Differences In False-positive Rate Research Articles

Thyroglobulin measurement in the needle washout for diagnosis of lymph node metastases of differentiated thyroid cancer

Introduction. Measurement of thyroglobulin (Tg) in a washout after fine-needle aspiration (FNA) is recommended for the diagnosis of metastases of differentiated thyroid cancer (DTC), but the method is not standardized and there is no recommended threshold value of Tg washout, which makes it difficult to interpret the results.Aim. To analyze Tg in the washout after FNA of lesions of different origin on the neck and to determine its optimal cutoff for the diagnosis of DTC metastases.Materials and methods. Ultrasound-guided FNA was performed in 1258 neck masses from 591 patients, 566 of them with confirmed DTC. In 1023 lesions, FNA was performed after thyroidectomy, 22 – after lobectomy, 213 – with preserved thyroid gland. The needle after FNA was washed in 1 ml of saline, in which the concentration of Tg was then examined. The level of Tg in the washout was compared with the histological (n = 522) or cytological diagnosis if no surgery was performed. The cut-off for Tg washout was determined by ROC analysis.Results. DTC lymph node metastases detected in 577 specimens. Nine specimens were obtained from metastases of thyroid cancer (TC) that does not express Tg (anaplastic TC, poorly differentiated TC, columnar-cell variant DTC), 22 – from neck metastases of other malignancy (lung cancer, mucinous soft tissues tumor, ovarian cancer, esophageal cancer, melanoma, neuroendocrine tumor), 6 – from other tumors of the neck (lymphoma, parathyroid adenoma, neurinoma). In 26 cases, the specimen was regarded as normal thyroid tissue left after thyroidectomy, 37 – postoperative seroma or granuloma, 1 – cyst of the neck, 578 – lymph node hyperplasia, 2 – sarcoidosis. The level of Tg washout from DTC metastasis and thyroid remnant significantly differed from that of non-thyroidal origin (p <0.0001). At the cut-off of 7.8 ng/ml, the sensitivity and specificity of Tg washout in the diagnosis of DTC metastases is 94 and 95 %, and at the cut-off of 20 ng/ml, 90 and 98 %, respectively. False-negative results were obtained from DTC with squamous metaplasia or sparse tumor cells in a specimen. False-positive results were obtained more often from lesions of level VI and IV compared with other localizations (8 % versus 4 %; p = 0.04). There were no differences in false positive rate in patients before and after thyroidectomy (p = 0.17), but in patients after thyroidectomy with a serum Tg >200 ng/ml, the false positive rate of Tg washout was significantly higher than that with a lower level of serum Tg (28 % versus 3 %; p = 0.0004). When comparing diagnostic performance of cytology and Tg washout, the advantage of the latter is in the diagnosis of cystic metastases, and the former is in the diagnosis of micrometastases and tumors that do not express Tg. Thyroglobulin in the washout increased the sensitivity of the cytology by 8 %. The combined use of these methods detected DTC metastases in 100 % of patients.Conclusion. Measurement of Tg in the washout is a useful addition to the cytology, increasing the diagnostic performance of the latter, mainly due to better detection of cystic metastases of DTC. The optimal suggested cut-off for Tg washout is 20 ng/mL, at which there are fewer false positives.

An Open Source Replication of a Winning Recidivism Prediction Model.

We present results of our winning solution to the National Institute of Justice recidivism forecasting challenge. Our team, "MCHawks," placed highly in both terms of accuracy (as measured via the Brier score), as well as the fairness criteria (weighted by differences in false positive rates between White and Black parolees). We used a non-linear machine learning model, XGBoost, although we detail our search of different model specifications, as many different models' predictive performance is very similar. Our solution to balancing false positive rates is trivial; we bias predictions to always be "low risk" so false positive rates for each racial group are zero. We discuss changes to the fairness metric to promote non-trivial solutions. By providing open-source replication materials, it is within the capabilities of others to build just as accurate models without extensive statistical expertise or computational resources.

The risk of racial bias while tracking influenza-related content on social media using machine learning.

