Row versus column correlations: avoiding the ecological fallacy in RNA/protein expression studies.

Biomedical researchers are often interested in computing the correlation between RNA and protein abundance. However, correlations can be computed between rows of a data matrix or between columns, and the results are not the same. The belief that these two types of correlation are estimating the same phenomenon is a special case of a well-known logical error called the ecological fallacy. In this article, we review different uses of correlation found in the literature, explain the differences between row and column correlations and argue that one of them has an undesirable interpretation in most applications. Through simulation studies and theoretical derivations, we show that the commonly used Pearson's coefficient, computed from protein and transcript data from a single sample, is only loosely related to the biological correlation that most researchers will be interested in studying. Beyond our basic exploration of the ecological fallacy, we examine how correlations are affected by relative quantification proteomics data and common normalization procedures, finding that double normalization is capable of completely masking true correlative relationships. We conclude with guidelines for properly identifying and computing consistent correlation coefficients.

Accelerating high-dimensional clustering with lossless data reduction.

MOTIVATION: For cluster analysis, high-dimensional data are associated with instability, decreased classification accuracy and high-computational burden. The latter challenge can be eliminated as a serious concern. For applications where dimension reduction techniques are not implemented, we propose a temporary transformation which accelerates computations with no loss of information. The algorithm can be applied for any statistical procedure depending only on Euclidean distances and can be implemented sequentially to enable analyses of data that would otherwise exceed memory limitations. RESULTS: The method is easily implemented in common statistical software as a standard pre-processing step. The benefit of our algorithm grows with the dimensionality of the problem and the complexity of the analysis. Consequently, our simple algorithm not only decreases the computation time for routine analyses, it opens the door to performing calculations that may have otherwise been too burdensome to attempt. AVAILABILITY AND IMPLEMENTATION: R, Matlab and SAS/IML code for implementing lossless data reduction is freely available in the Appendix. CONTACT: obrienj@hms.harvard.edu.

QuantFusion: Novel Unified Methodology for Enhanced Coverage and Precision in Quantifying Global Proteomic Changes in Whole Tissues.

Single quantitative platforms such as label-based or label-free quantitation (LFQ) present compromises in accuracy, precision, protein sequence coverage, and speed of quantifiable proteomic measurements. To maximize the quantitative precision and the number of quantifiable proteins or the quantifiable coverage of tissue proteomes, we have developed a unified approach, termed QuantFusion, that combines the quantitative ratios of all peptides measured by both LFQ and label-based methodologies. Here, we demonstrate the use of QuantFusion in determining the proteins differentially expressed in a pair of patient-derived tumor xenografts (PDXs) representing two major breast cancer (BC) subtypes, basal and luminal. Label-based in-spectra quantitative peptides derived from amino acid-coded tagging (AACT, also known as SILAC) of a non-malignant mammary cell line were uniformly added to each xenograft with a constant predefined ratio, from which Ratio-of-Ratio estimates were obtained for the label-free peptides paired with AACT peptides in each PDX tumor. A mixed model statistical analysis was used to determine global differential protein expression by combining complementary quantifiable peptide ratios measured by LFQ and Ratio-of-Ratios, respectively. With minimum number of replicates required for obtaining the statistically significant ratios, QuantFusion uses the distinct mechanisms to "rescue" the missing data inherent to both LFQ and label-based quantitation. Combined quantifiable peptide data from both quantitative schemes increased the overall number of peptide level measurements and protein level estimates. In our analysis of the PDX tumor proteomes, QuantFusion increased the number of distinct peptide ratios by 65%, representing differentially expressed proteins between the BC subtypes. This quantifiable coverage improvement, in turn, not only increased the number of measurable protein fold-changes by 8% but also increased the average precision of quantitative estimates by 181% so that some BC subtypically expressed proteins were rescued by QuantFusion. Thus, incorporating data from multiple quantitative approaches while accounting for measurement variability at both the peptide and global protein levels make QuantFusion unique for obtaining increased coverage and quantitative precision for tissue proteomes.

