Gamma-Glutamyl Transferase (GGT) Is the Leading External Quality Assurance Predictor of ISO15189 Compliance for Pathology Laboratories.

Pathology results are central to modern medical practice, informing diagnosis and patient management. To ensure high standards from pathology laboratories, regulators require compliance with international and local standards. In Australia, the monitoring and regulation of medical laboratories are achieved by conformance to ISO15189-National Pathology Accreditation Advisory Council standards, as assessed by the National Association of Testing Authorities (NATA), and an external quality assurance (EQA) assessment via the Royal College of Pathologists of Australasia Quality Assurance Program (RCPAQAP). While effective individually, integration of data collected by NATA and EQA testing promises advantages for the early detection of technical or management problems in the laboratory, and enhanced ongoing quality assessment. Random forest (RF) machine learning (ML) previously identified gamma-glutamyl transferase (GGT) as a leading predictor of NATA compliance condition reporting. In addition to further RF investigations, this study also deployed single decision trees and support vector machines (SVM) models that included creatinine, electrolytes and liver function test (LFT) EQA results. Across all analyses, GGT was consistently the top-ranked predictor variable, validating previous observations from Australian laboratories. SVM revealed broad patterns of predictive EQA marker interactions with NATA outcomes, and the distribution of GGT relative deviation suggested patterns by which to identify other strong EQA predictors of NATA outcomes. An integrated model of pathology quality assessment was successfully developed, via the prediction of NATA outcomes by EQA results. GGT consistently ranked as the best predictor variable, identified by combining recursive partitioning and SVM ML strategies.

Evolving Epidemiology of Reported Giardiasis Cases in the United States, 1995-2016.

BACKGROUND: Giardiasis is the most common intestinal parasitic disease of humans identified in the United States (US) and an important waterborne disease. In the United States, giardiasis has been variably reportable since 1992 and was made a nationally notifiable disease in 2002. Our objective was to describe the epidemiology of US giardiasis cases from 1995 through 2016 using National Notifiable Diseases Surveillance System data. METHODS: Negative binomial regression models were used to compare incidence rates by age group (0-4, 5-9, 10-19, 20-29, 30-39, 40-49, 50-64, and ≥ 65 years) during 3 time periods (1995-2001, 2002-2010, and 2011-2016). RESULTS: During 1995-2016, the average number of reported cases was 19 781 per year (range, 14 623-27 778 cases). The annual incidence of reported giardiasis in the United States decreased across all age groups. This decrease differs by age group and sex and may reflect either changes in surveillance methods (eg, changes to case definitions or reporting practices) or changes in exposure. Incidence rates in males and older age groups did not decrease to the same extent as rates in females and children. CONCLUSIONS: Trends suggest that differences in exposures by sex and age group are important to the epidemiology of giardiasis. Further investigation into the risk factors of populations with higher rates of giardiasis will support prevention and control efforts.

Rethinking ME/CFS Diagnostic Reference Intervals via Machine Learning, and the Utility of Activin B for Defining Symptom Severity.

Biomarker discovery applied to myalgic encephalomyelitis/chronic fatigue syndrome (ME/CFS), a disabling disease of inconclusive aetiology, has identified several cytokines to potentially fulfil a role as a quantitative blood/serum marker for laboratory diagnosis, with activin B a recent addition. We explored further the potential of serum activin B as a ME/CFS biomarker, alone and in combination with a range of routine test results obtained from pathology laboratories. Previous pilot study results showed that activin B was significantly elevated for the ME/CFS participants compared to healthy (control) participants. All the participants were recruited via CFS Discovery and assessed via the Canadian/International Consensus Criteria. A significant difference for serum activin B was also detected for ME/CFS and control cohorts recruited for this study, but median levels were significantly lower for the ME/CFS cohort. Random Forest (RF) modelling identified five routine pathology blood test markers that collectively predicted ME/CFS at ≥62% when compared via weighted standing time (WST) severity classes. A closer analysis revealed that the inclusion of activin B to the panel of pathology markers improved the prediction of mild to moderate ME/CFS cases. Applying correct WST class prediction from RFA modelling, new reference intervals were calculated for activin B and associated pathology markers, where 24-h urinary creatinine clearance, serum urea and serum activin B showed the best potential as diagnostic markers. While the serum activin B results remained statistically significant for the new participant cohorts, activin B was found to also have utility in enhancing the prediction of symptom severity, as represented by WST class.

Weighting of orthostatic intolerance time measurements with standing difficulty score stratifies ME/CFS symptom severity and analyte detection.

