Efficient parameter estimation in longitudinal data analysis using a hybrid GEE method.

The method of generalized estimating equations (GEEs) provides consistent estimates of the regression parameters in a marginal regression model for longitudinal data, even when the working correlation model is misspecified (Liang and Zeger, 1986). However, the efficiency of a GEE estimate can be seriously affected by the choice of the working correlation model. This study addresses this problem by proposing a hybrid method that combines multiple GEEs based on different working correlation models, using the empirical likelihood method (Qin and Lawless, 1994). Analyses show that this hybrid method is more efficient than a GEE using a misspecified working correlation model. Furthermore, if one of the working correlation structures correctly models the within-subject correlations, then this hybrid method provides the most efficient parameter estimates. In simulations, the hybrid method's finite-sample performance is superior to a GEE under any of the commonly used working correlation models and is almost fully efficient in all scenarios studied. The hybrid method is illustrated using data from a longitudinal study of the respiratory infection rates in 275 Indonesian children.

Testing intergroup concordance in ranking experiments with two groups of judges.

Across many areas of psychology, concordance is commonly used to measure the (intragroup) agreement in ranking a number of items by a group of judges. Sometimes, however, the judges come from multiple groups, and in those situations, the interest is to measure the concordance between groups, under the assumption that there is some within-group concordance. In this investigation, existing methods are compared under a variety of scenarios. Permutation theory is used to calculate the error rates and the power of the methods. Missing data situations are also studied. The results indicate that the performance of the methods depend on (a) the number of items to be ranked, (b) the level of within-group agreement, and (c) the level of between-group agreement. Overall, using the actual ranks of the items gives better results than using the pairwise comparison of rankings. Missing data lead to loss in statistical power, and in some cases, the loss is substantial. The degree of power loss depends on the missing mechanism and the method of imputing the missing data, among other factors.

A semiparametric two-component "compound" mixture model and its application to estimating malaria attributable fractions.

Malaria remains a major epidemiologic problem in many developing countries. Malaria is defined as the presence of parasites and symptoms (usually fever) due to the parasites. In endemic areas, an individual may have symptoms attributable either to malaria or to other causes. From a clinical viewpoint, it is important to correctly diagnose an individual who has developed symptoms so that the appropriate treatments can be given. From an epidemiologic and economic viewpoint, it is important to determine the proportion of malaria-affected cases in individuals who have symptoms so that policies on intervention program can be developed. Once symptoms have developed in an individual, the diagnosis of malaria can be based on the analysis of the parasite levels in blood samples. However, even a blood test is not conclusive as in endemic areas many healthy individuals can have parasites in their blood slides. Therefore, data from this type of study can be viewed as coming from a mixture distribution, with the components corresponding to malaria and non-malaria cases. A unique feature in this type of data, however, is the fact that a proportion of the non-malaria cases have zero parasite levels. Therefore, one of the component distributions is itself a mixture distribution. In this article, we propose a semiparametric likelihood approach for estimating the proportion of clinical malaria using parasite-level data from a group of individuals with symptoms. Our approach assumes the density ratio for the parasite levels in clinical malaria and nonclinical malaria cases can be modeled using a logistic model. We use empirical likelihood to combine the zero and nonzero data. The maximum semiparametric likelihood estimate is more efficient than existing nonparametric estimates using only the frequencies of zero and nonzero data. On the other hand, it is more robust than a fully parametric maximum likelihood estimate that assumes a parametric model for the nonzero data. Simulation results show that the performance of the proposed method is satisfactory. The proposed method is used to analyze data from a malaria survey carried out in Tanzania.

Adrenocortical adenoma and carcinoma: histopathological and molecular comparative analysis.

