Fifteen-minute Frequency of Glucose Measurements and the Use of Threshold Alarms: Impact on Mitigating Dysglycemia in Critically Ill Patients.

BACKGROUND: The use of near-continuous blood glucose (BG) monitoring has the potential to improve glycemic control in critically ill patients. The MANAGE IDE trial evaluated the performance of the OptiScanner (OS) 5000 in a multicenter cohort of 200 critically ill patients. METHODS: An Independent Group reviewed the BG run charts of all 200 patients and voted whether unblinded use of the OS, with alarms set at 90 and 130 to 150 mg/dL to alert the clinical team to impending hypoglycemia and hyperglycemia, respectively, would have eliminated episodes of dysglycemia: hypoglycemia, defined as a single BG <70 mg/dL; hyperglycemia, defined as >4 hours of BG >150 mg/dL; severe hyperglycemia, defined as >4 hours of BG >200 mg/dL and increased glucose variability (GV), defined as coefficient of variation (CV) >20%. RESULTS: At least one episode of dysglycemia occurred in 103 (51.5%) of the patients, including 6 (3.0%) with hypoglycemia, 83 (41.5%) with hyperglycemia, 18 (9.0%) with severe hyperglycemia, and 40 (20.0%) with increased GV. Unblinded use of the OS with appropriate alarms would likely have averted 97.1% of the episodes of dysglycemia: hypoglycemia (100.0%), hyperglycemia (96.4%), severe hyperglycemia (100.0%), and increased GV (97.5%). Point accuracy of the OS was very similar to that of the point of care BG monitoring devices used in the trial. CONCLUSION: Unblinded use of the OS would have eliminated nearly every episode of dysglycemia in this cohort of critically ill patients, thereby markedly improving the quality and safety of glucose control.

State of Harmonization of 24 Serum Albumin Measurement Procedures and Implications for Medical Decisions.

BACKGROUND: Measurements of serum and plasma albumin are widely used in medicine, including as indicators of quality of patient care in renal dialysis centers. METHODS: Pools were prepared from residual patient serum (n = 50) and heparin plasma (n = 48) from patients without renal disease, and serum from patients with kidney failure before hemodialysis (n = 53). Albumin was measured in all samples and in ERM-DA470k/IFCC reference material (RM) by 3 immunochemical, 9 bromcresol green (BCG), and 12 bromcresol purple (BCP) methods. RESULTS: Two of 3 immunochemical procedures, 5 of 9 BCG, and 10 of 12 BCP methods recovered the RM value within its uncertainty. One immunochemical and 3 BCG methods were biased vs the RM value. Random error components were small for all measurement procedures. The Tina-quant immunochemical method was chosen as the reference measurement procedure based on recovery and results of error analyses. Mean biases for BCG vs Tina-quant were 1.5% to 13.9% and were larger at lower albumin concentrations. BCP methods' mean biases were -5.4% to 1.2% irrespective of albumin concentration. Biases for plasma samples were generally higher than for serum samples for all method types. For most measurement procedures, biases were lower for serum from patients on hemodialysis vs patients without kidney disease. CONCLUSIONS: Significant differences among immunochemical, BCG, and BCP methods compromise interpretation of serum albumin results. Guidelines and calculations for clinical management of kidney and other diseases must consider the method used for albumin measurement until harmonization can be achieved.

The impact of measurement frequency on the domains of glycemic control in the critically ill--a Monte Carlo simulation.

The role of blood glucose (BG) measurement frequency on the domains of glycemic control is not well defined. This Monte Carlo mathematical simulation of glycemic control in a cohort of critically ill patients modeled sets of 100 patients with simulated BG-measuring devices having 5 levels of measurement imprecision, using 2 published insulin infusion protocols, for 200 hours, with 3 different BG-measurement intervals-15 minutes (Q15'), 1 hour (Q1h), and 2 hours (Q2h)-resulting in 1,100,000 BG measurements for 3000 simulated patients. The model varied insulin sensitivity, initial BG value and rate of gluconeogenesis. The primary outcomes included rates of hyperglycemia (BG > 180 mg/dL), hypoglycemia (BG < 70 and 40 mg/dL), proportion of patients with elevated glucose variability (within-patient coefficient of variation [CV] > 20%), and time in range (BG ranges 80-150 mg/dL and 80-180 mg/dL). Percentages of hyperglycemia, hypoglycemia at both thresholds, and patients with elevated glucose variability as well as time outside glycemic targets were substantially higher in simulations with measurement interval Q2h compared to those with measurement interval Q1h and moderately higher in simulations with Q1h than in those with Q15'. Higher measurement frequency mitigated the deleterious effect of high measurement imprecision, defined as CV ≥ 15%. This Monte Carlo simulation suggests that glycemic control in critically ill patients is more optimal with a BG measurement interval no longer than 1h, with further benefit obtained with use of measurement interval of 15'. These findings have important implications for the development of glycemic control standards.

