The diagnostic accuracy of quality control rules.

Internal quality control in clinical chemistry laboratories are based on analyzing samples of stable control materials among the patient samples. The control results are interpreted by using quality control rules that usually are designed to detect systematic errors. The best rules have a high probability of error detection (Ped), i.e. to detect the maximal allowable (critical) systematic error and a low probability of false rejection (Pfr, false alarm). In this work we show that quality control rules can be represented by points on a ROC curve which appears when Ped is plotted against Pfr and only the control limit is varied. Further, we introduce a new method for choosing the optimal control limit, analogous to choosing the optimal operating point on the ROC curve of a diagnostic test. This decision needs knowledge of the pretest probability of a critical systematic error, the benefit of detecting it when it occurs and the cost of false alarm. The ROC curve analysis showed that if rules based on N = 2 are used, mean rules outperform Westgard rules because the ROC curve of the mean rules was lying above the ROC curves of the Westgard rules. A mean rule also had a lower maximum expected increase in the number of unacceptable patient results reported during the presence of an out-of-control error condition (Max E(NUF)) than comparable Westgard rules.

Stability of serum placental growth factor (PlGF) and soluble fms-like tyrosine kinase 1 (sFlt-1).

Placental growth factor (PlGF) and soluble fms-like tyrosine kinase 1 (sFlt-1) are biomarkers used for diagnosis and risk estimation of preeclampsia. Stability in room temperature (RT) may affect the usefulness of these analyses, as shipping at ambient temperature is the most practical and suitable way to ship samples. To date, scientific studies of such stability are lacking. We aimed to assess the stability of PlGF and sFlt-1 at RT in serum from pregnant women. In addition, a smaller study of stability at 4 °C was performed. Serum was collected from 69 pregnant women and stored at RT or at 4 °C for up to 192 h. Analytes were considered stable if the mean percent change ± 90 confidence interval of the mean was within the baseline concentration ± allowable bias. Allowable bias was calculated from data on biological variation. In addition, an instability equation was calculated to assess loss of stability, in line with recent European Federation of Clinical Chemistry and Laboratory Medicine (EFLM) recommendations. The mean percent change was <3.5% for PlGF, <1% for sFlt-1 and <4.5% for sFlt-1/PlGF ratio up to 192 h. PlGF was considered stable for 168 h, and sFlt-1 and sFlt-1/PlGF ratios were considered stable for 192 h at RT. At 4 °C, PlGF was considered stable for 120 h, sFlt-1 for 168 h and sFlt-1/PlGF ratio for 120 h. Both PlGF and sFlt-1 as well as sFlt-1/PlGF ratio show sufficient stability (minimum 168 h) for samples to be shipped at RT.

Interpreting two TSH results from the same patient.

OBJECTIVES: When the patient's mean (setpoint) concentration of an analyte is unknown and the physician tries to judge the clinical condition from the analyte concentration in two separate specimens taken a time apart, we believe that the two values should be judged against a bivariate reference interval derived from clinically healthy and stable individuals, rather than using univariate reference limits and comparing the difference between the values against reference change values (RCVs). In this work we compared the two models, using s-TSH as an example. METHODS: We simulated two s-TSH measurement values for 100,000 euthyreot subjects, and plotted the second value against the first, along with a markup of the central 50, 60, 70, 80, 90, and 95 % of the bivariate distribution, in addition to the 2.5 and 97.5 percentile univariate reference limits and the 2.5 and 97.5 percentile RCVs. We also estimated the diagnostic accuracy of the combination of the 2.5 and 97.5 univariate percentile reference limits and the 2.5 and 97.5 percentile RCVs against the central 95 % of the bivariate distribution. RESULTS: Graphically, the combination of the 2.5 and 97.5 univariate reference limits and the 2.5 and 97.5 percentile RCVs did not accurately delineate the central 95 % of the bivariate distribution. Numerically, the sensitivity and specificity of the combination were 80.2 and 92.2 %, respectively. CONCLUSIONS: The concentrations of s-TSH measured in two samples taken at separate times from a clinically healthy and stable individual cannot be accurately interpreted using the combination of univariate reference limits and RCVs.

Reference change values of FIB-4.