Group testing for case identification with correlated responses.

This article examines group testing procedures where units within a group (or pool) may be correlated. The expected number of tests per unit (i.e., efficiency) of hierarchical- and matrix-based procedures is derived based on a class of models of exchangeable binary random variables. The effect on efficiency of the arrangement of correlated units within pools is then examined. In general, when correlated units are arranged in the same pool, the expected number of tests per unit decreases, sometimes substantially, relative to arrangements that ignore information about correlation.

Deletion diagnostics for alternating logistic regressions.

Deletion diagnostics are introduced for the regression analysis of clustered binary outcomes estimated with alternating logistic regressions, an implementation of generalized estimating equations (GEE) that estimates regression coefficients in a marginal mean model and in a model for the intracluster association given by the log odds ratio. The diagnostics are developed within an estimating equations framework that recasts the estimating functions for association parameters based upon conditional residuals into equivalent functions based upon marginal residuals. Extensions of earlier work on GEE diagnostics follow directly, including computational formulae for one-step deletion diagnostics that measure the influence of a cluster of observations on the estimated regression parameters and on the overall marginal mean or association model fit. The diagnostic formulae are evaluated with simulations studies and with an application concerning an assessment of factors associated with health maintenance visits in primary care medical practices. The application and the simulations demonstrate that the proposed cluster-deletion diagnostics for alternating logistic regressions are good approximations of their exact fully iterated counterparts.

Non-small cell lung cancer: prognostic importance of positive FDG PET findings in the mediastinum for patients with N0-N1 disease at pathologic analysis.

PURPOSE: To assess the prognostic implications of mediastinal positron emission tomographic (PET) findings in patients undergoing curative resection of non-small cell lung cancer (NSCLC) who have histologically negative mediastinal lymph nodes (LNs), with the hypothesis that positive findings at PET are prognostic even in patients with negative histologic findings in the LNs. MATERIALS AND METHODS: Records of patients with a preoperative PET undergoing curative surgery, without adjuvant radiation, for pathologic T1-3N0-1 NSCLC at the University of North Carolina between 2000 and 2006 were reviewed as an institutional review board-approved HIPAA-compliant retrospective study. Ninety patients were evaluable (all histologically negative in mediastinum; 44 with both mediastinoscopy and surgery); 13 patients had positive mediastinal PET findings, and 77 had negative mediastinal PET findings. Local-regional and distant failure rates in patients with and those without mediastinal abnormalities at preoperative PET were compared by using logistic regression and log-rank tests. RESULTS: Median follow-up was 54.3 months (range, 1-99 months). There were higher rates of local-regional (P = .001) and distant (P < .001) failure as well as death (P = .001) in patients with postive PET findings than in patients with negative findings. In multivariable analysis (adjusting for other prognostic factors), positive PET findings in the mediastinum remained prognostic for distant failure (P < .001, hazard ratio = 6.9) and were marginally prognostic for local-regional failure (P = .093, hazard ratio = 1.9). CONCLUSION: Positive findings at preoperative PET in the mediastinum appear to have prognostic implications despite the mediastinal LNs being histologically negative. The high rate of local-regional and distant failure suggests that postoperative radiation therapy and/or chemotherapy may be particularly helpful in patients with positive mediastinal findings at preoperative PET.

An R2 statistic for fixed effects in the linear mixed model.

Statisticians most often use the linear mixed model to analyze Gaussian longitudinal data. The value and familiarity of the R(2) statistic in the linear univariate model naturally creates great interest in extending it to the linear mixed model. We define and describe how to compute a model R(2) statistic for the linear mixed model by using only a single model. The proposed R(2) statistic measures multivariate association between the repeated outcomes and the fixed effects in the linear mixed model. The R(2) statistic arises as a 1-1 function of an appropriate F statistic for testing all fixed effects (except typically the intercept) in a full model. The statistic compares the full model with a null model with all fixed effects deleted (except typically the intercept) while retaining exactly the same covariance structure. Furthermore, the R(2) statistic leads immediately to a natural definition of a partial R(2) statistic. A mixed model in which ethnicity gives a very small p-value as a longitudinal predictor of blood pressure (BP) compellingly illustrates the value of the statistic. In sharp contrast to the extreme p-value, a very small R(2) , a measure of statistical and scientific importance, indicates that ethnicity has an almost negligible association with the repeated BP outcomes for the study.