Falsely increased chloride and missed anion gap elevation during treatment with sodium thiosulfate.

BACKGROUND: Sodium thiosulfate (STS) is used to treat calciphylaxis and cyanide poisoning, but can lead to a serious anion-gap acidosis. We suspected that the calculated anion gap in a patient treated with STS for calciphylaxis was decreased to normal by a falsely increased chloride, and we hypothesized that STS directly interfered with chloride measurements. METHODS: Plasma pools were prepared with 12 concentrations of STS from 0 to 20 mmol/l. Chloride was measured in each sample on 9 analyzers: Architect 16200, StatProfile pHOx Plus, RapidLab 1265®, Vitros 350®, Advia 1800, Roche Modular, iSTAT1, RAPIDpoint 500, and Radiometer ABL735. RESULTS: Statistically significant, dose-dependent increases in reported chloride concentrations were seen with all analyzers except the RAPIDpoint 500 and Vitros. The increases ranged from 5 to 75 mmol/l at the peak thiosulfate concentrations (33 mmol/l) expected in treated patients. The CLIA-allowable error of 5% was exceeded by 4 analyzers (Architect 16200, iSTAT1, StatProfile pHOx Plus, and Radiometer ABL735). The RAPIDpoint 500 showed a 3-mmol/l decrease in measured chloride over the tested range. The Vitros analyzer showed no interference. CONCLUSIONS: Interference of STS in chloride measurement in several common analyzers may lead to erroneous anion-gap calculations and confound the diagnosis of STS-induced anion-gap acidosis.

Effects of measurement frequency on analytical quality required for glucose measurements in intensive care units: assessments by simulation models.

BACKGROUND: Total error allowances have been proposed for glucose meters used in tight-glucose-control (TGC) protocols. It is unclear whether these proposed quality specifications are appropriate for continuous glucose monitoring (CGM). METHODS: We performed Monte Carlo simulations of patients on TGC protocols. To simulate use of glucose meters, measurements were made hourly. To simulate CGM, glucose measurements were made every 5 min. Glucose was measured with defined bias (varied from -20% to 20%) and imprecision (0% to 20% CV). The measured glucose concentrations were used to alter insulin infusion rates according to established treatment protocols. Changes in true glucose were calculated hourly on the basis of the insulin infusion rate, the modeled patient's insulin sensitivity, and a model of glucose homeostasis. We modeled 18 000 patients, equally divided between the hourly and every-5-min measurement schemas and distributed among 45 combinations of bias and imprecision and 2 treatment protocols. RESULTS: With both treatment protocols and both measurement frequencies, higher measurement imprecision increased the rates of hypoglycemia and hyperglycemia and increased glycemic variability (SD). These adverse effects of measurement imprecision were lower at the higher measurement frequency. The rate of hypoglycemia at an imprecision (CV) of 5% with hourly measurements was similar to the rate of hypoglycemia at 10% CV when measurements were made every 5 min. With measurements every 5 min, imprecision up to 10% had minimal effects on hyperglycemia or glycemic variability. Effects of simulated analytical bias on glycemia were unaffected by measurement frequency. CONCLUSIONS: Quality specifications for imprecision of glucose meters are not transferable to CGM.

From slide sets to sound bites: teaching and learning pathology in the digital age.

Educational evolution is particularly important in pathology, particularly cytopathology, due to the vast amounts of independent learning required to master this field. In this study, learning challenges faced by pathology residents were addressed through a variety of educational modalities including 24 short (∼10 minute) online tutorials (dubbed "Sound Bites") covering selected topics in cytopathology as well as other areas of anatomic and clinical pathology. Additionally, residents were provided with an annotated glass slide set covering pediatric pathology with an associated multiple choice self-assessment as well as multiheaded microscope slide review sessions. Use of these modalities was tracked and residents surveyed about their experiences using them. All 20 residents (100%) reported using Sound Bites either from work computers, home computers, or mobile devices. Residents reported that easy accessibility, brevity, and opportunities for self-assessment were important variables contributing to this use, and that Sound Bite use would make them more likely to benefit from in-person teaching through lectures and/or slide sessions. Within 12 months of the release of the first Sound Bite, individual Sound Bites were accessed a total of 1169 times (mean: 49 times per Sound Bite). In contrast, slide sets were only accessed about once a month and were only employed by 30% of residents (6 of 20) for independent study; only 20% (4 of 20) completed the accompanying multiple choice self-assessment. All residents attended multiheaded microscope slide review sessions. Whereas traditional educational methods remain valuable tools in pathology education, these data suggest that short, web-based tutorials represent a valuable adjuvant teaching tool.