When comparing two analytical results for the same analyte, the clinicians may benefit from knowing the reference change values (RCVs) of the analyte. For Fibrosis-4 Index (FIB-4), a noninvasive test used for assessing the risk of liver fibrosis, no RCVs have been published for non-cirrhotic individuals. Therefore, we estimated RCVs for adults, using retrospectively collected data from outpatients with AST, ALT, and thrombocytes within the respective reference intervals. FIB-4 was calculated as (age × AST)/(thrombocytes × ALT0.5). From two FIB-4 values in each patient we calculated the RCVs parametrically and non-parametrically. For both methods, we estimated the limits of the central 90% of the distribution of the ratio between the second and the first measurement. We obtained data on 599 outpatients with two blood tests taken 3 - 972 (median 258) days apart. The RCVs were 0.72 - 1.40 and 0.72 - 1.43, respectively, using the parametric and non-parametric methods. The 5 and 95 percentiles were not statistically significantly associated with sex, age, level of analyte, or the time between the measurements. The within-subject biological variation of FIB-4 was estimated to be 13.9%. Conclusion: In 90% of the patients the ratio between the second and the first FIB-4 result was approximately 0.7 - 1.4.

Calprotectin and CRP as biomarkers of cardiovascular disease risk in patients with chronic kidney disease: a follow-up study at 5 and 10 years.

Chronic kidney disease (CKD) and low-grade inflammation are associated with increased risk of cardiovascular disease (CVD). Calprotectin, a protein mainly secreted by activated neutrophils during inflammatory conditions, has been linked to CVD risk in general populations. The aim of this study was to evaluate the association of calprotectin with CVD risk in CKD patients, relative to C-reactive protein (CRP). One hundred and fifty-three patients with moderate CKD were prospectively followed up at 5 and 10 years. We used Cox regression modelling with stepwise adjustments for other relevant covariates (age, sex, cystatin C, previous CVD, systolic blood pressure, HDL cholesterol and HbA1c) to assess the association of baseline calprotectin and CRP with the risk of fatal or non-fatal CVD events. Twenty-nine and 44 patients experienced a CVD event during median follow-up of 4.8 and 10.9 years, respectively. Higher calprotectin was associated with increased CVD risk at both time points, which remained statistically significant after multivariable adjustments, including adjustment for CRP. For CRP, the associations did not remain statistically significant after final multivariable adjustments. In conclusion, we have shown that in patients with CKD, calprotectin was independently associated with the risk of future CVD events, suggesting that calprotectin may provide prognostic information of CVD risk.

Testing the limits: the diagnostic accuracy of reference change values.

Reference change values (RCVs) are used by the physician to judge whether a change in analyte concentration from one sample to the next may represent a clinically significant change. Published RCVs are usually given as fixed percentages of the analyte concentration in the first sample. The accuracy of published RCVs is not well known. We obtained public-use data from the US National Health and Nutrition Examination Survey (NHANES) 2001-2002 to study the distribution of changes in the concentration of eight commonly used analytes. Specimens were obtained on two occasions 7-47 days apart from 279 to 411 individuals with an analyte concentration within the reference interval in both samples. The analytes were albumin, calcium, cholesterol, phosphate, potassium, sodium, hemoglobin and thrombocytes. For each analyte, normal within-subject biological coefficient of variation from the EFLM Working Group on Biological Variation and the NHANES analytical coefficient of variation were used to calculate the 5 and 95 percentile RCVs. These RCVs were calculated as fixed percentages of the analyte concentrations in the first sample and compared to the empirical 5 and 95 percentiles. The sensitivity of the RCVs in detecting changes outside the empirical percentiles ranged from 0.35 for sodium to 0.80 for albumin. The specificity of the RCVs in detecting changes inside the empirical percentiles ranged from 0.85 for potassium to 0.97 for thrombocytes. Calculating RCVs as fixed percentages of the analyte concentration in the first sample lessened the diagnostic accuracy. RCVs given as a function of the first result would perform better.

Diagnosing empty iron stores in women: unbound iron binding capacity (UIBC) versus soluble transferrin receptor (sTFR).