Machine learning is used to understand and track influenza-related content on social media. Because these systems are used at scale, they have the potential to adversely impact the people they are built to help. In this study, we explore the biases of different machine learning methods for the specific task of detecting influenza-related content. We compare the performance of each model on tweets written in Standard American English (SAE) vs African American English (AAE). Two influenza-related datasets are used to train 3 text classification models (support vector machine, convolutional neural network, bidirectional long short-term memory) with different feature sets. The datasets match real-world scenarios in which there is a large imbalance between SAE and AAE examples. The number of AAE examples for each class ranges from 2% to 5% in both datasets. We also evaluate each model's performance using a balanced dataset via undersampling. We find that all of the tested machine learning methods are biased on both datasets. The difference in false positive rates between SAE and AAE examples ranges from 0.01 to 0.35. The difference in the false negative rates ranges from 0.01 to 0.23. We also find that the neural network methods generally has more unfair results than the linear support vector machine on the chosen datasets. The models that result in the most unfair predictions may vary from dataset to dataset. Practitioners should be aware of the potential harms related to applying machine learning to health-related social media data. At a minimum, we recommend evaluating fairness along with traditional evaluation metrics.

Evaluation of a Preparticipation Cardiovascular Screening Program Among 1,686 National Collegiate Athletic Association Division I Athletes: Comparison of the Seattle, Refined, and International Electrocardiogram Screening Criteria.

To: (1) analyze the results of 5 years of preparticipation cardiac screening including 12-lead electrocardiogram (ECG) of National Collegiate Athletic Association (NCAA) Division I athletes; and (2) assess the rates of ECG screening abnormalities and false-positive rates among 3 ECG screening criteria. Retrospective chart review. National Collegiate Athletic Association Division I University. One thousand six hundred eighty-six first-year athletes presenting for their preparticipation examination including 12-lead resting ECG. At the completion of the study period, all ECGs were retrospectively reviewed using the Seattle, Refined, and International Criteria. (1) Prevalence of pathologic cardiac conditions identified by screening; and (2) number of ECG screening abnormalities by criteria. Three athletes (0.2%) were found to have conditions that are associated with sudden cardiac death. Retrospective review of ECGs using Seattle, Refined, and International criteria revealed an abnormal ECG rate of 3.0%, 2.1%, and 1.8%, respectively. International criteria [odds ratios (OR), 0.58; P = 0.02] demonstrated a lower false-positive rate compared with the Seattle criteria. There was no significant difference in false-positive rates between the Seattle and Refined (OR, 0.68; P = 0.09) or the International and Refined criteria (OR, 0.85; P = 0.5). There was a low rate of significant cardiac pathology in this population, and no athletes were permanently restricted from play as a result of screening. Our results suggest that the International criteria have the lowest false-positive rate of athlete-specific ECG criteria, and thus, it is the preferred method for preparticipation ECG screening in NCAA athletes.

Ventricular scar channel entrances identified by new wideband cardiac magnetic resonance sequence to guide ventricular tachycardia ablation in patients with cardiac defibrillators.

Ventricular tachycardia (VT) substrate-based ablation has become a standard procedure. Electroanatomical mapping (EAM) detects scar tissue heterogeneity and define conduction channels (CCs) that are the ablation target. Late gadolinium enhancement cardiac magnetic resonance (LGE-CMR) is able to depict CCs and increase ablation success. Most patients undergoing VT ablation have an implantable cardioverter-defibrillator (ICD) that can cause image artefacts in LGE-CMR. Recently wideband (WB) LGE-CMR sequence has demonstrated to decrease these artefacts. The aim of this study is to analyse accuracy ofWB-LGE-CMR in identifying the CC entrances. Thirteen consecutive ICD-patients who underwent VT ablation after WB-LGE-CMR were included. Number and location of CC entrances in three-dimensional EAM and in WB-LGE-CMR reconstruction were compared. Concordance was compared with a historical cohort matched by cardiomyopathy, scar location, and age (26 patients) with LGE-CMR prior to ICD and VT ablation. In WB-CMR group, 101 and 93 CC entrances were identified in EAM and WB-LGE-CMR, respectively. In historical cohort, 179 CC entrances were identified in both EAM and LGE-CMR. The EAM/CMR concordance was 85.1% and 92.2% in the WB and historical group, respectively (P = 0.66). There were no differences in false-positive rate (CC entrances detected in CMR and absent in EAM: 7.5% vs 7.8% in WB vs. conventional CMR, P = 0.92) nor in false-negative rate (CC entrances present in EAM not detected in CMR: 14.9% vs.7.8% in WB vs. conventional CMR, P = 0.23). Epicardial CCs was predictor of poor CMR/EAM concordance (OR 2.15, P = 0.031). Use of WB-LGE-CMR sequence in ICD-patients allows adequate VT substrate characterization to guide VT ablation with similar accuracy than conventional LGE-CMR in patients without an ICD.