Empiric validation of simulation models for estimating glucose meter performance criteria for moderate levels of glycemic control.

BACKGROUND: We used simulation modeling to relate glucose meter performance criteria to insulin dosing errors for patients on a moderate glycemic control protocol (glucose target, 110-150 mg/dL) and empirically validated assumptions from simulation models using observed glucose meter and laboratory glucose values obtained nearly simultaneously. SUBJECTS AND METHODS: The 25,948 glucose values from 1,513 patients on a moderate glycemic control protocol were used to represent the expected distribution of glucose values in this patient population. Simulation models were used to relate glucose meter analytical performance to insulin dosing errors assuming 10%, 15%, or 20% total allowable error (TEa). In addition, 4,017 paired glucose meter and serum laboratory glucose measurements drawn within 5 min of each other were used to generate an empiric dataset to validate simulation model assumptions relating glucose meter performance to insulin dosing errors. RESULTS: Large (three or more category) insulin dosing errors are predicted to occur only under the 20% TEa condition. Two category insulin dosing errors were common (6-20% of all insulin dosing decisions) when 20% TEa was assumed, but frequency decreased to only 0.2% of dosing decisions when 10% TEa was modeled. When insulin dosing error rates were measured empirically by comparing paired glucose meter and laboratory glucose values, insulin dosing error rates were very similar to those predicted for the 20% TEa condition. CONCLUSIONS: Both simulation models and empiric data demonstrate that glucose meters that perform at ≥20% TEa allow large insulin dosing errors during a moderate glycemic control protocol.

Failure of current laboratory protocols to detect lot-to-lot reagent differences: findings and possible solutions.

BACKGROUND: Maintaining consistency of results over time is a challenge in laboratory medicine. Lot-to-lot reagent changes are a major threat to consistency of results. METHODS: For the period October 2007 through July 2012, we reviewed lot validation data for each new lot of insulin-like growth factor 1 (IGF-1) reagents (Siemens Healthcare Diagnostics) at Mayo Clinic, Rochester, MN, and the University of Virginia, Charlottesville, VA. Analyses of discarded patient samples were used for comparison of lots. For the same period, we determined the distributions of reported patient results for each lot of reagents at the 2 institutions. RESULTS: Lot-to-lot validation studies identified no reagent lot as significantly different from the preceding lot. By contrast, significant lot-to-lot changes were seen in the means and medians of 105 668 reported patient IGF-I results during the period. The frequency of increased results increased nearly 2-fold to a high of 17%, without detectable changes in the underlying patient demographics. Retrospective statistical analysis indicated that lot-to-lot comparison protocols were underpowered and that validation studies for this assay required testing >100 samples to achieve 90% power to detect reagent lots that would significantly alter the distributions of patient results. CONCLUSIONS: The number of test samples required for adequate lot-to-lot validation protocols is high and may be prohibitively large, especially for low-volume or complex assays. Monitoring of the distributions of patient results has the potential to detect lot-to-lot inconsistencies relatively quickly. We recommend that manufacturers implement remote monitoring of patient results from analyzers in multiple institutions to allow rapid identification of between-lot result inconsistency.

Hepatocyte cytokeratin 7 expression in chronic allograft rejection.

We examined hepatocyte cytokeratin 7 (CK7) expression in chronic allograft rejection (CR), a ductopenic condition in which this has not been systematically evaluated, in 20 patients with the clinicopathologic diagnosis of CR and age-, sex-, and native-disease-matched control subjects. We also studied baseline biopsy specimens from both groups. Three pathologists independently reviewed H&E- and CK7-stained sections, counting interlobular bile ducts (BDs) and portal tracts (PTs), noting the morphologic pattern of injury and scoring hepatocyte CK7 expression (0, none; 1+, rare; 2+, multifocal, predominantly periportal; 3+, extension into the lobule; 4+, diffuse). Mean BD/PT ratios and CK7 scores were calculated. The mean BD/PT ratio (0.58) and CK7 score (1.01) for the "CR, diagnostic" group were significantly different from all other group means (P < .05); no other comparisons were significant (P > .05). A CK7 score of 1 or more was observed in 9 (56%) of 16 CR specimens and in 3 (7%) of 41 remaining specimens. Hepatocyte CK7 expression is frequently noted in CR, and it would appear to reflect ductopenia. CK7 staining may be a useful diagnostic adjunct in evaluation of transplant liver biopsy specimens.