We compared histomorphological features and molecular expression profiles of adrenocortical adenomas (ACAd) and carcinomas (ACCa). A critical histopathological review (mean, 11 slides per patient) was conducted of 37 ACAd and 67 ACCa. Paraffin-embedded tissue cores of ACAd (n = 33) and ACCa (n = 38) were arrayed in triplicate on tissue microarrays. Expression profiles of p53, mdm-2, p21, Bcl-2, cyclin D1, p27, and Ki-67 were investigated by immunohistochemistry and correlated with histopathology and patient outcome using standard statistical methodology. Median follow-up period was 5 years. Tumor necrosis, atypical mitoses, and >1 mitosis per 50 high-power fields were factors that were highly specific for ACCa (P <.001). Number (0 to 4) of unfavorable markers [Ki-67 (+), p21 (+), p27 (+), mdm-2(-)] expressed was significantly associated with mitotic activity and morphologic index (i.e., number of adverse morphologic features) and highly predictive of malignancy (P <.001). Ki-67 overexpression occurred in 0 ACAd and 36% ACCa (P <.001) and was significantly associated with mitotic rate and unfavorable morphologic index (P <.001). Tumor necrosis, atypical mitoses, >5 mitoses per 50 high-power fields, sinusoidal invasion, histologic index of >5, and presence of more than two unfavorable molecular markers were associated significantly with metastasis in ACCa. Well-established histopathologic criteria and Ki-67 can specifically distinguish ACCAd from ACCa. Tumor cell proliferation (Ki-67) correlates with mitotic activity and morphologic index. Tumor morphology is a better predictor of metastatic risk in ACCa than current immunohistochemistry-detected cell cycle regulatory and proliferation-associated proteins.

Tissue microarray molecular profiling of early, node-negative adenocarcinoma of the rectum: a comprehensive analysis.

PURPOSE: Early-stage adenocarcinoma of the rectum treated with curative intent has a favorable overall prognosis; however, 20%-30% of the patients recur, and the majority ultimately die of disease. Recurrence and tumor-related mortality may be attributable to molecular abnormalities in primary tumors accounting for their more aggressive biological behavior. This study evaluates such molecular phenotypes with regard to cell cycle regulation and proliferation and determines their significance for patient outcome. EXPERIMENTAL DESIGN: One hundred patients with primary T(2-3), N(0) adenocarcinoma of the rectum uniformly treated by surgery alone were studied. Core biopsies of pathological specimens were assembled on tissue microarrays, and expression of p53, mdm-2, p21, Bcl-2, p27, cyclin D1, and Ki-67 was analyzed by immunohistochemistry. Molecular profiles were correlated with disease-free (DFS) and disease-specific survival (DSS). RESULTS: Despite previously described prognostic relevance of some of the investigated molecules in analyses where different stages of colorectal cancer were included, none of the cell cycle-regulatory or proliferation-related markers was associated with recurrence or survival. However, patients with tumors demonstrating down-regulation of p27, a cyclin-dependent kinase inhibitor and tumor suppressor gene associated with development of metastases, showed a trend toward reduced DFS and DSS (P = 0.06 and P = 0.07, respectively). CONCLUSIONS: In this homogeneous group of patients with early-stage, node-negative adenocarcinoma of the rectum uniformly treated by surgery alone, the investigated cell cycle-regulatory and proliferation-associated proteins appear to have no prognostic significance. However, down-regulation of p27 appears to be associated with a trend toward reduced DFS and DSS, which suggests further investigation of other p27-related pathways potentially relevant for metastatic disease.

Primary adult soft tissue sarcoma: time-dependent influence of prognostic variables.

PURPOSE: To define prognostic factors for postrelapse survival and their time-dependent influence for adult soft tissue sarcoma (STS). PATIENTS AND METHODS: We analyzed 2,123 patients with completely resected localized primary STS treated from 1982 to 1999. Variables studied included tumor site, size, depth, grade, and resection margin but not treatment other than resection. Landmark time frames were used to assess the influence of disease-free interval (DFI) on disease-specific survival (DSS). DSS was estimated with the Kaplan-Meier method. Univariate and multivariate analyses were performed using log-rank test and the Cox proportional hazards regression model. Time-dependent stepwise regression analysis evaluated the time-dependent influence of each variable. RESULTS: Two thirds of recurrences developed within 2 years of initial resection. Tumor size (P <.001), grade (P <.001), and microscopic resection margin (P <.001) independently predicted DSS for all STS. Size and grade independently predicted early (DFI 3 years) DSS. Risk of tumor-related death was the same across all sites 3 years postresection and decreased significantly for extremity/trunk STS when DFI exceeded 3 years (P <.001). Influence of initial high-risk factors for tumor-related mortality in extremity/trunk STS decreased by 40% 3 years postresection, but their influence over DSS for non-extremity/trunk sites remained constant over time. Likelihood of complete resection after recurrence (all sites) increased with DFI (9% and 33% for DFI < 6 and > 36 months, respectively). CONCLUSION: Tumor size, grade, and resection margin predict outcome for completely resected STS, and their influence for DSS is time- and site-dependent. The influence of prognostic factors changes over the natural history of extremity/trunk STS. Time to recurrence exerts significant influence over complete resection rates for recurrent disease.