The effects of nonignorable missing data on label-free mass spectrometry proteomics experiments.

An idealized version of a label-free discovery mass spectrometry proteomics experiment would provide absolute abundance measurements for a whole proteome, across varying conditions. Unfortunately, this ideal is not realized. Measurements are made on peptides requiring an inferential step to obtain protein level estimates. The inference is complicated by experimental factors that necessitate relative abundance estimation and result in widespread non-ignorable missing data. Relative abundance on the log scale takes the form of parameter contrasts. In a complete-case analysis, contrast estimates may be biased by missing data and a substantial amount of useful information will often go unused. To avoid problems with missing data, many analysts have turned to single imputation solutions. Unfortunately, these methods often create further difficulties by hiding inestimable contrasts, preventing the recovery of interblock information and failing to account for imputation uncertainty. To mitigate many of the problems caused by missing values, we propose the use of a Bayesian selection model. Our model is tested on simulated data, real data with simulated missing values, and on a ground truth dilution experiment where all of the true relative changes are known. The analysis suggests that our model, compared with various imputation strategies and complete-case analyses, can increase accuracy and provide substantial improvements to interval coverage.

Comparison of subject-specific and population averaged models for count data from cluster-unit intervention trials.

Maximum likelihood estimation techniques for subject-specific (SS) generalized linear mixed models and generalized estimating equations for marginal or population-averaged (PA) models are often used for the analysis of cluster-unit intervention trials. Although both classes of procedures account for the presence of within-cluster correlations, the interpretations of fixed effects including intervention effect parameters differ in SS and PA models. Furthermore, closed-form mathematical expressions relating SS and PA parameters from the two respective approaches are generally lacking. This paper investigates the special case of correlated Poisson responses where, for a log-linear model with normal random effects, exact relationships are available. Equivalent PA model representations of two SS models commonly used in the analysis of nested cross-sectional cluster trials with count data are derived. The mathematical results are illustrated with count data from a large non-randomized cluster trial to reduce underage drinking. Knowledge of relationships among parameters in the respective mean and covariance models is essential to understanding empirical comparisons of the two approaches.

Validation of the palliative performance scale in the acute tertiary care hospital setting.

BACKGROUND: Physicians are often asked to prognosticate patient survival. However, prediction of survival is difficult, particularly with critically ill and dying patients within the hospitals. The Palliative Performance Scale (PPS) was designed to assess functional status and measure progressive decline in palliative care patients, yet it has not been validated within hospital health care settings. OBJECTIVE: This study explores the application of the PPS for its predictive ability related to length of survival. Other variables examined were correlates of symptom distress in a tertiary academic setting. METHODS: Patients were assigned a score on the PPS ranging from 0% to 100% at initial consultation. Standardized symptom assessments were carried out daily, and survival was determined by medical record review and search of the National Death Index. RESULTS: Of 261 patients seen since January 2002, 157 had cancer and 104 had other diagnoses. PPS scores ranged from 10% to 80% with 92% of the scores between 10% and 40%. Survival ranged from 0 to 30 months, with a median of 9 days. By 90 days, 83% of patients had died. Proportional hazards regression estimates showed that a 10% decrement in PPS score was associated with a hazard ratio of 1.65 (95% confidence interval [CI]: 1.42-1.92). Proportional odds regression models showed that a lower PPS was significantly associated with higher levels of dyspnea. CONCLUSION: The PPS correlated well with length of survival and with select symptom distress scores. We consider it to be a useful tool in predicting outcomes for palliative care patients.