Common reference intervals for aspartate aminotransferase (AST), alanine aminotransferase (ALT) and γ-glutamyl transferase (GGT) in serum: results from an IFCC multicenter study.

BACKGROUND: Aspartate aminotransferase (AST), alanine aminotransferase (ALT) and γ-glutamyl transferase (GGT) measurements are important for the assessment of liver damage. The aim of this study was to define the reference intervals (RIs) for these enzymes in adults, paying attention to standardization of the methods used and careful selection of the reference population. METHODS: AST, ALT and GGT were measured with commercial analytical systems standardized to the IFCC-recommended reference measurement systems. Three centers (two in Italy and one in China) measured their own freshly collected samples; one of these centers also measured frozen samples from the Nordic Countries RI Project and from a Turkish center. RIs were generated using non-parametric techniques from the results of 765 individuals (411 females and 354 males, 18-85 years old) selected on the basis of the results of other laboratory tests and a specific questionnaire. RESULTS: AST results from the four regions (Milan, Beijing, Bursa and Nordic Countries) were statistically different, but these differences were too small to be clinically relevant. Likewise, differences between the upper reference limits for genders was only 1.7 U/L (0.03 µkat/L), allowing a single RI of 11-34 U/L (0.18-0.57 µkat/L) to be defined. Interregional differences were not statistically significant for ALT, but partitioning was required due to significant gender differences. RIs for ALT were 8-41 U/L (0.13-0.68 µkat/L) for females and 9-59 U/L (0.15-0.99 µkat/L) for males, respectively. The upper reference limits for GGT from the Nordic Country population were higher than those from the other three regions and results from this group were excluded from final calculations. The GGT RIs were 6-40 U/L (0.11-0.66 µkat/L) for females and 12-68 U/L (0.20- 1.13 µkat/L) for males, respectively. CONCLUSIONS: For AST and ALT, the implementation of common RIs appears to be possible, because no differences between regions were observed. However, a common RI for GGT that is applicable worldwide appears unlikely due to differences among populations.

An appraisal of statistical procedures used in derivation of reference intervals.

When conducting studies to derive reference intervals (RIs), various statistical procedures are commonly applied at each step, from the planning stages to final computation of RIs. Determination of the necessary sample size is an important consideration, and evaluation of at least 400 individuals in each subgroup has been recommended to establish reliable common RIs in multicenter studies. Multiple regression analysis allows identification of the most important factors contributing to variation in test results, while accounting for possible confounding relationships among these factors. Of the various approaches proposed for judging the necessity of partitioning reference values, nested analysis of variance (ANOVA) is the likely method of choice owing to its ability to handle multiple groups and being able to adjust for multiple factors. Box-Cox power transformation often has been used to transform data to a Gaussian distribution for parametric computation of RIs. However, this transformation occasionally fails. Therefore, the non-parametric method based on determination of the 2.5 and 97.5 percentiles following sorting of the data, has been recommended for general use. The performance of the Box-Cox transformation can be improved by introducing an additional parameter representing the origin of transformation. In simulations, the confidence intervals (CIs) of reference limits (RLs) calculated by the parametric method were narrower than those calculated by the non-parametric approach. However, the margin of difference was rather small owing to additional variability in parametrically-determined RLs introduced by estimation of parameters for the Box-Cox transformation. The parametric calculation method may have an advantage over the non-parametric method in allowing identification and exclusion of extreme values during RI computation.

Assessing the impact of biomarkers on patient outcome: an obligatory step.

Payers for healthcare increasingly require evidence about health outcomes of medical interventions. Outcomes research uses various study designs to provide such evidence, with the highest level of evidence provided by randomized controlled trials (RCTs). Among published studies of biomarkers, however, relatively few determine the relationship of biomarker testing to outcomes, and only a small fraction of those studies are RCTs, and fewer still follow the CONSORT standards for reporting of trials. Outcomes studies of biomarkers are difficult to carry out. During an outcomes study, clinicians may be expected to use the results of the test (e.g., troponin) along with other information (e.g., symptoms of an acute coronary syndrome) to decide about use of another intervention (such as cardiac catheterization) that is hoped to improve an outcome (e.g., mortality rate) at some time in the future. Studies of diagnostic tests frequently lack evidence that test results were acted upon at all, much less according to a defined protocol. The potential for a biomarker to improve outcomes depends upon a wide range of variables. These variables include the diagnostic accuracy of the test and the effectiveness of the therapeutic intervention, both of which will, predictably, vary with the patient population studied. Thus outcomes studies performed in one patient population leave unanswered questions regarding outcomes in other populations. The questions are infinite, but resources are finite. Simulation modelling studies are attractive as an adjunct to patient studies to address multiple patient variables and multiple treatment approaches without the expense of multiple clinical studies.