urine certified reference materials for albumin and creatinine on multiple clinical analyzers using different statistical models.

Urine albumin concentration and albumin-creatinine ratio are important for the screening of early-stage kidney damage. Commutable urine certified reference materials (CRMs) for albumin and creatinine are necessary for standardization of urine albumin and accurate measurement of albumin-urine ratio. Two urine CRMs for albumin and creatinine with certified values determined using higher-order reference measurement procedures were evaluated for their commutability on five brands/models of clinical analyzers where different reagent kits were used, including Roche Cobas c702, Roche Cobas c311, Siemens Atellica CH, Beckman Coulter AU5800, and Abbott Architect c16000. The commutability study was conducted by measuring at least 26 authentic patient urine samples and the human urine CRMs using both reference measurement procedures and the routine methods. Both the linear regression model suggested by the Clinical and Laboratory Standard Institute (CLSI) guidelines and log-transformed model recommended by the International Federation of Clinical Chemistry and Laboratory Medicine (IFCC) Commutability Working Group were used to evaluate the commutability of the human urine CRMs. The commutability of the human urine CRMs was found to be generally satisfactory on all five clinical analyzers for both albumin and creatinine, suggesting that they are suitable to be used routinely by clinical laboratories as quality control or for method validation of urine albumin and creatinine measurements.

Does the use of fish oil-based lipid emulsion in the clinical setting of total parenteral nutrition and lipid rescue therapy interfere with common laboratory analytes on Roche Cobas 6000?

BACKGROUND: Lipaemic interference on automated analysers has been widely studied using soy-based emulsion such as Intralipid. Due to the greater adoption of fish oil-based lipid emulsion for total parenteral nutrition in view of improved clinical outcomes, we seek to characterize the optical properties of SMOFlipid 20% (Fresenius Kabi, Bad Homburg, Germany), a fish oil-based emulsion, on the Roche Cobas 6000 chemistry analyser (Roche Diagnostic, Basel, Switzerland). METHOD: Various amounts of SMOFlipid were spiked into pooled serums. We plotted Roche Cobas Serum Index Gen.2 Lipaemia Index (L-index) against the amount of SMOFlipid added. We then studied the interference thresholds for aspartate aminotransferase, alanine aminotransferase, albumin and renal panel analytes using SMOFlipid. We subjected five levels of spiked lipaemia to high-speed centrifugation and analysed the specimens pre- and post-centrifugation. To postulate whether fish oil-based lipid emulsion interferes with laboratory results in the clinical setting, we calculated concentrations of SMOFlipid post-lipid rescue therapy and steady-state concentration of a typical total parenteral nutrition regime using pharmacokinetic principles. RESULTS: SMOFlipid optical behaviour is similar to Intralipid using the Serum Index Gen.2 L-index, with 1 mg/dL of SMOFlipid representing 1 unit of L-index. Manufacturer-stated interference thresholds are accurate for alanine aminotransferase, aspartate aminotransferase, albumin, urea and creatinine. High-speed centrifugation at 60 min 21,100g facilitates the removal of fish oil-based SMOFlipid. CONCLUSION: Based on the interference thresholds we verified and pharmacokinetics parameters provided by SMOFlipid manufacturer, total parenteral nutrition may not interfere with chemistry analytes given sufficient clearance, but lipid rescue therapy will interfere. Further studies assessing lipaemic interference on immunoassays are needed.

Commutable whole blood reference materials for hemoglobin A1c validated on multiple clinical analyzers.

Background The measurement of hemoglobin A1c (HbA1c) is important for diagnosing diabetes mellitus as well as assessing glycemic control in diabetic patients. Commutable whole blood certified reference materials (CRMs) are needed in the measurement of HbA1c for method validation and/or as quality controls. Methods We developed three levels of hemolyzed whole blood CRMs for HbA1c. The certified values were determined using liquid chromatography-isotope dilution tandem mass spectrometry method (LC-IDMS/MS) where two "signature" hexapeptides of HbA1c and hemoglobin A0 (HbA0) were used as the calibration standards. The concentrations of the hexapeptide solutions were determined by amino acid analysis by the LC-IDMS/MS method using amino acid CRMs as the calibration standards. The commutability study was conducted by measuring 25 patient specimens and the whole blood CRMs by both LC-IDMS/MS method and various routine methods using six different clinical analyzers. Results The certified values were determined to be 35.1±2.0, 50.3±1.9 and 65.8±2.6 mmol/mol, respectively. These CRMs showed good commutability on five of the six clinical analyzers but showed poor commutability on one of the clinical analyzers that used similar method as two other analyzers where good commutability was observed. Conclusions With certified target values based on metrological traceability and good commutability on most of the clinical analyzers, the developed whole blood CRMs can be used for method validation or as quality control materials in the measurement of HbA1c. The commutability study results also underscored the need of commutability testing of clinical CRMs using various clinical analyzers.

