F-test P-value Research Articles

Reproducibility and usability assessment of the novel Fine Birth device for threatened preterm labor diagnosis

Preterm delivery is considered the leading cause of mortality worldwide in children under five. Forty-five million pregnant women are hospitalized yearly for Threatened Preterm Labor (TPTL). However, only 50% of TPTLs end in delivery before the estimated date, classifying the rest as false TPTL. The ability of current diagnostic methods to predict TPTL is low (low positive predictive value), ranging between 8-30%. This highlights the need for a solution that accurately detects and differentiates between false and real TPTLs in women who attend obstetric clinics and hospital emergency departments with delivery symptoms. This is a pilot study aiming at first, to assess the reproducibility and usability of a novel medical device, the Fine Birth, aimed at accurately diagnosing TPTL through the objective quantification of pregnant women's cervical consistency. Secondarily, we aimed to evaluate the impact of training and the incorporation of a lateral micro camera on the device's reliability and usability outcomes. 77 singleton pregnant women were recruited during their follow-up visit to the obstetric and gynecological departments at five Spanish hospitals. Eligibility criteria were: pregnant women ≥ 18 years old, with a normal fetus and uncomplicated pregnancy, without prolapse of membranes, uterine anomalies, previous cervical surgery or latex allergy, and signing the informed written consent. Cervical tissue stiffness was assessed using the Fine Birth device, whose technology is based on the propagation of torsional waves through the studied tissue. Cervical consistency measurements were taken for each woman until obtaining two valid measurements by two different operators. The intra-observer and inter-observer reproducibility of the Fine Birth measurements was assessed using the Intraclass Correlation Coefficients (ICC) with a 95% confidence interval (CI) and the F-test p-value. The usability was evaluated based on clinicians' and participants' feedback. There was a good intra-observer reproducibility (ICC=0.88, 95% CI 0.84 to 0.95; F-test p-value < 0.05). Since the results obtained for the inter-observer reproducibility did not reach the desired acceptable values (ICC<0.75), a lateral micro camera was added to the Fine Birth intravaginal probe, and the operators involved in the clinical investigation received the corresponding training with the modified device. Analysis of 16 additional subjects demonstrated an excellent inter-observer reproducibility (ICC=0.93, 95% CI 0.78 to 0.97) and a clear improvement after the intervention (p<0.0001). The robust reproducibility and usability results obtained after the insertion of a lateral micro camera and the corresponding training make the Fine Birth a promising novel device to objectively quantify the patient's cervical consistency, diagnose TPTL and thus, predict the risk of spontaneous preterm birth. Further research is needed to demonstrate the clinical utility of the device.

Covid-19: Regional policies and local infection risk: Evidence from Italy with a modelling study.

Background: Policy-makers have attempted to mitigate the spread of covid-19 with national and local non-pharmaceutical interventions. Moreover, evidence suggests that some areas are more exposed than others to contagion risk due to heterogeneous local characteristics. We study whether Italy’s regional policies, introduced on 4th November 2020, have effectively tackled the local infection risk arising from such heterogeneity.Methods: Italy consists of 19 regions (and 2 autonomous provinces), further divided into 107 provinces. We collect 35 province-specific pre-covid variables related to demographics, geography, economic activity, and mobility. First, we test whether their within-region variation explains the covid-19 incidence during the Italian second wave. Using a LASSO algorithm, we isolate variables with high explanatory power. Then, we test if their explanatory power disappears after the introduction of the regional-level policies.Findings: The within-region variation of seven pre-covid characteristics is statistically significant (F-test p-value ) and explains 19% of the province-level variation of covid-19 incidence, on top of region-specific factors, before regional policies were introduced. Its explanatory power declines to 7% after the introduction of regional policies, but is still significant (p-value ), even in regions placed under stricter policies (p-value ).Interpretation: Even within the same region, Italy’s provinces differ in exposure to covid-19 infection risk due to local characteristics. Regional policies did not eliminate these differences, but may have dampened them. Our evidence can be relevant for policy-makers who need to design non-pharmaceutical interventions. It also provides a methodological suggestion for researchers who attempt to estimate their causal effects.Funding: None.