BACKGROUND: Myalgic Encephalomyelitis/Chronic Fatigue Syndrome (ME/CFS) is clinically defined and characterised by persistent disabling tiredness and exertional malaise, leading to functional impairment. METHODS: This study introduces the weighted standing time (WST) as a proxy for ME/CFS severity, and investigates its behaviour in an Australian cohort. WST was calculated from standing time and subjective standing difficulty data, collected via orthostatic intolerance assessments. The distribution of WST for healthy controls and ME/CFS patients was correlated with the clinical criteria, as well as pathology and cytokine markers. Included in the WST cytokine analyses were activins A and B, cytokines causally linked to inflammation, and previously demonstrated to separate ME/CFS from healthy controls. Forty-five ME/CFS patients were recruited from the CFS Discovery Clinic (Victoria) between 2011 and 2013. Seventeen healthy controls were recruited concurrently and identically assessed. RESULTS: WST distribution was significantly different between ME/CFS participants and controls, with six diagnostic criteria, five analytes and one cytokine also significantly different when comparing severity via WST. On direct comparison of ME/CFS to study controls, only serum activin B was significantly elevated, with no significant variation observed for a broad range of serum and urine markers, or other serum cytokines. CONCLUSIONS: The enhanced understanding of standing test behaviour to reflect orthostatic intolerance as a ME/CFS symptom, and the subsequent calculation of WST, will encourage the greater implementation of this simple test as a measure of ME/CFS diagnosis, and symptom severity, to the benefit of improved diagnosis and guidance for potential treatments.

Enhancement of hepatitis virus immunoassay outcome predictions in imbalanced routine pathology data by data balancing and feature selection before the application of support vector machines.

BACKGROUND: Data mining techniques such as support vector machines (SVMs) have been successfully used to predict outcomes for complex problems, including for human health. Much health data is imbalanced, with many more controls than positive cases. METHODS: The impact of three balancing methods and one feature selection method is explored, to assess the ability of SVMs to classify imbalanced diagnostic pathology data associated with the laboratory diagnosis of hepatitis B (HBV) and hepatitis C (HCV) infections. Random forests (RFs) for predictor variable selection, and data reshaping to overcome a large imbalance of negative to positive test results in relation to HBV and HCV immunoassay results, are examined. The methodology is illustrated using data from ACT Pathology (Canberra, Australia), consisting of laboratory test records from 18,625 individuals who underwent hepatitis virus testing over the decade from 1997 to 2007. RESULTS: Overall, the prediction of HCV test results by immunoassay was more accurate than for HBV immunoassay results associated with identical routine pathology predictor variable data. HBV and HCV negative results were vastly in excess of positive results, so three approaches to handling the negative/positive data imbalance were compared. Generating datasets by the Synthetic Minority Oversampling Technique (SMOTE) resulted in significantly more accurate prediction than single downsizing or multiple downsizing (MDS) of the dataset. For downsized data sets, applying a RF for predictor variable selection had a small effect on the performance, which varied depending on the virus. For SMOTE, a RF had a negative effect on performance. An analysis of variance of the performance across settings supports these findings. Finally, age and assay results for alanine aminotransferase (ALT), sodium for HBV and urea for HCV were found to have a significant impact upon laboratory diagnosis of HBV or HCV infection using an optimised SVM model. CONCLUSIONS: Laboratories looking to include machine learning via SVM as part of their decision support need to be aware that the balancing method, predictor variable selection and the virus type interact to affect the laboratory diagnosis of hepatitis virus infection with routine pathology laboratory variables in different ways depending on which combination is being studied. This awareness should lead to careful use of existing machine learning methods, thus improving the quality of laboratory diagnosis.

The kinetics of haemoglobin and ferritin in longitudinal community patients with iron deficiency or hypoxia.

BACKGROUND: Red cell distribution width (RDW) is well recognised as a marker of iron-deficient anaemia, as well as useful to the distinction between some anaemic states. A role in the prediction of patient mortality and for the laboratory diagnosis of organ dysfunction has been also investigated. RDW has recently been suggested as a marker of acute and chronic hypoxia. METHODS: In this paper we use RDW kinetics to identify different patient groups and then investigate the relationship between RDW, ferritin and haemoglobin kinetics in a large cross-sectional community patient dataset. RESULTS: A novel mathematical model of this relationship is developed that captures all aspects of variation in the data. A linear regression of RDW/log(ferritin) on days is combined with a multi-level random structure including random intercepts and slopes for each patient. CONCLUSIONS: No evidence of an age affect was found in the data. On the other hand, significant patterns in the rises and falls of log(ferritin) and haemoglobin with RDW over time are identified.