Unbound iron binding capacity (UIBC) is more accurate than total iron binding capacity (TIBC) and percent transferrin saturation in diagnosing empty iron stores. It is unknown whether UIBC is more or less accurate than soluble transferrin receptor (sTFR). We obtained public-use data from the U.S. National Health and Nutrition Examination Survey (NHANES) 2005-2006 to compare the accuray of UIBC and sTFR in diagnosing empty iron stores in 2337 women aged 12-49 years. We grouped the women according to CRP less than 5 mg/L and pregnancy (four groups) and used three definitions of empty iron stores: Serum ferritin less than 10, 15, and 20 µg/L. Receiver operating characteristic (ROC) curve analysis was used to estimate the diagnostic accuracy. UIBC showed a better diagnostic accuracy than sTFR in all groups and definitions of empty iron stores, except in nonpregnant women with CRP at least 5 mg/L when empty iron stores were defined as ferritin less than 10 and 15 µg/L. Two differences reached statistical significance: In nonpregnant women without inflammation the area under the ROC curve for UIBC was 0.830 compared to 0.793 for sTFR (p = .007) when empty iron stores were defined as ferritin less than 20 µg/L. The corresponding figures for pregnant women without inflammation were 0.843 for UIBC and 0.739 for sTFR (p = .003). In conclusion, UIBC is a more accurate test than sTFR in diagnosing empty iron stores in women without inflammation.

Using blood calprotectin as a measure of blood neutrophils.

Presently, bed-side or at home quantification of neutrophils in blood (b-neutrophils) is not practical, because cytometric methods are too expensive and technically demanding. We have explored whether calprotectin concentration in whole blood (b-calprotectin) might be a valid measure of b-neutrophils because this principle might be used in a simple and robust immunoassay device. We obtained heparin blood samples from 77 patients with possible neutropenia, most of them cancer patients treated with cytostatic drugs, and compared b-calprotectin with their b-neutrophils in a simultaneously taken EDTA-blood sample. The Spearman rank correlation coefficient between b-calprotectin and b-neutrophils was 0.986 (p < .0001). In a regression model of b-neutrophils as a function of age, gender, type of hematology instrument, total leukocyte count minus neutrophils, b-calprotectin, and plasma calprotectin (p-calprotectin), only b-calprotectin was a statistically significant predictor. B-neutrophils below 1 × 109/L was unlikely if b-calprotectin was above 50 mg/L. In conclusion, b-calprotectin, without adjusting for p-calprotectin, correlates closely with b-neutrophils and could be used to detect b-neutrophils below 1 × 109/L.

Allowable bias derived from the NOBIDA reference values.

Figures of allowable bias are used to make rational choices of quality control rules, to judge the validity of published reference values, and to determine the stability of sample materials. Usually, allowable bias is parametrically defined as 0.25 times the total biological standard deviation, because that is half the width of the 90% confidence interval of parametrically estimated reference limits from 120 reference values. The published figures are mostly derived from very small populations, less than 120. We estimated allowable bias non-parametrically as the least of 4 percentile differences in distributions of reference values from the large Nordic reference interval project biobank and database (NOBIDA). The percentile differences are equivalent to 0.25 times the total biological standard deviation in Gaussian distributions. We also estimated allowable bias from the distributions of non-parametrically estimated reference limits after resampling 120 reference values from the same datasets. Clearly larger allowable bias was derived from the resampling method than from the percentile difference method, showing that non-parametric estimation of reference limits from 120 reference values implies a larger allowable bias than 0.25 times the normal biological standard deviation. With some exceptions, the figures of allowable bias using the percentile difference method were in the same order of magnitude as parametrically derived figures in other studies, and lend some support to the results from those smaller studies. Whether such bias specifications, if met, guarantee measurements of sufficient clinical quality is unknown.

A new index of clinical utility for diagnostic tests.