COVID-19: Are Regional Policies Local Enough? Evidence from Italy

Background: Policy-makers have attempted to mitigate the spread of covid- 19 with national and local non-pharmaceutical interventions. Moreover, evidence suggests that some areas are more exposed than others to contagion risk due to heterogeneous local characteristics. We study whether Italy’s regional policies, introduced on 4th November 2020, have effectively tackled the local infection risk arising from such heterogeneity.Methods: Italy consists of 20 regions, further divided into 107 provinces. We collect 35 province-specific pre-covid variables related to demographics, geography, economic activity, and work mobility. First, we test whether their within-region variation explains the covid-19 incidence during the Italian second wave. We use the LASSO algorithm to isolate variables with high explanatory power. Then, we test if their explanatory power disappears after the introduction of the regional-level policies.Findings: The within-region variation of seven pre-covid characteristics is statistically significant (F-test p-value Interpretation: Even within the same region, Italy's provinces differ in exposure to covid-19 infection risk due to local characteristics. Regional policies did not eliminate these differences, but may have dampened them. Our evidence can be relevant for policymakers who need to design non-pharmaceutical interventions.It also provides a methodological suggestion for researchers who attempt to estimate their causal effects.Funding: None.Conflict of Interest: The authors have no conflict of interest of any sort.Ethical Approval: No ethical issue requiring the approval of the ethical committee of the authors' institutions has arisen during the data collection or research phase.

Abstract TP346: Quantitative Analysis of Intracranial Blood Volumes and Relationship With Stress Induced Hyperglycemia and Outcomes in Intracerebral Hemorrhage

Introduction: Stress-induced hyperglycemia (SIH) is independently associated with poor clinical outcome in intracerebral hemorrhage (ICH). It is unknown if quantitative radiological parameters determine SIH in ICH population. Hypothesis: Investigate whether SIH is associated with change in intracranial volume due to ICH using unbiased quantitative estimation, intraventricular hemorrhage (IVH) and ICH location. Methods: Retrospective cohort of 328 patients with spontaneous ICH over 5-year. Glycemic Gap (GG) is used to define SIH. Receiver Operator Curve (ROC) determined threshold for adverse outcomes (hospice or death). Blood volume was estimated in each intracranial compartment on admission head Computed Tomography (CT) using image segmentation software. Linear regression models and Pearson’s correlation analysis were used to study inter-relationship between SIH, ICH and IVH volume. Results: A GG of 29.5 mg/dL and higher defined SIH in this study cohort. Patients with SIH had significantly larger hemorrhage (30.5 mL vs 17.6 mL, p&lt;0.005). ICH, IVH volumes and SIH showed positive correlation (r=0.185, p&lt;0.005; r=0.233, p&lt;0.005 respectively). Subcortically located ICH patients were more likely to have SIH as compared to other locations (ANOVA F-test p-value 0.011). ICH score also showed positive correlation with SIH (r =0.333, p&lt;0.005). For each point increase in ICH score, GG increased by 10.6 mg/dL. Conclusion: Our single center retrospective study reports dose dependent SIH with ICH/IVH volumes and subcortical hemorrhage location. Such observation likely alludes to involvement of hypothalamo-pituitary endocrine axis in pathogenesis of SIH after ICH.

P13.13 Clinical factors involved in brain metastasis edema: results of a retrospective cohort and evaluation of the best edema descriptor