Divergent Specificity Development of IgG1 and IgG4 Autoantibodies in Endemic Pemphigus Foliaceus (Fogo Selvagem).

We have shown that although the IgG response in fogo selvagem (FS) is mainly restricted to desmoglein (Dsg) 1, other keratinocyte cadherins are also targeted by FS patients and healthy control subjects living in the endemic region of Limão Verde, Brazil (endemic controls). Evaluating nonpathogenic IgG1 and pathogenic IgG4 subclass responses to desmosomal proteins may reveal important differences between pathogenic and nonpathogenic responses, and how these differences relate to the pathogenic IgG4 response and resultant FS. In this study, we tested by ELISA >100 sera from each FS patient, endemic control, and nonendemic control for IgG1 and IgG4 autoantibodies to keratinocyte cadherins besides Dsg1. IgG1 and IgG4 subclass responses in endemic controls are highly correlated between Dsg1 and other keratinocyte cadherins. This correlation persists in the IgG1 response among FS patients, but diminishes in IgG4 response, suggesting that IgG1 binds highly conserved linear epitopes among cadherins, whereas IgG4 binds mainly specific conformational epitopes on Dsg1. A confirmatory test comparing serum samples of 11 individuals before and after their FS onset substantiated our findings that IgG1 recognizes primarily linear epitopes on Dsg1 both before and after disease onset, whereas IgG4 recognizes primarily linear epitopes before disease onset, but recognizes more conformational epitopes on Dsg1 after the onset of disease. This study may provide a mechanism by which a specificity convergence of the IgG4 response to unique Dsg1 epitopes, most likely conformational pathogenic epitopes, leads to the onset of FS disease.

Cancer Risk after Fat Transfer: A Multicenter Case-Cohort Study.

BACKGROUND: Fat transfer is an increasingly popular method for refining postmastectomy breast reconstructions. However, concern persists that fat transfer may promote disease recurrence. Adipocytes are derived from adipose-derived stem cells and express adipocytokines that can facilitate active breast cancer cells in laboratory models. The authors sought to evaluate the association between fat transfer to the reconstructed breast and cancer recurrence in patients diagnosed with local or regional invasive breast cancers. METHODS: A multicenter, case-cohort study was performed. Eligible patients from four centers (Memorial Sloan Kettering, M. D. Anderson Cancer Center, Alvin J. Siteman Cancer Center, and the University of Chicago) were identified by each site's institutional tumor registry or cancer data warehouse. Eligibility criteria were as follows: mastectomy with immediate breast reconstruction between 2006 and 2011, age older than 21 years, female sex, and incident diagnosis of invasive ductal carcinoma (stage I, II, or III). Cases consisted of all recurrences during the study period, and controls consisted of a 30 percent random sample of the study population. Cox proportional hazards regression was used to evaluate for association between fat transfer and time to recurrence in bivariate and multivariate models. RESULTS: The time to disease recurrence unadjusted hazard ratio for fat transfer was 0.99 (95 percent CI, 0.56 to 1.7). After adjustment for age, body mass index, stage, HER2/Neu receptor status, and estrogen receptor status, the hazard ratio was 0.97 (95 percent CI, 0.54 to 1.8). CONCLUSION: In this population of breast cancer patients who had mastectomy with immediate reconstruction, fat transfer was not associated with a higher risk of cancer recurrence. CLINICAL QUESTION/LEVEL OF EVIDENCE: Therapeutic, III.

Prevalence of anti-desmoglein-3 antibodies in endemic regions of Fogo selvagem in Brazil.