Clinical utility of a diagnostic test depends on its diagnostic accuracy, the pretest probability of disease and the clinical consequences of the test results. Tools for evaluating clinical utility are scarce. We propose a new clinical utility index (CUI), which is the expected gain in utility (EGU) of the test divided by the EGU of an ideal test, both adjusted for EGU of the optimal clinical action without testing. The index expresses the relative benefit of using the test compared to using an optimal test when making a clinical decision. To illustrate how the index may be used, we estimated CUI for fasting glucose, both as a continuous and as a dichotomous test, at several values of pretest probability of diabetes mellitus and at two levels of cost/benefit-ratio. In the same clinical situations we also estimated CUI for the 2 h glucose tolerance test. Hemoglobin A1c ≥ 48 mmol/mol was used as a reference standard for diabetes mellitus. In this model, fasting glucose was clinically more useful as a continuous test than as a dichotomous one, based on CUIs. At pretest probability above the treatment threshold, fasting glucose as a continuous test was even more useful than the complete glucose tolerance test. These results are not necessarily generalizable; however, they show how the CUI can be used to select the most useful test in certain clinical situations.

Measuring calprotectin in plasma and blood with a fully automated turbidimetric assay.

Calprotectin in plasma and blood might prove to be a useful biomarker of inflammation and infection; however, automated methods for analysing the concentration of calprotectin in those materials are lacking. We have validated a fully automated turbidimetric method and present health-related reference limits. Calprotectin was measured by Siemens Advia XPT with the Bühlmann fCAL® turbo test (Bühlmann Laboratories AG, Schönenbuch, Switzerland), a particle enhanced turbidimetric immunoassay for quantification of calprotectin in fecal extracts. Plasma and serum samples were analysed directly, while whole blood was first extracted with M-PER® Mammalian Protein Extraction Reagent (ThermoFisher) and diluted with B-CAL-EX (Bühlmann). We studied analytical imprecision, estimated health-related reference limits and examined the correlation between neutrophil-calprotectin (blood-calprotectin adjusted for plasma-calprotectin) and the neutrophil count. The intermediate ('day-to-day') coefficient of variation was 3.5 and 1.0% for heparin-plasma-calprotectin at 0.52 mg/L and 3.53 mg/L, respectively, and 4.9% for heparin-blood-calprotectin at 50.2 mg/L. Health-related reference limits were 0.470-3.02 mg/L for calprotectin in heparin-plasma, 50.8-182 mg/L for calprotectin in heparin-blood, 0.534-2.41% for the ratio between them and 24.7-33.3 pg for the mean amount of calprotectin per neutrophil. Compared to heparin-plasma, calprotectin concentrations were significantly lower in EDTA-plasma and higher in serum (p < .05). Correlation between neutrophil-calprotectin and the neutrophil count was excellent. We have shown that the Bühlmann fCAL® turbo test can be used to measure calprotectin in plasma and blood.

The individual systematic difference between CoaguChek and STA-SPA.

Using CoaguChek to measure PT-INR and comparing the results with those from the hospital laboratory, some patients get consistent results while others do not. The extent of this problem is unknown. Our study aimed to quantify the between-subject variation of the systematic PT-INR difference between CoaguChek and a hospital laboratory method. We used register data with PT-INR results from both CoaguChek and a hospital laboratory method (STA-SPA+) in samples taken simultaneously from 108 patients. After excluding five patients with outlying results, we used mixed-effects models to estimate individual slopes and intercepts to describe the systematic relationship between the two methods for each patient, and calculated the fraction of patients having a systematic difference greater than 0.3 INR units. The included 103 patients had from 3 to 16, median seven data pairs measured over a time span from 15 to 2319, median 234 days. The mean of individual slopes was 1.113, with a standard deviation of 0.137. Corresponding values for the intercept were -0.151 and 0.208, respectively. Adjusted for the average systematic difference, the proportion of patients with a systematic difference greater than 0.3 INR units increased from 15% at a PT-INR level of 2.5 to 50% at a PT-INR level of 4. The systematic difference between CoaguChek and STA-SPA + varies considerably between patients. This precludes using a single, common formula to make the CoaguChek results directly comparable to the results from the hospital laboratory.

New reference intervals for cortisol, cortisol binding globulin and free cortisol index in women using ethinyl estradiol.