Abstract BACKGROUND Brain metastasis (BM) are frequently surrounded by brain edema with a variable extension. This pathological increase in the water mass from the interstitial or intracellular space is the source of neurological symptoms but some studies also showed a prognosis impact of this edema. Nevertheless, the current edema evaluation methods are poorly reproducible and were not extensively evaluated. From a series of patients with brain metastasis, we analyzed factors influencing perilesional edema extension in function of three edema descriptors which were qualitatively evaluated: the absolute cumulative intracranial edema volume (CIEV, in mL), the edema on tumor volume ratio (ER), and the Edema Theoretical Thickness (ETT, in mm), new factor designed to be more independent from geometrical considerations. MATERIAL AND METHODS Clinical, biological and imaging factors were retrospectively recorded from patients referred for upfront stereotactic radiotherapy of one to three BMs. CIEV, ER and ETT were calculated for each patient. Non-colinear factors were selected using the Farrar-Glauber test. Impact of these factors on edema was tested with an univariate then a multivariate linear regression for each edema descriptor. Each resulting regression model was qualitatively evaluated using the F-test, R square value and residuals calculation. RESULTS Between January 2012 to December 2017, 182 patients were included. Upon 20 potentially prognostic factors recorded, 10 were kept, including clinical, biological and imaging factors. In the multivariate analysis, the CIEV was influenced by the cumulative intracranial tumor volume (CITV) (coefficient of 1.95, p&lt;.001). The CIEV F-statistic p-value and R square were &lt;10–15 and 0.30, respectively. The ETT was significantly influenced by the CITV (0.31, p&lt;.001) and the extracerebral disease control (-1.58, p=.049). The F-test p-value and R square were &lt;10-10 and 0.25, respectively. No significant linear regression was found for the ER. Residuals were more clinically relevant for the ETT compared to the CIEV. CONCLUSION Being significantly influenced by the cumulative brain tumor volume and the extracerebral disease control, the ETT might be a factor to include in future BM works because its strong relationships with confounding factors and its ability to provide robust intra and inter subjects comparisons.

Association between availability of children's book and the literacy-numeracy skills of children aged 36 to 59 months: secondary analysis of the UNICEF Multiple-Indicator Cluster Surveys covering 35 countries.

BackgroundStimulating home environments that have children’s books, pictures and play toys facilitate caregiver-child interactions and enhance children’s development. Although this has been demonstrated in small-scale intervention studies, it is important to document whether book ownership is beneficial at large scale in low and middle-income settings.MethodsWe conducted a secondary analysis using data from the multiple-indicator cluster survey, covering 100 012 children aged 36-59 months, from 35 countries. The outcome was children being on-track for a literacy-numeracy index (LNI) constructed from three questions assessing children’s ability to identify/name at least 10 letters of the alphabet, read at least four simple popular words and know the names and symbols of all numbers from 1-10. The main exposure was availability of children’s book to the child within household. Analysis considered the survey design, assessed and ranked risk ratios of being on track, adjusting for potential confounders such as child’s age (in months), maternal education, household wealth index quintile and area of residence (rural/urban). Ecological analysis was performed using meta-regression after grouping countries by World Bank income groups (low- to high-income).ResultsOnly half (51.8%) of children from all the countries analysed have at least one children’s book at home and less than one-third (29.9%; 95% confidence interval (CI) = 23.5%, 36.3%) are on track for literacy-numeracy. After adjusting for confounders, the likelihood of being on track in literacy-numeracy almost doubled if at least one book was available at home compared to when there was none: RR = 1.89 (95% CI = 1.75, 2.03). There was an economic gradient showing that the likelihood of children being on track for LNI decreased with the country’s income group: adjusted-RR ranged from 1.65 in upper middle income to 2.23 in LIC (F-test P-value <0.0001). Only three high-income countries were included, and children’s books were universally available resulting in wide confidence intervals for the effect.ConclusionsThese findings are policy-relevant, as they corroborate the results from small scale experiments. Making children’s book available to children is a cheap and feasible intervention that could change home dynamics to improve the future economic fortunes of children especially in the poorest countries.

Modeling accrual of older adults to cancer clinical trials: (Alliance A151736).