Secondary prostate cancer screening outcomes by race in the Prostate, Lung, Colorectal, and Ovarian (PLCO) Screening Trial.

Despite disparities in prostate cancer incidence and mortality rates between black and white men, there is still insufficient data available to assess potential differences in the benefits and harms of prostate cancer screening by race. Although the Prostate, Lung, Colorectal, and Ovarian (PLCO) Screening Trial is underpowered to detect a difference by race in prostate-cancer specific mortality, because of the large study size, there are still sufficient numbers to examine secondary screening outcomes. The objective of this analysis is to examine whether differences exist between black and white participants with respect to screening false-positive rates, biopsy follow-up of men with positive screens, tumor characteristics, and overdiagnosis of prostate cancer. Participants from the PLCO included men aged 55-74 years at baseline. Cancer diagnoses and deaths were identified through study update questionnaires, records of biopsy procedures, and linkage with the National Death Index. Cancer characteristics were obtained by medical abstractors. We used chi-squared tests to assess differences in false-positive rates, biopsy follow-up, and tumor characteristics. We used Cox proportional hazards models to compare incidence and mortality rates adjusting for age and survival rates adjusting for Gleason scores. Black men were slightly more likely (14.5%) to have a false-positive PSA test compared to white men (12.4%; P = 0.02) but less likely to have a false-positive digital rectal exam (DRE) (10.9% vs 14.2%, respectively; P < 0.001). Among all men who were screened, black men were significantly more likely to undergo a biopsy than white men (16.5% vs 13.8%, respectively [P = 0.003]) but there was no difference when limited to those with a positive PSA test. Prostate cancer tumors were more likely to be aggressive and to have metastasized in black men compared to white men. Disparities in incidence, mortality, and survival rates were comparable to those seen in population-based data. There was evidence that false-positive test results differed by race and screening test. Consistent with previous studies, cancer outcomes, and tumor characteristics were all more unfavorable in black men.

Comparison of Prostate Imaging Reporting and Data System versions 1 and 2 for the Detection of Peripheral Zone Gleason Score 3 + 4 = 7 Cancers

The objective of our study was to compare Prostate Imaging Reporting and Data System version 1 (PI-RADSv1) and Prostate Imaging Reporting and Data System version 2 (PI-RADSv2) for the detection of peripheral zone (PZ) Gleason score 3 + 4 = 7 cancers. Forty-seven consecutive patients with 52 PZ Gleason score 3 + 4 = 7 cancers that were 0.5 cm3 or larger underwent radical prostatectomy (RP) and 3-T MRI between 2012 and 2015. Two blinded radiologists (readers 1 and 2) retrospectively assigned PI-RADSv1 sequence (T2-weighted imaging, DWI, dynamic contrast-enhanced MRI [DCE-MRI]) and sum scores and PI-RADSv2 assessment categories. A third blinded radiologist (reader 3) measured apparent diffusion coefficient (ADC) ratio (ADC of tumor / ADC of normal PZ) using RP-MRI maps. Sensitivity, false-positive rate, and overall accuracy were compared using McNemar test. Pearson correlation was performed. Using PI-RADSv1, reader 1 detected 86.5% (45/52) of the cancers and reader 2, 76.9% (40/52) of the cancers. Using PI-RADSv2, reader 1 detected 78.9% (41/52) and reader 2, 67.3% (35/52). Reader 1 detected 7.7% (4/52) and reader 2 detected 9.6% (5/52) more tumors using PI-RADSv1 due to T2-weighted imaging score ≥ 4 or DCE-MRI score ≥ 3. Sensitivity was higher for PI-RADSv1 (p = 0.01 and 0.03, readers 1 and 2). False-positive rates were higher with PI-RADSv1 than with PI-RADSv2 (1.8% vs 0.9% for reader 1; 3.6% vs 1.8% for reader 2) without significant differences in false-positive rate (p = 0.41 and 0.25) or overall accuracy (p = 0.06 and 0.23). PI-RADSv1 sum scores correlated strongly with PI-RADSv2 categories (B = 0.78-0.93, p < 0.0001). The mean ADC ratio was 0.61 ± 0.14 mm2/s with no difference between visible and nonvisible tumors (p = 0.06-0.5). Interobserver agreement was moderate for PI-RADSv2 (κ = 0.41) and ranged from slight to substantial for PI-RADSv1 (T2-weighted imaging, κ = 0.32; DWI, κ = 0.52; DCE-MRI, κ = 0.13). There was no difference in overall detection of cancers comparing PI-RADSv1 and PI-RADSv2; however, PI-RADSv1 sequence scores on T2-weighted imaging and DCE-MRI detected approximately 10% more tumors that were otherwise underestimated on DWI and using PI-RADSv2 decision-tree rules.