Healthy women using contraceptives containing a low dose of an estrogen may have a higher serum concentration of cortisol (s-cortisol) and cortisol binding globulin (s-CBG) than the commonly used upper reference limits. There are no published reference intervals for s-cortisol, s-CBG, serum free cortisol index (s-FCI) or cortisol in saliva (sa-cortisol) for these women. The aim was to establish the above-mentioned reference intervals and document the differences in s-cortisol and s-CBG in one group of women using and another group not using ethinyl estradiol (EE). In this cross-sectional study, the reference limits presented were given as the 2.5 and 97.5 percentiles of the distribution of reference values in a population of 277 healthy volunteer women, aged 18-45 years. 157 women were not using any type of estrogen, while 120 women were using contraceptives containing a daily dose of 15-35 µg of EE. Serum and salivary cortisol, and serum CBG were measured using standard laboratory methods. S-FCI was calculated as s-cortisol/s-CBG. The reference intervals for s-cortisol in samples collected at 0800-1030 am in women using and not using EE contraception were: 284-994 nmol/L and 159-569 nmol/L respectively, and for s-CBG: 847-3366 nmol/L and 860-1940 nmol/L, respectively. For s-FCI and sa-cortisol, no clinically significant differences were found. Sa-cortisol may be the preferred measurand for evaluation of possible hypercortisolism in women using estrogens, since cortisol in saliva is not influenced by estrogen. If assessing morning s-cortisol and s-CBG in women using EE, we recommend using separate - and not the commonly used - reference intervals.

Comparison of procalcitonin, C-reactive protein, white blood cell count and clinical status in diagnosing pneumonia in patients hospitalized with acute exacerbations of COPD: A prospective observational study.

Lower respiratory tract infection is the most common cause of acute exacerbations of chronic obstructive pulmonary disease (AECOPD). The aim of the present study was to compare the accuracy of procalcitonin (PCT), C-reactive protein (CRP) and white blood cell count (WBC) as single diagnostic tests and in combination with clinical signs and symptoms to diagnose pneumonia in patients hospitalized with AECOPD. This was a prospective, single centre observational study. Patients with spirometry-confirmed COPD who were hospitalized due to AECOPD were consecutively recruited at the hospital's Emergency Unit. Pneumonia was defined as a new pulmonary infiltrate on chest X-ray. The values of PCT, CRP and WBC were determined at admission. Receiver operating characteristic (ROC) curve analysis was used to study the accuracy of various diagnostic tests. Of the 113 included patients, 35 (31%) had pneumonia at admission. Area under the ROC curve (AUC) for PCT, CRP and WBC as a single test to distinguish between patients with and without pneumonia was 0.67 (95% CI 0.55-0.79), 0.73 (95% CI 0.63-0.84) and 0.67 (95% CI 0.55-0.79), respectively ( p = 0.42 for the test of difference). The AUC for a model of clinical signs and symptoms was 0.84 (95% CI 0.76-0.92). When biomarkers were added to the clinical model, the AUCs of the combined models were not significantly different from that of the clinical model alone ( p = 0.54). PCT had about the same accuracy as CRP and WBC in predicting pneumonia in patients hospitalized with AECOPD both as a single test and in combination with clinical signs and symptoms.

Iron loading in HFE p.C282Y homozygotes found by population screening: relationships to HLA-type and T-lymphocyte subsets.

Iron loading in p.C282Y homozygous HFE hemochromatosis subjects is highly variable, and it is unclear what factors cause this variability. Finding such factors could aid in predicting which patients are at highest risk and require closest follow-up. The degree of iron loading has previously been associated with certain HLA-types and with abnormally low CD8 + cell counts in peripheral blood. In 183 Norwegian, p.C282Y homozygotes (104 men, 79 women) originally found through population screening we determined HLA type and measured total T-lymphocytes, CD4 + and CD8 + cells, and compared this with data on iron loading. In p.C282Y homozygous men, but not in homozygous women, we found that the presence of two HLA-A*03 alleles increased the iron load on average by approximately 2-fold compared to p.C282Y homozygous men carrying zero or one A*03 allele. On the other hand, the presence of two HLA-A*01 alleles, in male subjects, apparently reduced the iron loading. In p.C282Y homozygous individuals, the iron loading was increased if the CD8 + cell number was below the 25 percentile or if the CD4 + cell number was above the 75 percentile. This effect appeared to be additive to the effect of the number of HLA-A*03 alleles. Our data indicate that homozygosity for the HLA-A*03 allele significantly increases the risk of excessive iron loading in Norwegian p.C282Y homozygous male patients. In addition, low CD8 + cell number or high CD4 + cell number further increases the risk of excessive iron loading.