e18132 Background: Older adults (65+ years) make up &gt; 50% of cancer patients in the US,1 yet, older patient enrollment onto trials remains &lt; 30%.2 This study’s aim was to determine the association between trial level variables and accrual rates of adults 65+ to better design trials that can more improve enrollment of these patients. Methods: We analyzed enrollment percentages of patients age 65+ in non-age specific Alliance trials with &gt; 50 patients, that completed accrual between 1995- 2015. Enrollment percentage by 65+ for each study was compared to cancer incidence percentage by 65+ according to the Surveillance, Epidemiology, and End Results (SEER) Program for the same time periods. Enrollment Disparity Difference (EDD)3, defined as the difference between SEER incidence percentage of patients 65+ and the trial percentage enrollment of 65+ was calculated for each trial ( SEER % – trial %). Univariate analyses (UVA) were performed using simple linear regression to determine trial variables associated with larger EDDs (higher disparity). Variables that had an F-test p-value &lt; 0.20 were included in a multivariable fixed-effects linear model for multivariate analysis (MVA). Results: Median age of 66,708 patients across 237 trials with 10 cancer types was 60 years. Average % of 65+ in trials was lower than average SEER comparison % 65+ per trial (37.6% vs. 57.8%). The median trial EDD, was 16.7%. On UVA, cancer type was significantly associated with higher EDD (see Table). Use of fewer modalities (1, 2, vs 3+) was associated with lower estimated EDDs (1 = -11 [+/-3.6], 2 = -8.9 [+/-3.9], F-test p = .012). On MVA, non-genitourinary (GU) cancer type, number of treatment modalities, and the phase (II or III compared to observational/other) were all significantly associated with higher EDD. Conclusions: Older adults were under-represented in this large sample of trials. This information can be used as a platform to better design disease-specific trials to improve older patient accrual, particularly in disease areas which are under-performing. Support: U10CA180821, U10CA180882, UG1CA189823: https://acknowledgments.alliancefound.org .

Accuracy of Gene Expression Prediction From Genotype Data With PrediXcan Varies Across and Within Continental Populations.

Using genetic data to predict gene expression has garnered significant attention in recent years. PrediXcan has become one of the most widely used gene-based methods for testing associations between predicted gene expression values and a phenotype, which has facilitated novel insights into the relationship between complex traits and the component of gene expression that can be attributed to genetic variation. The gene expression prediction models for PrediXcan were developed using supervised machine learning methods and training data from the Depression Genes and Networks (DGN) study and the Genotype-Tissue Expression (GTEx) project, where the majority of subjects are of European descent. Many genetic studies, however, include samples from multi-ethnic populations, and in this paper we evaluate the accuracy of PrediXcan for predicting gene expression in diverse populations. Using transcriptomic data from the GEUVADIS (Genetic European Variation in Disease) RNA sequencing project and whole genome sequencing data from the 1000 Genomes project, we evaluate and compare the predictive performance of PrediXcan in an African population (Yoruban) and four European ancestry populations for thousands of genes. We evaluate a range of models from the PrediXcan weight databases and use Pearson's correlation coefficient to assess gene expression prediction accuracy with PrediXcan. From our evaluation, we find that the predictive performance of PrediXcan varies substantially among populations from different continents (F-test p-value < 2.2 × 10−16), where prediction accuracy is lower in the Yoruban population from West Africa compared to the European-ancestry populations. Moreover, not only do we find differences in predictive performance between populations from different continents, we also find highly significant differences in prediction accuracy among the four European ancestry populations considered (F-test p-value < 2.2 × 10−16). Finally, while there is variability in prediction accuracy across different PrediXcan weight databases, we also find consistency in the qualitative performance of PrediXcan for the five populations considered, with the African ancestry population having the lowest accuracy across databases.

HNSCC Gene Expression Subtypes, Including a HPV Subtype, Demonstrate Differential Immune Cell and Biomarker Associations