A simulation model to estimate the risk of transfusion-transmitted arboviral infection.

BACKGROUND: The arboviruses West Nile virus (WNV), dengue virus (DENV) and Ross River virus (RRV) have been demonstrated to be blood transfusion-transmissible. A model to estimate the risk of WNV to the blood supply using a Monte Carlo approach has been developed and also applied to Chikungunya virus. Also, a probabilistic model was developed to assess the risk of DENV to blood safety, which was later adapted to RRV. To address efficacy and limitations within each model we present a hybrid model that promises improved accuracy, and is broadly applicable to assess the risk of arboviral transmission by blood transfusion. MATERIAL AND METHODS: Data were drawn from the Cairns Public Health Unit (Australia) and published literature. Based on the published models and using R code, a novel 'combined' model was developed and validated against the BP model using sensitivity testing. RESULTS: The mean risk per 10,000 of the combined model is 0.98 with a range from 0.79 to 1.25, while the maximum risk was 4.45 ranging from 2.62 to 7.67 respectively. These parameters for the BP model were 1.20 ranging from 0.84 to 1.55, and 2.86 ranging from 1.33 to 5.23 respectively. CONCLUSION: The combined simulation model is simple and robust. We propose it can be applied as a 'generic' arbovirus model to assess the risk from known or novel arboviral threats to the blood supply.

Assessment of machine-learning techniques on large pathology data sets to address assay redundancy in routine liver function test profiles.

BACKGROUND: Routine liver function tests (LFTs) are central to serum testing profiles, particularly in community medicine. However there is concern about the redundancy of information provided to requesting clinicians. Large quantities of clinical laboratory data and advances in computational knowledge discovery methods provide opportunities to re-examine the value of individual routine laboratory results that combine for LFT profiles. METHODS: The machine learning methods recursive partitioning (decision trees) and support vector machines (SVMs) were applied to aggregate clinical chemistry data that included elevated LFT profiles. Response categories for γ-glutamyl transferase (GGT) were established based on whether the patient results were within or above the sex-specific reference interval. Single decision tree and SVMs were applied to test the accuracy of GGT prediction by the highest ranked predictors of GGT response, alkaline phosphatase (ALP) and alanine amino-transaminase (ALT). RESULTS: Through interrogating more than 20,000 individual cases comprising both sexes and all ages, decision trees predicted GGT category at 90% accuracy using only ALP and ALT, with a SVM prediction accuracy of 82.6% after 10-fold training and testing. Bilirubin, lactate dehydrogenase (LD) and albumin did not enhance prediction, or reduced accuracy. Comparison of abnormal (elevated) GGT categories also supported the primacy of ALP and ALT as screening markers, with serum urate and cholesterol also useful. CONCLUSIONS: Machine-learning interrogation of massive clinical chemistry data sets demonstrated a strategy to address redundancy in routine LFT screening by identifying ALT and ALP in tandem as able to accurately predict GGT elevation, suggesting that GGT can be removed from routine LFT screening.

The early detection of anaemia and aetiology prediction through the modelling of red cell distribution width (RDW) in cross-sectional community patient data.

BACKGROUND: Red cell distribution width (RDW) is a marker of iron-deficient anaemia that can also assist differentiation of other anaemias. RDW also has been suggested as an effective marker for earlier anaemia detection. The RDW-anaemia relationship was investigated in cross-sectional community patient data, and the capacity of RDW to predict the diagnostic value of second tier anaemia markers assessed. METHODS: Routine and second tier assay data were provided by the laboratory Sullivan Nicolaides Pathology. The cohort was divided into male and female groups stratified by age, and correlation analyses assessed associations of RDW to haemoglobin and ferritin. Analysis of covariance (ANCOVA) was performed for both routine and second tier markers to investigate their significance for RDW prediction. RESULTS: RDW had statistically significant negative correlation with haemoglobin for both sexes and age ranges (p<0.01). The RDW relationship with serum ferritin was non-linear, representing two populations. ANCOVA showed categorical ferritin as a significant RDW predictor for younger females, with vitamin B12 a significant RDW predictor for older men. Haemoglobin, mean corpuscular haemoglobin (MCH) and second tier iron markers (e.g., transferrin) were significant RDW predictors for both sexes and ages investigated. An individual longitudinal female case study showed RDW as very sensitive to haemoglobin decrease, with ferritin not as responsive. CONCLUSIONS: RDW had a significant negative association with haemoglobin in cross-sectional community patient data. ANCOVA showed ferritin as a significant RDW predictor for younger females only. This study confirms the utility of RDW as a marker for early anaemia detection, and useful to accelerated diagnoses of anaemia aetiology.