Fogo selvagem (FS), the endemic form of pemphigus foliaceus (PF), is an autoimmune blistering disease characterized by autoantibodies against desmoglein 1. The Terena reservation of Limao Verde in Mato Grosso do Sul, Brazil, is a previously identified focus of disease. Autoantibodies against desmoglein 3 (Dsg3) have also been detected in sera from patients with FS. In an effort to further characterize the serological, geographical, and clinical epidemiology of the disease, we sought to determine the prevalence of anti-Dsg3 autoantibodies in sera from normal subjects living outside of and in an endemic area using an ELISA. Anti-Dsg3 antibodies were detected in 53 of 146 normal subjects from Limao Verde (36%), and in eight of 140 normal subjects from surrounding areas (6%). A significant trend was observed in the proportion of positive tests relative to distance from the endemic area (P < 0.001). Our seroepidemiological observations support the concept that the likely environmental trigger of the antibody response in FS is located in this endemic area, and that the population at risk to develop FS may also be at risk to develop an endemic form of pemphigus vulgaris as reported by our co-investigators from Brasilia.

The molecular portraits of breast tumors are conserved across microarray platforms.

BACKGROUND: Validation of a novel gene expression signature in independent data sets is a critical step in the development of a clinically useful test for cancer patient risk-stratification. However, validation is often unconvincing because the size of the test set is typically small. To overcome this problem we used publicly available breast cancer gene expression data sets and a novel approach to data fusion, in order to validate a new breast tumor intrinsic list. RESULTS: A 105-tumor training set containing 26 sample pairs was used to derive a new breast tumor intrinsic gene list. This intrinsic list contained 1300 genes and a proliferation signature that was not present in previous breast intrinsic gene sets. We tested this list as a survival predictor on a data set of 311 tumors compiled from three independent microarray studies that were fused into a single data set using Distance Weighted Discrimination. When the new intrinsic gene set was used to hierarchically cluster this combined test set, tumors were grouped into LumA, LumB, Basal-like, HER2+/ER-, and Normal Breast-like tumor subtypes that we demonstrated in previous datasets. These subtypes were associated with significant differences in Relapse-Free and Overall Survival. Multivariate Cox analysis of the combined test set showed that the intrinsic subtype classifications added significant prognostic information that was independent of standard clinical predictors. From the combined test set, we developed an objective and unchanging classifier based upon five intrinsic subtype mean expression profiles (i.e. centroids), which is designed for single sample predictions (SSP). The SSP approach was applied to two additional independent data sets and consistently predicted survival in both systemically treated and untreated patient groups. CONCLUSION: This study validates the "breast tumor intrinsic" subtype classification as an objective means of tumor classification that should be translated into a clinical assay for further retrospective and prospective validation. In addition, our method of combining existing data sets can be used to robustly validate the potential clinical value of any new gene expression profile.

A phase II trial of weekly paclitaxel and gemctiabine infused at a constant rate in patients with advanced non-small cell lung cancer.

BACKGROUND: Gemcitabine and paclitaxel both have significant single agent activity in non-small cell lung cancer (NSCLC). Because both are cell cycle and phase specific in their mechanism of action, frequent exposure should optimize activity. Phase I data support that gemcitabine is maximally converted to the active metabolite when it is infused at a rate of 10 mg/(m2 min). Based on this, we designed a phase II trial to examine gemcitabine 800 mg/m2 infused over 80 min with paclitaxel 110 mg/m2 infused over 3 h both on days 1, 8 and 15 every 28 days as first line therapy in patients with advanced NSCLC. The primary objectives were to assess the response rate, toxicity and survival of the combination in advanced NSCLC. Secondary objectives were to determine the effect of paclitaxel on the pharmacokinetic (PK) distribution of gemcitabine, the ability to achieve a concentration of 10-20 microM when gemcitabine was infused at a rate of 10 mg/(m2 min), as well as to assess the quality of life (QOL) with the functional assessment of cancer therapy-lung (FACT-L) questionnaire. METHODS: Patients with NSCLC, no prior treatment, ECOG performance status (PS) 0-1, adequate bone marrow, renal, and hepatic function were eligible for this trial. Paclitaxel 110 mg/m2 was infused over 3 h, followed by gemcitabine 800 mg/m2 infused over 80 min on days 1, 8, and 15 every 28 days for the first 2 patients, and then amended to days 1 and 8 every 21 days after the first 2 patients required day 15 dose omissions due to myelosupression. RESULTS: Thirty-nine patients were treated. Nine PS = 0; 28 PS = 1; Stage IIIB = 3, Stage IV = 36; median age 62 (range: 39-77). A median of six cycles (range: 0-10) was delivered. Grade 3-4 toxicities observed in > or =10% of patients included leucopenia in 26%, neutropenia in 28%, dyspnea in 13%, febrile neutropenia in 3% (1 patient). Fourteen of 39 (35%, 95% CI: 21-53%) patients had a partial response (PR), 14 of 39 (35%, 95% CI: 21-53%) had stable disease (SD) and 5 patients (13%, 95% CI: 4-27%) had progressive (PD). Median survival was 10.4 months (95% CI: 5.3-13.6). One-and two-year survival rates were 35% (95% CI: 21-53%) and 5% (95% CI: 0.6-17%), respectively. QOL as measured by the FACT-L and the trial outcome index (TOI) did not change significantly from baseline over the course of therapy. CONCLUSIONS: Paclitaxel and gemcitabine is an active and well-tolerated combination in advanced NSCLC. Patients on this trial had no significant change in their QOL as assessed by the FACT-L questionnaire.