Human papilloma virus (HPV) RNA expression in combination with a gene expression signature of 144 genes was used to classify head and neck squamous cell carcinoma (HNSCC) into 5 distinct subtypes. The subtypes are investigated for differential associations with clinical features, immune infiltration, and immune marker expression. Using previously published Bindea et al. immune cell gene signatures (24 in total) and the TCGA Head & Neck cancer gene expression dataset (HNSCC n=520), we examined immune cell expression in relation to 5 HNSCC gene expression subtypes (atypical, classical, basal, mesenchymal, and HPV subtype). The HPV tumor subtype was determined by alignment of RNAseq with HPV sequences and by evaluation of its gene expression profile. Signatures of multiple immune cells as well as single immune-biomarkers, (CTLA4, PDCD1(PD-1), and CD274(PD-L1) were examined across the 5 subtypes. Differential gene expression was examined using the Kruskal-Wallis test. Immune cell signature associations with tumor subtype and with CD274 expression were evaluated using linear regression. Survival-based associations were evaluated with cox proportional hazard models. Immune cell expression was significantly different across the subtypes (Tcells KW P=1.7e-17) and in general was highest in HPV positive tumors. The classical expression subtype demonstrated lower Tcell immune expression. CD274 (PD-L1) expression was less variable across the subtypes but lowest expression was again seen in the classical subtype (KW P=3.7e-06). Mutation burden varied across the subtypes with the highest mutation burden observed in the atypical subtype and the lowest mutation burden in the HPV tumors (KW P=1.7e-10). Subtype and HPV status were more strongly associated with immune expression than CD274 (PD-L1) expression (median F-test P-value and adjusted R-squared were 2.0e-16 and 0.14 for subtype versus 6.7e-07 and 0.04 for CD274). Cox models with adjustment for stage suggested higher immune cell expression is associated with better survival, and most strongly in the HPV tumor subtype including T gamma delta cells HR 0.2 (95% CI 0.06-0.68) and T cells HR 0.33 (95% CI 0.14-0.79). Biologic subtypes of HNSCC reveal key differences in immune cell expression, which were not always correlated with CD274 (PD-L1) expression. Evaluation of subtypes as potential biomarkers for immunotherapy response in HNSCC should be investigated.

Combined FDG and CT Radiomics Features Predict FMISO Uptake in Head and Neck Cancer

18F-fluoromisonidazole (FMISO) is a hypoxia PET imaging agent that has been used in head and neck to alter management. However, FMISO is still an investigational agent that requires an institutional review board with an IND and therefore is not widely available. Having a surrogate biomarker based on an approved imaging technique that could indirectly identify hypoxia would enable hypoxia-based treatment modulation in a routine clinical setting. In this work we assessed whether a radiomics biomarker built from FDG PET and CT features could predict FMISO uptake on a per-lesion basis. One hundred head and neck cancer patients undergoing chemoradiotherapy treatment at a single institution between 2011 and 2016 received pre-treatment FDG and FMISO PET/CT. A total of 134 lesions delineated by physicians on the treatment planning CT and with a volume larger than 10 cc were analyzed. Forty-six lesions were held out as a validation set. The level of hypoxia was defined as the maximum tumor-to-blood ratio (TBRmax) of FMISO uptake measured at ∼150 minutes post injection. Approximately 200 radiomics features were extracted from FDG PET (first order and run length features only) and treatment planning CT (all orders, including run length), and used to produce a prediction of FMISO TBRmax. The most frequent features were selected by lasso using 100×10-fold cross validation (CV). The resulting features and their bilinear interaction terms were used again in a lasso regression model. The final combination of predictors was found by minimizing the mean square error on nested 10-fold CV. Performance was assessed by means of (i) R2, (ii) F-test P-value for a constant-model null hypothesis, and (iii) area under the curve (AUC) analysis, assuming a hypoxia threshold of TBRmaxthresh = 1.4. The mean FMISO TBRmax of the population was 1.82 (SD = 0.65). The resulting model contained two individual FDG PET features (median SUV and low gray-level run emphasis) and three interaction terms with other PET run-length and CT Haralick correlation features. In the training dataset (n = 88) the resulting prediction had an adjusted R2 = 0.42. For a hypoxia threshold of TBRmaxthresh = 1.4, the AUC was 0.86. In the validation dataset the prediction had an adjusted R2 of 0.17, with F-test P = 0.002. Assuming TBRmaxthresh = 1.4, the AUC for the validation dataset was 0.79, with an optimal operating point of FPR = 0.21 and TPR = 0.78. We found that FDG PET and CT provide complementary information that can be used to form a radiomics biomarker that correlates with tumor hypoxia as evaluated by FMISO PET. Further validation using multi-institutional datasets is required. Clinically, this is a promising approach to identify a subset of head and neck radiotherapy patients who may benefit from hypoxia-based dose modification.

