Multivariate test power approximations for balanced linear mixed models in studies with missing data.

Multilevel and longitudinal studies are frequently subject to missing data. For example, biomarker studies for oral cancer may involve multiple assays for each participant. Assays may fail, resulting in missing data values that can be assumed to be missing completely at random. Catellier and Muller proposed a data analytic technique to account for data missing at random in multilevel and longitudinal studies. They suggested modifying the degrees of freedom for both the Hotelling-Lawley trace F statistic and its null case reference distribution. We propose parallel adjustments to approximate power for this multivariate test in studies with missing data. The power approximations use a modified non-central F statistic, which is a function of (i) the expected number of complete cases, (ii) the expected number of non-missing pairs of responses, or (iii) the trimmed sample size, which is the planned sample size reduced by the anticipated proportion of missing data. The accuracy of the method is assessed by comparing the theoretical results to the Monte Carlo simulated power for the Catellier and Muller multivariate test. Over all experimental conditions, the closest approximation to the empirical power of the Catellier and Muller multivariate test is obtained by adjusting power calculations with the expected number of complete cases. The utility of the method is demonstrated with a multivariate power analysis for a hypothetical oral cancer biomarkers study. We describe how to implement the method using standard, commercially available software products and give example code. Copyright © 2015 John Wiley & Sons, Ltd.

Ultrasound-based clinical prediction rule model for detecting papillary thyroid cancer in cervical lymph nodes: A pilot study.

PURPOSE: To identify sonographic features of cervical lymph nodes (LNs) that are associated with papillary thyroid cancer (PTC) and to develop a prediction model for classifying nodes as metastatic or benign. METHODS: This retrospective study included the records of postthyroidectomy patients with PTC who had undergone cervical ultrasound and LN biopsy. LN location, size, shape, hilum, echopattern, Doppler flow, and microcalcifications were assessed. Model selection was used to identify features associated with malignant LNs and to build a predictive, binary-outcome, generalized linear mixed model. A cross-validated receiver operating characteristic analysis was conducted to assess the accuracy of the model for classifying metastatic nodes. RESULTS: We analyzed records from 71 LNs (23 metastatic) in 44 patients (16 with PTC). The predictive model included a nonhomogeneous echopattern (odds ratio [OR], 5.73; 95% confidence interval [CI], 1.07-30.74; p = 0.04), microcalcifications (OR, 4.91; 95% CI, 0.91-26.54; p = 0.06), and volume (OR, 2.57; 95% CI, 0.66-9.99; p = 0.16) as predictors. The model had an area under the curve of 0.74 (95% CI, 0.60-0.85), sensitivity of 65% (95% CI, 50% to 78%), and specificity of 85% (95% CI, 73% to 94%) at the Youden optimal cut point of 0.38. CONCLUSIONS: Nonhomogeneous echopattern, microcalcifications, and node volume were predictive of malignant LNs in patients with PTC. A larger sample is needed to validate this model.

Effect of Change in Portal Venous Blood Flow Rates on the Performance of a 2.45-GHz Microwave Ablation Device.

PURPOSE: To investigate the effect of change in portal venous blood flow rates on the size and shape of ablations created by a 2.45-GHz microwave ablation device. MATERIALS AND METHODS: This study was exempt from review by the institutional animal care and use committee. An in vitro bovine liver model perfused with autologous blood via the portal vein at five flow rates (60, 70, 80, 90, and 100 mL/min per 100 g of liver) was used to evaluate the effect of change in flow rates on the size and shape of coagulation created by a 2.45-GHz, 140-W microwave ablation device operated for 5 and 10 minutes. Three ablations per ablation time were conducted in each of 10 livers, with two livers perfused at each flow rate. Short- and long-axis diameters were measured from gross specimens, and volume and sphericity index were calculated. General linear mixed models that accounted for correlations within the liver were used to evaluate the effects of lobe, flow, and ablation time on size and sphericity index of ablations. RESULTS: Flow did not have a significant effect on the size or shape of coagulation created at 5 or 10 minutes (P > .05 for all tests). The mean short- and long-axis diameters and volume were 3.2 cm (95% confidence interval [CI]: 3.1, 3.3), 5.6 cm (95% CI: 5.4, 5.8), and 30.2 cm(3) (95% CI: 28.4, 32.1) for the 5-minute ablations and 3.8 cm (95% CI: 3.7, 3.9), 6.5 cm (95% CI: 6.3, 6.7), and 49.3 cm(3) (95% CI: 47.5, 51.2), for the 10-minute ablations, respectively. The mean sphericity index for both 5- and 10-minute ablations was 34.4% (95% CI: 32%, 36.7%). CONCLUSION: Change in portal venous blood flow rates did not have an effect on the size and shape of ablations created by a 2.45-GHz microwave ablation device.