Unanticipated hospital admissions during or soon after radiation therapy: Incidence and predictive factors.

PURPOSE: Unplanned hospital admissions in cancer patients undergoing treatment is an understudied area with important implications for both health care costs and patient outcomes. The goal of this retrospective study was to evaluate the rate, reasons for, and predictors of unplanned hospital admissions during or soon after palliative or curative radiation therapy for cancer, with or without chemotherapy. METHODS AND MATERIALS: A total of 1116 consecutive patients who received external beam radiation therapy for a malignancy at the University of North Carolina at Chapel Hill from January 1 through December 31, 2010, were studied. The primary outcome was unplanned hospitalization within 90 days of starting radiation therapy (ie, during or soon after). Multivariable logistic regression was used to examine patient and treatment factors associated with admissions. RESULTS: Twenty percent of patients experienced an unplanned admission, which was especially likely in patients with lung (25% of such patients admitted), head and neck (22%), and gastrointestinal (21%) cancers, as well as those treated with palliative intent (31%). The most common causes for admission were gastrointestinal symptoms, neurologic symptoms, respiratory symptoms, pain, and fever or infection. Forty-seven percent of admitted patients were seen in the clinic within 2 weeks of unplanned hospital admission, and 61% of those patients had a related complaint in the clinic. Multivariate analysis showed that married patients (odds ratio [OR] = 0.58; P < .001), curative intent (OR = 0.38; P < .001), and no concurrent chemotherapy (OR = 0.55; P < .001) were associated with decreased odds for admission. CONCLUSIONS: Unplanned admissions are relatively common during or soon after radiation therapy in our patient series. Additional work is needed to gather data from other centers and to better understand, and hopefully reduce, these unplanned admissions.

Anti-desmoglein-1 antibodies in onchocerciasis, leishmaniasis and Chagas disease suggest a possible etiological link to Fogo selvagem.

Pemphigus foliaceus (PF) and the endemic form Fogo Selvagem (FS) are mediated by pathogenic antibodies to the EC1-2 domains of desmoglein-1. There is a preclinical phase with antibodies to only EC5. Based on geographic clustering of cases, FS is thought to have an, as yet unidentified, environmental trigger. In this study we have searched for anti-desmoglein-1 antibodies in sera from parasitic (leishmaniasis, Chagas, and onchocerciasis), and infectious diseases (leprosy and South American (SA) blastomycosis), which are prevalent in the same geographic regions of Brazil as FS. A specific and sensitive desmoglein-1 ELISA detected antibodies in 34 of 41 onchocerciasis (83%), 38 of 88 leishmaniasis (43%), 18 of 31 Chagas disease (58%), 7 of 28 SA blastomycosis (25%), and 14 of 83 leprosy sera (17%). These sera recognized epitopes restricted to the EC5 domain. These findings identify several etiological factors for FS. It is hypothesized that a component of insect vector saliva, rather than the parasite itself may trigger an antibody response to EC-5. In persons with the known HLA susceptibility alleles and living in endemic areas, a response to the EC1-2 domains may subsequently develop by epitope spreading with associated clinical signs of FS.