Hürthle cell carcinoma: a 60-year experience.

BACKGROUND: The aim of this study was to define the clinical behavior and prognostic indicators of outcome in Hürthle cell cancer (HCC). METHODS: Diagnosis was confirmed for 56 patients with HCC treated between 1940 and 2000, who form the basis of this study. Primary end points were relapse-free survival (RFS) and disease-specific survival (DSS). Data were analyzed with the Kaplan-Meier method and by log-rank test. RESULTS: The extent of thyroid resection did not predict outcome. Recurrence was a significant predictor of tumor-related mortality. Significant adverse predictors of RFS and DSS were degree of invasion, size >4 cm, extrathyroidal extension, and initial nodal or distant metastases. The most significant predictor of outcome was extent of invasion. Eight-year RFS values for low- and high-risk groups were 100% and 24%. Corresponding rates of 8-year DSS were 100% and 58%. CONCLUSIONS: Widely invasive HCC is an aggressive malignancy that identifies patients who are at high risk for recurrence and tumor-related death. Patients with HCC have a prognosis that is reliably predicted by degree of invasion, tumor size, extrathyroidal disease extension, and initial nodal or distant metastasis. Recurrence portends a poor outcome. High-risk patients and those with recurrence should be considered for adjuvant therapy.

The value of splenic preservation with distal pancreatectomy.

HYPOTHESIS: Splenic-preserving distal pancreatectomy for benign or low-grade malignant disease is associated with decreased perioperative morbidity compared with conventional distal pancreatectomy with splenectomy. DESIGN: A retrospective review of a prospective database of patients. SETTING: Memorial Sloan-Kettering Cancer Center, New York, NY. PATIENTS: All patients (N = 211) undergoing distal pancreatectomy. MAIN OUTCOME MEASURES: Perioperative complications, length of postoperative stay, and overall survival times were analyzed. RESULTS: After excluding patients with adenocarcinoma and those who had other major organ resection, 125 patients underwent distal pancreatectomy for benign or low-grade malignant disease with splenectomy (n = 79) or splenic preservation (n = 46). Perioperative complications occurred in 39 (49%) of the 79 patients following splenectomy and 18 (39%) of the 46 patients following splenic preservation (P =.21). Perioperative infectious complications and severe complications were significantly higher in the splenectomy group (28% and 11%) compared with the splenic preservation group (9% and 2%) (P =.01 and.05), respectively. Length of hospital stay was 9 days (range, 5-41 days) following splenectomy and 7 days (range, 5-26 days) following splenic preservation (P<.01). No difference in length of surgery, units of blood transfused, or perioperative mortality was noted between groups. CONCLUSIONS: Splenic preservation following distal pancreatectomy for benign or low-grade malignant disease is safe and is associated with a reduction in perioperative infectious complications, severe complications, and length of hospital stay compared with conventional distal pancreatectomy with splenectomy. Therefore, splenic preservation should be considered in this group of patients.

Analysis of the prognostic significance of microscopic margins in 2,084 localized primary adult soft tissue sarcomas.

OBJECTIVE: To define the significance of positive microscopic resection margins in a large cohort treated for soft tissue sarcoma. METHODS: The authors analyzed 2,084 patients with localized primary soft tissue sarcoma (all anatomic sites) treated from 1982 to 2000. Clinicopathologic variables studied included tumor site, size, depth, histologic type, grade, and resection margin status. Treatment other than resection was not analyzed. Study endpoints included local and distant recurrence-free and disease-specific survival rates, estimated by the Kaplan-Meier method. Univariate and multivariate analyses were performed using the log-rank test and the Cox proportional hazards model. RESULTS: Median follow-up was 50 months. After primary resection, 1,624 (78%) patients had negative and 460 (22%) had positive resection margins. Having positive margins nearly doubled the risk of local recurrence and increased the risk of distant recurrence and disease-related death. Seventy-two percent of patients with positive margins had no recurrence. Resection margin did not predict local control for retroperitoneal sarcomas or fibrosarcomas. Resection margin remained significantly associated with distant recurrence-free survival and disease-specific survival across all subsets after adjusting for other prognostic variables. The overall 5-year disease-specific survival rates for negative and positive margins were 83% and 75%. CONCLUSIONS: Positive microscopic resection margins significantly decrease the local recurrence-free survival rate for other-than-primary fibrosarcoma and retroperitoneal sarcomas, and independently predict distant recurrence-free survival rates and disease-specific survival rates for all patient subsets. Adjuvant therapy should be considered in the management of soft tissue sarcoma to increase local control. Because 72% of positive margins did not equate with inevitable local recurrence, considerable clinical judgment is required in considering additional treatment. Microscopic resection margins should be considered for inclusion in staging systems and treatment algorithms that address local recurrence.