Association of circulating proinflammatory marker, leucine-rich-α2-glycoprotein (LRG1), following metabolic/bariatric surgery.

BACKGROUND: Obesity confers substantial excess risk for morbidity and mortality, especially for type 2 diabetes (T2D). Leucine-rich-α2-glycoprotein 1 (LRG1), a novel proinflammatory factor, was recently reported to be higher in patients with T2D with complications of peripheral arterial disease. Association of LRG1, obesity, and weight loss is unknown. We examined whether plasma LRG1 is associated with obesity in health screening participants and if it predicts future weight loss in morbidly obese patients after metabolic/bariatric surgery. METHODS: Cohort 1 was a cross-sectional study from a Health Screening program (n = 616) in a tertiary hospital. Cohort 2 was a prospective study of morbidly obese patients (n = 231) who underwent metabolic/bariatric surgery with follow-up weight measurements. Anthropometric data, baseline fasting glucose, plasma adiponectin, high sensitivity C-reactive protein (HsCRP), and LRG1 were measured. Postsurgery blood, after metabolic/bariatric surgery, were available for LRG1and HsCRP measurements in 57 patients. RESULTS: In the group with highest tertile of LRG1, body mass index (BMI), waist circumference, and HsCRP were significantly higher, while total cholesterol, high-density lipoprotein, low-density lipoprotein, and adiponectin were lower than tertiles 1 and 2. Generalized linear model analysis showed that female gender (P < .0001), non-Chinese ethnicity (P < .019), and higher HsCRP (P < .0001) levels were independent and significant determinants of higher plasma LRG1 levels. After adjustment for age, gender, ethnicity, and baseline BMI, female gender (P = .020), higher presurgery BMI (P = .001), and lower presurgery LRG1 (P = .002) remained statistically significant predictors for greater weight loss. Plasma LRG1 increased significantly [from 28.2 (21.9-36.8) to 34.9 (22.6-49.5)] µg/mL (P = .003) within 1.5 months, after metabolic/bariatric surgery. CONCLUSIONS: Our study demonstrates that LRG1 level is positively associated with obesity and a lower level of plasma LRG1 predicts weight loss in metabolic/bariatric surgery. Our novel findings suggest LRG1, itself or in combination with other known factors, is a potential biomarker of inflammation and obesity.

Relationship between measured average glucose by continuous glucose monitor and HbA1c measured by three different routine laboratory methods.

OBJECTIVES: The relationship between glycated hemoglobin A1c (HbA1c) and average glucose has been described by the empirically derived estimated average glucose (eAG) equation in the A1c-Derived Average Glucose (ADAG) study, with extensive calibration efforts in four secondary reference HbA1c methods. It is not known if this relationship is preserved when HbA1c is measured by routine laboratory methods under routine conditions. DESIGN AND METHODS: We measured average glucose (mAG) by six days of continuous glucose monitoring in 45 adults with stable HbA1c (<1% HbA1c change in the preceding two months). Subjects with medical conditions that may confound HbA1c measurement, including anemia and hemoglobinopathy, were excluded. HbA1c was measured using Bio-Rad Variant II (cation-exchange HPLC), Bio-Rad in2it (boronate affinity HPLC) and Roche Tina-quant (immunoassay) methods. RESULTS: The average differences between eAG derived from the routine HbA1c methods and mAG were 10.4% (Variant II), 6.0% (Tina-quant) and 1.0% (in2it). The regression coefficients between the mAG and HbA1c are different between in2it (mAG, mmol/L=0.58 × %HbA1c + 2.3), Tina-quant and Variant II (both mAG, mmol/L=0.66 × %HbA1c + 1.9). However, the 95% confidence intervals of the slope and bias of these methods overlap. The correlation between mAG and HbA1c was greatest when measured using the Variant II (r(2)=0.84), followed by Tina-quant (r(2)=0.82) and in2it (r(2)=0.71). CONCLUSIONS: The relationship between HbA1c and measured average glucose is method-dependent despite improved HbA1c standardization. The differences in relationship may reflect as discrepant eAG and home glucose monitoring results.