ORTH: R and SAS software for regression models of correlated binary data based on orthogonalized residuals and alternating logistic regressions.

This article describes a new software for modeling correlated binary data based on orthogonalized residuals, a recently developed estimating equations approach that includes, as a special case, alternating logistic regressions. The software is flexible with respect to fitting in that the user can choose estimating equations for association models based on alternating logistic regressions or orthogonalized residuals, the latter choice providing a non-diagonal working covariance matrix for second moment parameters providing potentially greater efficiency. Regression diagnostics based on this method are also implemented in the software. The mathematical background is briefly reviewed and the software is applied to medical data sets.

The fidelity of the ligation step determines how ends are resolved during nonhomologous end joining.

Nonhomologous end joining (NHEJ) can effectively resolve chromosome breaks despite diverse end structures; however, it is unclear how the steps employed for resolution are determined. We sought to address this question by analysing cellular NHEJ of ends with systematically mispaired and damaged termini. We show NHEJ is uniquely proficient at bypassing subtle terminal mispairs and radiomimetic damage by direct ligation. Nevertheless, bypass ability varies widely, with increases in mispair severity gradually reducing bypass products from 85 to 6%. End-processing by nucleases and polymerases is increased to compensate, although paths with the fewest number of steps to generate a substrate suitable for ligation are favoured. Thus, both the frequency and nature of end processing are tailored to meet the needs of the ligation step. We propose a model where the ligase organizes all steps during NHEJ within the stable paired-end complex to limit end processing and associated errors.

Validation of interobserver agreement in lung cancer assessment: hematoxylin-eosin diagnostic reproducibility for non-small cell lung cancer: the 2004 World Health Organization classification and therapeutically relevant subsets.

CONTEXT: Precise subtype diagnosis of non-small cell lung carcinoma is increasingly relevant, based on the availability of subtype-specific therapies, such as bevacizumab and pemetrexed, and based on the subtype-specific prevalence of activating epidermal growth factor receptor mutations. OBJECTIVES: To establish a baseline measure of interobserver reproducibility for non-small cell lung carcinoma diagnoses with hematoxylin-eosin for the current 2004 World Health Organization classification, to estimate interobserver reproducibility for the therapeutically relevant squamous/nonsquamous subsets, and to examine characteristics that improve interobserver reproducibility. DESIGN: Primary, resected lung cancer specimens were converted to digital (virtual) slides. Based on a single hematoxylin-eosin virtual slide, pathologists were asked to assign a diagnosis using the 2004 World Health Organization classification. Kappa statistics were calculated for each pathologist-pair for each slide and were summarized by classification scheme, pulmonary pathology expertise, diagnostic confidence, and neoplastic grade. RESULTS: The 12 pulmonary pathology experts and the 12 community pathologists each independently diagnosed 48 to 96 single hematoxylin-eosin digital slides derived from 96 cases of non-small cell lung carcinoma resection. Overall agreement improved with simplification from the comprehensive 44 World Health Organization diagnoses (κ  =  0.25) to their 10 major header subtypes (κ  =  0.48) and improved again with simplification into the therapeutically relevant squamous/nonsquamous dichotomy (κ  =  0.55). Multivariate analysis showed that higher diagnostic agreement was associated with better differentiation, better slide quality, higher diagnostic confidence, similar years of pathology experience, and pulmonary pathology expertise. CONCLUSIONS: These data define the baseline diagnostic agreement for hematoxylin-eosin diagnosis of non-small cell lung carcinoma, allowing future studies to test for improved diagnostic agreement with reflex ancillary tests.