Recommendations for choosing an analysis method that controls Type I error for unbalanced cluster sample designs with Gaussian outcomes.

We used theoretical and simulation-based approaches to study Type I error rates for one-stage and two-stage analytic methods for cluster-randomized designs. The one-stage approach uses the observed data as outcomes and accounts for within-cluster correlation using a general linear mixed model. The two-stage model uses the cluster specific means as the outcomes in a general linear univariate model. We demonstrate analytically that both one-stage and two-stage models achieve exact Type I error rates when cluster sizes are equal. With unbalanced data, an exact size α test does not exist, and Type I error inflation may occur. Via simulation, we compare the Type I error rates for four one-stage and six two-stage hypothesis testing approaches for unbalanced data. With unbalanced data, the two-stage model, weighted by the inverse of the estimated theoretical variance of the cluster means, and with variance constrained to be positive, provided the best Type I error control for studies having at least six clusters per arm. The one-stage model with Kenward-Roger degrees of freedom and unconstrained variance performed well for studies having at least 14 clusters per arm. The popular analytic method of using a one-stage model with denominator degrees of freedom appropriate for balanced data performed poorly for small sample sizes and low intracluster correlation. Because small sample sizes and low intracluster correlation are common features of cluster-randomized trials, the Kenward-Roger method is the preferred one-stage approach.

Survival outcomes in patients with advanced hepatocellular carcinoma treated with drug-eluting bead chemoembolization.

OBJECTIVE. The purpose of this study was to determine the overall survival rates in patients with advanced hepatocellular carcinoma (HCC) who undergo treatment with drug-eluting bead (DEB) therapy. MATERIALS AND METHODS. A retrospective review of the clinical HCC database of a single institution was undertaken for patients treated between September 2008 and December 2011. Demographic information, laboratory and imaging findings, procedural details, and outcomes after treatment were obtained. The primary outcome was overall survival, which was stratified by Barcelona Clinic Liver Cancer (BCLC) stage, Child-Pugh class, Eastern Cooperative Oncology Group (ECOG) score, serum bilirubin level, and ethnicity. Multiple secondary independent variables were also measured. RESULTS. Of 239 consecutive patients treated during the prescribed time frame, 43 patients met the inclusion criteria. Thirty patients met the criteria for BCLC stage C, and 13 met the criteria for BCLC stage D based largely on ECOG score. Eight patients had venous invasion or portal venous thrombosis, and four had limited extrahepatic metastases. Eight patients had Child-Pugh class C liver disease but remained candidates for liver transplant based on the Milan criteria. The median overall survival was 596 days; 23 patients are still alive, 12 of whom underwent liver transplant. The only independent variables affecting survival were serum bilirubin value of 2.0 mg/dL or greater (hazard ratio [HR] = 3.96; 95% CI, 1.46-10.7; p = 0.007) and Child-Pugh class B or C disease (HR = 3.33; 95% CI, 1.07-10.34; p = 0.037). CONCLUSION. The use of DEBs for TACE therapy is safe and effective in carefully selected patients with advanced HCC.

Reducing decision errors in the paired comparison of the diagnostic accuracy of screening tests with Gaussian outcomes.

BACKGROUND: Scientists often use a paired comparison of the areas under the receiver operating characteristic curves to decide which continuous cancer screening test has the best diagnostic accuracy. In the paired design, all participants are screened with both tests. Participants with suspicious results or signs and symptoms of disease receive the reference standard test. The remaining participants are classified as non-cases, even though some may have occult disease. The standard analysis includes all study participants, which can create bias in the estimates of diagnostic accuracy since not all participants receive disease status verification. We propose a weighted maximum likelihood bias correction method to reduce decision errors. METHODS: Using Monte Carlo simulations, we assessed the method's ability to reduce decision errors across a range of disease prevalences, correlations between screening test scores, rates of interval cases and proportions of participants who received the reference standard test. RESULTS: The performance of the method depends on characteristics of the screening tests and the disease and on the percentage of participants who receive the reference standard test. In studies with a large amount of bias in the difference in the full areas under the curves, the bias correction method reduces the Type I error rate and improves power for the correct decision. We demonstrate the method with an application to a hypothetical oral cancer screening study. CONCLUSION: The bias correction method reduces decision errors for some paired screening trials. In order to determine if bias correction is needed for a specific screening trial, we recommend the investigator conduct a simulation study using our software.