769: Clinical accuracy of abnormal cell-free fetal DNA (cfDNA) results for the sex chromosomes

769: Clinical accuracy of abnormal cell-free fetal DNA (cfDNA) results for the sex chromosomes

Cell‐free DNA testing: how to choose which laboratory to use?

Cell‐free <scp>DNA</scp> testing: how to choose which laboratory to use?

Taking blood cultures from a newly established intravenous catheter in the emergency department does not increase the rate of contaminated blood cultures

It has been suggested that blood cultures drawn from vascular catheters have a higher false positive rate than those drawn by venepuncture. In the face of institutionally imposed practice change prohibiting obtaining blood cultures from intravenous (i.v.) catheters in the ED, our aim of was to compare the rate of contaminated blood cultures between those taken from recently placed i.v. catheters and those taken by direct venepuncture. Prospective, non-randomised, observational study comparing the rate of contaminated blood cultures for specimens taken from recently placed (<1 h) i.v. catheters and direct venepuncture in adult ED patients. Outcome of interest was the rate of false positive cultures. Analysis was by comparison of proportions (χ(2) -test). Four hundred seventy-two blood culture sets were studied. There were 65 positive cultures, of which 49 (75%; 95% confidence interval [CI], 63-85%) were classified as true positive. The overall rate of contaminated blood cultures was 3.4% (95% CI, 2.0-5.6%). There was no difference in false positive rate between blood cultures taken via venepuncture and those taken from a recently placed i.v. cannula (P = 0.52; odds ratio, 0.9; 95% CI, 0.33-2.44). We found no difference in contaminated blood culture rate between recently placed i.v. catheters and direct venepuncture when infection control procedures were followed.

Pseudo-Outbreak of Vancomycin-Resistant-Enterococcus (VRE) Colonization in a Neonatal Intensive Care Unit Using Spectra VRE Surveillance Medium

From November 2011 through March 2012, we surveyed 272 babies in our neonatal intensive care unit for rectal colonization with vancomycin-resistant enterococci (VRE). Using Spectra VRE medium (Remel Diagnostics, Lenexa, KS), we identified one neonate colonized with vancomycin-resistant Enterococcus faecium. In addition, 55 (13%) of the surveillance cultures yielded false-positive results with vancomycin-susceptible Enterococcus faecalis. During the same time period, 580 rectal swabs were collected from adult patients resulting in 20 (3%) false-positive cultures. The difference in false-positive rates between cultures from babies and adults was statistically significant (P < 0.001), prompting an investigation of factors that might influence the elevated false-positive rate in the neonates including patient demographics, nutrition, and topical ointments applied at the time of testing. Older neonates, with a median age of 6 weeks, were more likely to have false-positive cultures than younger neonates with a median age of 3 weeks (P < 0.001). The younger neonates receiving Similac Expert Care products were less likely to have false-positive surveillance cultures than those receiving other formulas (P < 0.001). Application of topical products was not associated with false-positive cultures. The false-positive E. faecalis strains were typed by Diversilab Rep-PCR (bioMérieux, Marcy l'Etoile, France) and found to represent eight different groups of isolates. The utility of the Spectra VRE media appeared to be significantly impacted by the age of the patients screened.