An 1H-MRS framework predicts the onset of Alzheimer's disease symptoms in PSEN1 mutation carriers.

BACKGROUND: Alzheimer's disease (AD) is the most common cause of dementia; the main risk factors are age and several recently identified genes. A major challenge for AD research is the early detection of subjects at risk. The aim of this study is to develop a predictive model using proton magnetic resonance spectroscopy (1H-MRS), a noninvasive technique that evaluates brain chemistry in vivo, for monitoring the clinical outcome of carriers of a fully penetrant mutation that causes AD. METHODS: We studied 75 subjects from the largest multigenerational pedigree in the world (∼5000 people) that segregates a unique form of early-onset Alzheimer's disease (EOAD) caused by a fully penetrant mutation in the Presenilin-1 gene (PSEN1 p.Glu280Ala [E280 A]). Forty-four subjects were carriers of the mutation, and 31 were noncarriers. Seventeen carriers had either mild cognitive impairment (MCI) or early-stage AD (collectively MCI-AD). In right and left parietal white mater and parasagittal parietal gray matter (RPPGM and LPPGM) of the posterior cingulate gyrus and precuneus, we measured levels of the brain metabolites N-acetylaspartate (NAA), inositol (Ins), choline (Cho), and glutamate-glutamine complex (Glx) relative to creatine (Cr) levels (NAA/Cr, Ins/Cr, Cho/Cr, and Glx/Cr, respectively) with two-dimensional 1H-MRS. Using advanced recursive partition analysis and random forest analysis, we built classificatory decision trees for both mutation carrier status and the presence of MCI-AD symptoms, fitting them to 1H-MRS data while controlling for age, educational level, and sex. RESULTS: We found that (1) the combination of LPPGM Cho/Cr<0.165 and RPPGM Glx/Cr>1.54 fully excluded carriers; (2) LPPGM Cho/Cr>0.165, RPPGM Glx/Cr<1.54, and left parietal white mater NAA/Cr>1.16 identified asymptomatic carriers with sensitivity of 97.7% and specificity of 77.4%; and (3) RPPGM NAA/Cr>1.05 defined asymptomatic subjects (independent of carrier status) with sensitivity of 100% and a specificity of 96.6%. CONCLUSIONS: Brain metabolites measured by 1H-MRS in the posterior cingulate gyrus and precuneus are optimally sensitive and specific potential noninvasive biomarkers of subclinical emergence of AD caused by the PSEN1 p.Glu280Ala (E280 A) mutation.

Infection status outcome, machine learning method and virus type interact to affect the optimised prediction of hepatitis virus immunoassay results from routine pathology laboratory assays in unbalanced data.

BACKGROUND: Advanced data mining techniques such as decision trees have been successfully used to predict a variety of outcomes in complex medical environments. Furthermore, previous research has shown that combining the results of a set of individually trained trees into an ensemble-based classifier can improve overall classification accuracy. This paper investigates the effect of data pre-processing, the use of ensembles constructed by bagging, and a simple majority vote to combine classification predictions from routine pathology laboratory data, particularly to overcome a large imbalance of negative Hepatitis B virus (HBV) and Hepatitis C virus (HCV) cases versus HBV or HCV immunoassay positive cases. These methods were illustrated using a never before analysed data set from ACT Pathology (Canberra, Australia) relating to HBV and HCV patients. RESULTS: It was easier to predict immunoassay positive cases than negative cases of HBV or HCV. While applying an ensemble-based approach rather than a single classifier had a small positive effect on the accuracy rate, this also varied depending on the virus under analysis. Finally, scaling data before prediction also has a small positive effect on the accuracy rate for this dataset. A graphical analysis of the distribution of accuracy rates across ensembles supports these findings. CONCLUSIONS: Laboratories looking to include machine learning as part of their decision support processes need to be aware that the infection outcome, the machine learning method used and the virus type interact to affect the enhanced laboratory diagnosis of hepatitis virus infection, as determined by primary immunoassay data in concert with multiple routine pathology laboratory variables. This awareness will lead to the informed use of existing machine learning methods, thus improving the quality of laboratory diagnosis via informatics analyses.