Downstaging disease in patients with hepatocellular carcinoma outside of Milan criteria: strategies using drug-eluting bead chemoembolization.

PURPOSE: To assess downstaging rates in patients with United Network for Organ Sharing stage T3N0M0 hepatocellular carcinoma (HCC) treated with doxorubicin-eluting bead transarterial chemoembolization to meet Milan criteria for transplantation. MATERIALS AND METHODS: A single-center retrospective review of 239 patients treated with doxorubicin-eluting bead (DEB) chemoembolization between September 2008 and December 2011 was undertaken. Baseline and follow-up computed tomography or magnetic resonance imaging was assessed for response based on the longest enhancing axial dimension of each tumor (ie, modified Response Evaluation Criteria In Solid Tumors measurements), and medical records were reviewed. Fisher exact tests and exact logistic regression were used to test the association of patient and disease characteristics with downstaging. RESULTS: After exclusions, 22 patients remained in the analysis, 17 of whom (77%) had their HCC downstaged to meet Milan criteria. Among those whose disease was downstaged, seven underwent transplantation, one remained listed for transplantation, six had disease progression beyond Milan criteria, two underwent conventional transarterial chemoembolization, and one underwent radiofrequency ablation. The seven patients who received transplants were still living, but recurrent HCC developed in two. Baseline age (P = .25), Model for End-stage Liver Disease score (P = .77), and α-fetoprotein (AFP) level (P = 1.00) were similar between patients with and without downstaged HCC. No associations were observed between the odds of downstaging and sex (P = .21), Child-Pugh class (P = .14), Child-Pugh class controlling for baseline tumor multiplicity (P = .15), Eastern Cooperative Oncology Group performance status (P = 1.00), tumor burden (P = .31), multiple tumors (P = .31), or hepatitis C virus infection (P = 1.00). Fifteen patients who did not receive transplants were alive at 1 year, with two progression-free. Baseline AFP levels differed between those who survived 1 year and those who did not (P = .02), but did not differ by progression-free survival status (P = .62). CONCLUSIONS: T3N0M0 HCC treatment with DEB chemoembolization has a high likelihood (77%) of downstaging the disease to meet Milan criteria.

GLIMMPSE: Online Power Computation for Linear Models with and without a Baseline Covariate.

GLIMMPSE is a free, web-based software tool that calculates power and sample size for the general linear multivariate model with Gaussian errors (http://glimmpse.SampleSizeShop.org/). GLIMMPSE provides a user-friendly interface for the computation of power and sample size. We consider models with fixed predictors, and models with fixed predictors and a single Gaussian covariate. Validation experiments demonstrate that GLIMMPSE matches the accuracy of previously published results, and performs well against simulations. We provide several online tutorials based on research in head and neck cancer. The tutorials demonstrate the use of GLIMMPSE to calculate power and sample size.

A power approximation for the Kenward and Roger Wald test in the linear mixed model.

We derive a noncentral [Formula: see text] power approximation for the Kenward and Roger test. We use a method of moments approach to form an approximate distribution for the Kenward and Roger scaled Wald statistic, under the alternative. The result depends on the approximate moments of the unscaled Wald statistic. Via Monte Carlo simulation, we demonstrate that the new power approximation is accurate for cluster randomized trials and longitudinal study designs. The method retains accuracy for small sample sizes, even in the presence of missing data. We illustrate the method with a power calculation for an unbalanced group-randomized trial in oral cancer prevention.

Digital Breast Tomosynthesis Utilization in the United States: A Survey of Physician Members of the Society of Breast Imaging.

PURPOSE: To assess utilization of digital breast tomosynthesis (DBT) and examine criteria for offering DBT to patients. METHODS: We created an online survey for physician members of the Society of Breast Imaging to assess their use of DBT. The questions covered availability of DBT at the participant's practice, whether DBT was used for clinical care or research, clinical decision rules guiding patient selection for DBT, costs associated with DBT, plans to obtain DBT, and breast imaging practice characteristics. Fisher's exact tests and logistic regression were used to compare DBT users and nonusers. RESULTS: In all, 670 members responded (response rate = 37%). Of these, 200 (30.0%) respondents reported using DBT, with 89% of these using DBT clinically. Participants were more likely to report DBT use if they worked at an academic practice (odds ratio [OR], 2.07; 95% confidence interval [CI], 1.41 to 3.03; P < .001), a practice with more than 3 breast imagers (OR, 2.36; 95% CI, 1.62 to 3.43; P < .001), or a practice with 7 or more mammography units (OR, 3.05; 95% CI, 2.11 to 4.39; P < .001). Criteria used to select patients to undergo DBT varied, with 107 (68.2%) using exam type (screening versus diagnostic), 25 (15.9%) using mammographic density, and 25 (15.9%) using breast cancer risk. Fees for DBT ranged from $25 to $250. In addition, 62.3% of nonusers planned to obtain DBT. CONCLUSION: DBT is becoming more common but remains a limited resource. Clinical guidelines would assist practices in deciding whether to adopt DBT and in standardizing which patients should receive DBT.