Concentraciones de hormona estimulante de la tiroides y tiroxina libre en ovinos jóvenes

El objetivo de esta investigación fue establecer valores de referencia en las concentraciones de la hormona estimulante de la tiroides (TSH, del inglés Thyroid-Stimulating Hormone) y tiroxina (T4) libre, para los pequeños rumiantes menores de 7 meses, analizando la influencia del sexo sobre las determinaciones. Se obtuvieron muestras de sangre de 48 ovinos en estado de ayuno (25 machos y 23 hembras menores de 7 meses de edad), las concentraciones de TSH y T4 libre se determinaron mediante inmunoensayo enzimático. Los valores de TSH para las hembras jóvenes (μUI/ml) fueron: promedio, mínimo, máximo y desviación estándar de 0,34; 0,0; 0,57; 0,15, respectivamente. Para los machos jóvenes fueron: 0,09; 0,0; 0,57; 0,15, respectivamente. El valor p del test F fue 0,147 (p ≥ 0,05), por lo cual no hay diferencia estadísticamente significativa con una confiabilidad de 95 % para la TSH en ovinos jóvenes con respecto al género. Los valores de T4 libre para las hembras jóvenes (ng/dl) fueron: promedio, mínimo, máximo y desviación estándar de 0,44; 0,13; 0,73; 0,21, respectivamente. Para los machos jóvenes fueron: 0,66; 0,31; 1,17; 0,24, respectivamente. El valor de p del test F es 0,04 (p &lt; 0,05), lo cual evidencia diferencia significativa con una confiabilidad del 95 % para los valores de T4 libre en ovinos jóvenes. Se concluye que el sexo causa variación en la secreción de T4 libre en los pequeños rumiantes jóvenes.

The intravertebral distribution of bone density: correspondence to intervertebral disc health and implications for vertebral strength

This study's goal was to determine associations among the intravertebral heterogeneity in bone density, bone strength, and intervertebral disc (IVD) health. Results indicated that predictions of vertebral strength can benefit from considering the magnitude of the density heterogeneity and the congruence between the spatial distribution of density and IVD health. This study aims to determine associations among the intravertebral heterogeneity in bone density, bone strength, and IVD health Regional measurements of bone density were performed throughout 30 L1 vertebral bodies using micro-computed tomography (μCT) and quantitative computed tomography (QCT). The magnitude of the intravertebral heterogeneity in density was defined as the interquartile range and quartile coefficient of variation in regional densities. The spatial distribution of density was quantified using ratios of regional densities representing different anatomical zones (e.g., anterior to posterior regional densities). Cluster analysis was used to identify groups of vertebrae with similar spatial distributions of density. Vertebral strength was measured in compression. IVD health was assessed using two scoring systems. QCT- and μCT-based measures of the magnitude of the intravertebral heterogeneity in density were strongly correlated with each other (p < 0.005). Accounting for the interquartile range in regional densities improved predictions of vertebral strength as compared to predictions based only on mean density (R (2) = 0.59 vs. 0.43; F-test p-value = 0.018). Specifically, after adjustment for mean density, vertebral bodies with greater heterogeneity in density exhibited higher strength. No single spatial distribution of density was associated with high vertebral strength. Analyses of IVD scores suggested that the health of the adjacent IVDs may modulate the effect of a particular spatial distribution of density on vertebral strength. Noninvasive measurements of the intravertebral distribution of bone density, in conjunction with assessments of IVD health, can aid in predictions of bone strength and in elucidating biomechanical mechanisms of vertebral fracture.

Accuracy of p-values of approximate tests in testing for equality of means under unequal variances

Abstract Seemingly, testing for fixed effects in linear models with variance-covariance components has been solved for decades. However, even in simple situations such as in fixed one-way model with heteroscedastic variances (a multiple means case of the Behrens-Fisher problem) the questions of statistical properties of various approximations of test statistics are still alive. Here we present a brief overview of several approaches suggested in the literature as well as those available in statistical software, accompanied by a simulation study in which the accuracy of p-values is studied. Our interest is limited here to the Welch’s test, the Satterthwaite-Fai-Cornelius test, the Kenward-Roger test, the simple ANOVA F-test, and the parametric bootstrap test. We conclude that for small sample sizes, regardless the number of compared means and the heterogeneity of variance, the ANOVA F-test p-value performs the best. For higher sample sizes (at least 5 per group), the parametric bootstrap performs well, and the Kenward-Roger test also performs well.