Rates of downstream invasive coronary angiography and revascularization: computed tomographic coronary angiography vs. Tc-99m single photon emission computed tomography

Computed tomographic coronary angiography (CTA) appears to be a useful modality for the detection of obstructive coronary artery disease (CAD). Recent data suggest that CTA may reduce the frequency of normal invasive coronary angiograms. However, there remains concern that the implementation of CTA could increase referrals to invasive coronary angiography (ICA). To further support the clinical acceptance of CTA, it is important to compare CTA to another accepted modality such as single photon emission computed tomography (SPECT). We followed a cohort of 64-slice CTA patients and a matched cohort of Tc-99m SPECT patients to determine downstream referrals for ICA and revascularization. Consecutive CTA patients (without history of revascularization or cardiac transplantation) were prospectively enrolled and compared with a Tc-99m SPECT cohort (matched for age, gender, and Morise score). Each CTA and SPECT was evaluated for obstructive CAD and patients were followed for downstream ICA and revascularization. Of the 1221 patients in each cohort, 129 (10.6%) CTA patients and 125 (10.2%) SPECT patients were referred to ICA. Of those referred to ICA, obstructive CAD was confirmed in 105 (81.4%) CTA patients and in 88 (70.4%) SPECT patients. Differences in false positive rates were significantly lower in the CTA than the SPECT cohort (9.7 and 25.8%, respectively, P = 0.009). Rates of revascularization were similar in the CTA and SPECT cohorts (6.2 vs. 5.9%, respectively). Compared with SPECT, CTA had similar referrals for ICA and revascularization rates but lower false positive rates. Computed tomographic coronary angiography appears to be a viable non-invasive diagnostic modality and does not appear to negatively impact upon ICA resources.

Differences between first and subsequent rounds of the MRISC breast cancer screening program for women with a familial or genetic predisposition

Within the Dutch MRI Screening (MRISC) study, a Dutch multicenter screening study for hereditary breast cancer, the authors investigated whether previously reported increased diagnostic accuracy of magnetic resonance imaging (MRI) compared with mammography would be maintained during subsequent screening rounds. From November 1999 to October 2003, 1909 eligible women were included in the study. Screening parameters and tumor characteristics of different rounds were calculated and compared. The authors defined 3 different types of imaging screening rounds: first round in women never screened by imaging before, first round in women screened by imaging (mainly mammography) before, and subsequent rounds. The difference in sensitivity for invasive cancers between mammography and MRI was largest in the first round of women previously screened with mammography (20.0 vs. 93.3%; P=.003), but also in subsequent rounds, there was a significant difference in favor of MRI (29.4 vs. 76.5%; P=.02). The difference in false-positive rate between mammography and MRI was also largest in the first round of women previously screened with mammography (5.5 vs. 14.0%; P<.001), and it remained significant in subsequent rounds (4.6 vs. 8.2%; P<.001). Screen-detected tumors were smaller and more often lymph node negative than symptomatic tumors in age-matched control patients, but no major differences in tumor stage were found between tumors detected at subsequent rounds compared with those in the first round. In subsequent rounds, a significantly higher sensitivity and better discriminating capacity of MRI compared with mammography was maintained, and a favorable tumor stage compared with age-matched symptomatic controls. As results of these subsequent screening rounds were most predictive for long-term effects, the authors expect that this screening program will contribute to a decrease of breast cancer mortality in these high-risk women.

The effect of fetal gender on the false-positive rate of Down syndrome by maternal serum screening

(1) To further explore if there is a difference in maternal serum levels of alpha-fetoprotein (AFP), human chorionic gonadotrophin (hCG) and estriol (uE3) between fetal genders. (2) To determine if these differences influence false-positive rates of Down syndrome screening in pregnancies with male or female fetuses. This is a descriptive study of women screened at the Ontario Maternal Serum Screening program between 1993 and 1995. The women were grouped by fetal gender and ethnicity. Serum levels of the three markers and screening false-positive rates for Down syndrome were compared between fetal genders in women of different ethnicity respectively. Complete data were available for 110 306 pregnancies. In all three ethnic groups, MSAFP levels were significantly decreased and MShCG levels were significantly increased in women with female fetuses. The level of MSuE3 was similar between genders. The difference in false-positive rates of Down syndrome between genders was not statistically significant. This is the largest study comparing false-positive rates between fetal genders. In contrast to previous studies, the differences in the serum marker levels between fetal genders do not influence the false-positive rates for Down syndrome.