Correlation between early 18F-FDG PET/CT response to BRAF and MEK inhibition and survival in patients with BRAF-mutant metastatic melanoma.

PURPOSE: Metabolic response to treatment measured by fluorine-18 fluorodeoxyglucose ((18)F-FDG) PET has prognostic implications in many cancers. This study investigated the association between survival and early changes on (18)F-FDG PET/computed tomography (CT) for patients with BRAF-mutant melanoma receiving combined BRAF and MEK inhibition therapy. MATERIALS AND METHODS: Overall, 24 patients with advanced BRAF-mutant melanoma were included. Patients were treated with a BRAF inhibitor (vemurafenib or dabrafenib) and a MEK inhibitor (cobimetinib or trametinib), and were imaged at baseline and shortly thereafter with (18)F-FDG PET/CT. Each scan yielded two values of maximum standardized uptake value (SUVmax): one for the most metabolically active focus and one for the least responsive focus. Short-term treatment response was assessed by evaluating the target lesions using the EROTC criteria. A Cox proportional hazards model was used to examine associations between overall survival (OS) and progression-free survival (PFS) and changes in SUVmax. RESULTS: The mean time to follow-up (18)F-FDG PET/CT was 26 days. At follow-up, two patients achieved a complete response. For the most metabolically active focus, 22 patients showed a partial response. For the least responsive focus, 18 patients showed a partial response, two had stable disease, and two had progressive disease.A total of 16 patients were alive at the end of the study. For the most metabolically active tumor, no association was observed between changes in SUVmax and OS (P=0.73) or PFS (P=0.17). For the least responsive tumor, change in SUVmax was associated with PFS [hazard ratio (HR)=1.34, 95% confidence interval (CI): 1.06-1.71, P=0.01], but not OS (P=0.52). The ECOG score was associated with OS (HR=11.81, 95% CI: 1.42-97.60, P=0.02) and PFS (HR=24.72, 95% CI: 3.23-189.42, P=0.002). CONCLUSION: Change in SUVmax for the least responsive tumor and baseline functional performance may be useful prognostic indicators for PFS in patients with BRAF-mutant melanoma.

Digital breast tomosynthesis utilization in the United States: a survey of physician members of the Society of Breast Imaging.

PURPOSE: To assess utilization of digital breast tomosynthesis (DBT) and examine criteria for offering DBT to patients. METHODS: We created an online survey for physician members of the Society of Breast Imaging to assess their use of DBT. The questions covered availability of DBT at the participant's practice, whether DBT was used for clinical care or research, clinical decision rules guiding patient selection for DBT, costs associated with DBT, plans to obtain DBT, and breast imaging practice characteristics. Fisher's exact tests and logistic regression were used to compare DBT users and nonusers. RESULTS: In all, 670 members responded (response rate = 37%). Of these, 200 (30.0%) respondents reported using DBT, with 89% of these using DBT clinically. Participants were more likely to report DBT use if they worked at an academic practice (odds ratio [OR], 2.07; 95% confidence interval [CI], 1.41 to 3.03; P < .001), a practice with more than 3 breast imagers (OR, 2.36; 95% CI, 1.62 to 3.43; P < .001), or a practice with 7 or more mammography units (OR, 3.05; 95% CI, 2.11 to 4.39; P < .001). Criteria used to select patients to undergo DBT varied, with 107 (68.2%) using exam type (screening versus diagnostic), 25 (15.9%) using mammographic density, and 25 (15.9%) using breast cancer risk. Fees for DBT ranged from $25 to $250. In addition, 62.3% of nonusers planned to obtain DBT. CONCLUSION: DBT is becoming more common but remains a limited resource. Clinical guidelines would assist practices in deciding whether to adopt DBT and in standardizing which patients should receive DBT.

GLIMMPSE Lite: calculating power and sample size on smartphone devices.

Researchers seeking to develop complex statistical applications for mobile devices face a common set of difficult implementation issues. In this work, we discuss general solutions to the design challenges. We demonstrate the utility of the solutions for a free mobile application designed to provide power and sample size calculations for univariate, one-way analysis of variance (ANOVA), GLIMMPSE Lite. Our design decisions provide a guide for other scientists seeking to produce statistical software for mobile platforms.