Immunoglobulin-associated creatine kinase masquerading as macro-creatine kinase type 2 in a statin user.

Macro-creatine kinase (CK) is a cause of falsely elevated CK. Macro-CK type 1 is immunoglobulin-associated CK; type 2 is polymeric mitochondrial-CK. An elderly asymptomatic lady had an elevated CK level after receiving statin therapy. Her CK gel electrophoresis analysis demonstrated coexisting macro-CK type 1 and type 2 patterns. Further analysis by immunofixation and mixing this patient's serum with CK control material revealed an IgG-associated macro-CK that mimicked the electrophoretic pattern of macro-CK type 2. This highly unusual discovery suggests the possibility of the misinterpretation of macro-CK type 1 as macro-CK type 2. Falsely elevated CK is still common despite modern laboratory instrumentation and should be investigated.

Lowering the criterion for impaired fasting glucose: impact on disease prevalence and associated risk of diabetes and ischemic heart disease.

OBJECTIVE: To determine the effect of lowering the fasting plasma glucose (FPG) criterion for impaired fasting glucose (IFG) on the prevalence of IFG, the risks of diabetes, and cardiovascular disease (CVD) associated with IFG. RESEARCH DESIGN AND METHODS: Three studies were used: 1). the 1998 National Health Survey (NHS98), a randomly selected cross-sectional sample of 4723 subjects; 2). the Singapore Impaired Glucose Tolerance (IGT) Follow-up Study, a cohort study comprising 295 IGT and 292 normal glucose tolerance subjects (frequency matched for age, sex, and ethnic group) followed up from 1992 to 2000; and 3). the Singapore CVD Cohort Study, comprising 5920 subjects from three cross-sectional studies in whom the first ischemic heart disease (IHD) event was identified through linkage to registry databases. Risk of diabetes (Singapore IGT Follow-up study) was estimated using logistic regression adjusted for age, sex, and ethnicity. Risk of IHD (Singapore CVD cohort) was estimated using stratified (by study, from which data were derived) Cox's proportional hazards models adjusted for age, sex, and ethnicity. RESULTS: Lowering the criterion for diagnosing IFG to 5.6 mmol/l increased the prevalence of IFG from 9.5 to 32.3% in the NHS98. The lower cutoff identified more subjects at risk of diabetes and IHD, but the relative risk was lower than that for IGT. CONCLUSIONS: Greater efforts to identify those with IGT, or a group at similar risk of diabetes and CVD, may be a more efficient public health measure than lowering the FPG criterion for diagnosing IFG.

The Singapore impaired glucose tolerance follow-up study: does the ticking clock go backward as well as forward?

OBJECTIVE: To 1). document the change in glucose tolerance for subjects with normal glucose tolerance (NGT) and impaired glucose tolerance (IGT) over time, 2). identify baseline factors associated with worsening of glucose tolerance, and 3). determine whether cardiovascular disease (CVD) risk factors associated with IGT improved in tandem with glucose tolerance. RESEARCH DESIGN: Subjects with IGT and NGT (matched for age, sex, and ethnic group) were identified from a cross-sectional survey conducted in 1992. Subjects with IGT (297) and NGT (298) (65.0%) were reexamined in 2000. Glucose tolerance (assessed by 75-g oral glucose tolerance test), anthropometric data, serum lipids, blood pressure, and insulin resistance were determined at baseline and at the follow-up examination. RESULTS: For NGT subjects, 14.0% progressed to IGT and 4.3% to diabetes over 8 years. For IGT subjects, 41.4% reverted to NGT, 23.0% remained impaired glucose tolerant, and 35.1% developed diabetes. Obesity, hypertriglyceridemia, higher blood pressure, increased insulin resistance, and lower HDL cholesterol at baseline were associated with worsening of glucose tolerance in both IGT and NGT subjects. Those with IGT who reverted to NGT remained more obese and had higher blood pressure than those with NGT in both 1992 and 2000. However, serum triglyceride, HDL cholesterol, and insulin resistance values in 2000 became indistinguishable from those of subjects who maintained NGT throughout the study period. CONCLUSIONS: Some, but not all, CVD risk factors associated with IGT and with the risk of future diabetes normalize when glucose tolerance normalizes. Continued surveillance and treatment in subjects with IGT, even after they revert to NGT, may be important in the prevention of CVD.