Variability of polymerase chain reaction detection of the bcl-2-IgH translocation in an international multicentre study

The capacity of the polymerase chain reaction (PCR) to detect very low numbers of cells bearing a t(14;18) translocation has led to its application in assessment of the results of treatment for follicular lymphoma, and suggestions that therapy might be guided by molecular studies. To test the reliability of PCR a collaborative study was undertaken to compare results from different laboratories in Europe and North America. Twenty laboratories with records of publication in molecular diagnostics were sent blood from normal donors with varying numbers of t(14;18)-bearing cells added from a cell line with a translocation in the major breakpoint region (MBR) of the bcl-2 gene. Samples contained 1000, 100, 10, 1 or 0 cells per ml of whole blood and were sent blinded in duplicate. PCR methodology varied widely, with the total number of amplification cycles between 30 and 70, and 13 different primers used for the MBR region. Twelve laboratories used nested PCR and eight single round amplification. The sensitivity of nested and single round PCR was similar at 100 cells/ml but below this the nested method proved significantly more sensitive. The false positive rate was 28%, with 11 samples from 9 laboratories reported as positive when no t(14;18) cells were added. PCR product size and sequence analysis showed that false positives were due to contamination from cell-line DNA rather than background translocations in the donors. There was no significant difference in false positive rates between nested and single round techniques. The polymerase chain reaction to detect bcl-2-IgH rearrangements is presently carried out with widely disparate results. Further effort is required to bring forward a standard PCR protocol which can be re-tested in different laboratories to improve accuracy and reproducibility. The application of quantitative techniques such as real-time PCR may resolve many of the problems presently encountered.

For appreciable ethnic differences in false positive rates, use different cut off points

EDITOR,—Lucy Gilbert and colleagues report ethnic differences in biochemical screening for Down's syndrome and the relative ineffectiveness of the programme in Indian Asians.1 At national meetings of the Association of Clinical Biochemists we presented our findings on the ethnic differences in maternal serum (alpha) fetoprotein and free β human chorionic gonadotrophin concentrations2 and the differences in false positive rates in the various groups, which remain unchanged even after the use of “ethnospecific” medians.3 The false positive rate …

Comparison of the effects of season and prematurity on the enzymatic newborn screening tests for galactosemia and biotinidase deficiency

In the Commonwealth of Virginia, 299,420 newborns were screened for biotinidase deficiency and galactosemia during the years 1989–1991 using enzymatic assays for galactose-l-phosphate uridyl transferase activity and biotinidase activity on dried blood filter paper cards. Retrospectively, false-positive initial samples for galactosemia or biotinidase deficiency were identified, and monthly frequencies of false-positive results were determined. The false-positive rates for galactosemia and biotinidase deficiency screening decreased over the 3-year period and were significantly higher in galactosemia screening for each year examined P < 0.001). For galactosemia screening, monthly false-positive rates were highest during May through September, seasonal differences in false-positive rates were significant for each year evaluated ( P < 0.001). and seasonal false-positive rates for the 3 years combined were highest during the summer and lowest during the winter. For biotinidase deficiency, differences in seasonal false-positive rates were significant in 1989 ( P<0.001) and 1991 ( P<0.01), and the highest seasonal false-positive rate for the 3 years combined occurred in the spring. Average monthly temperature and year of measurement were significant factors in galactosemia screening only ( P< 0.0001) and accounted for much of the variance in false-positive rates ( r 2 = 0.43). Approximately half of the infants with false-positive results for biotinidase deficiency were premature, significantly more than the incidence of prematurity in Virginia (7–8%) during the 3 years ( P<0.001). Prematurity in false-positive galactosemia screens was not significantly different from expected rates.