Postoperative nomogram for 12-year sarcoma-specific death.

PURPOSE: Few published studies have analyzed risk factors for sarcoma-specific death. We developed and internally validated a nomogram that combines the factors to predict the probability of 12-year sarcoma-specific death using a database of 2,136 prospectively followed adult patients treated at a single institution. PATIENTS AND METHODS: Nomogram predictor variables included age at diagnosis, tumor size (< or = 5, 5 to 10, or > 10 cm), histologic grade (high or low), histologic subtype (fibrosarcoma, leiomyosarcoma, liposarcoma, malignant fibrous histiocytoma, malignant peripheral nerve tumor, synovial, or other), depth (superficial or deep), and site (upper extremity, lower extremity, visceral, thoracic or trunk, retrointraabdominal, or head or neck). Death from sarcoma or treatment complication was the predicted end point. Three prediction methods were compared, Kaplan-Meier analysis of all possible subsets, recursive partitioning, and Cox proportional hazards regression analysis. The concordance index was used as an accuracy measure with bootstrapping to correct for optimistic bias. RESULTS: Sarcoma-specific death at 12 years was 36% (95% confidence interval, 33% to 39%). The bootstrap-corrected concordance indices were as follows: Kaplan-Meier, 0.69; recursive partitioning, 0.74; and Cox regression, 0.77. A nomogram was drawn on the basis of the Cox regression model. This nomogram was internally validated using bootstrapping and shown to have excellent calibration. CONCLUSION: A nomogram has been developed to predict 12-year sarcoma-specific death. This tool may be useful for patient counseling, follow-up scheduling, and clinical trial eligibility determination.

Clinical significance of molecular expression profiles of Hürthle cell tumors of the thyroid gland analyzed via tissue microarrays.

Hürthle cell tumors are rare thyroid neoplasms for which disease biology is poorly understood and diagnosis of carcinoma can be challenging. The aim of the study was to characterize molecular expression profiles of Hürthle cell tumors and to determine the clinical significance of identified phenotypes. Paraffin-embedded tissue cores of normal thyroid (n = 18), and histopathologically well-defined Hürthle cell adenomas (n = 27), Hürthle cell tumors of unknown malignant behavior (n = 7), and minimally (n = 14) and widely (n = 21) invasive Hürthle cell carcinomas were arrayed in triplicate on tissue microarrays. Expression profiles of p53, mdm-2, p21, Bcl-2, cyclin D1, and Ki-67 were detected by immunohistochemistry and correlated with clinicopathological data and patient outcome using standard statistical methodology. Median follow-up time was 8 years. High Ki-67 proliferative index was evident only in the clinically aggressive widely invasive Hürthle cell carcinomas and was associated with significantly reduced relapse-free (P = 0.001) and disease-specific (P < 0.001) survival. The molecular phenotype of Hürthle cell tumors, independent of histopathological subtype diagnosis, was characterized by p53(-), mdm-2(+), p21(+/-), cyclin D1(-), and Bcl-2(+/-). Normal thyroid tissue demonstrated a p53(-), mdm-2(-), p21(-), cyclin D1(-), and Bcl-2(+) phenotype. The Bcl-2(+) phenotype was associated with improved relapse-free survival (P = 0.04) and disease-specific survival (P = 0.01) in widely invasive carcinomas and the Ki-67(+)/Bcl-2(-) phenotype was associated with the diagnosis of widely invasive Hürthle cell carcinoma (P < 0.001). This study demonstrates that tissue microarray-based profiling allows identification of molecular markers that are associated with patient prognosis. High Ki-67 proliferative index was associated with adverse outcome in Hürthle cell neoplasms. Together with down-regulation of Bcl-2, high Ki-67 proliferative index may be useful for diagnosing widely invasive Hürthle cell carcinomas. Molecular alterations in the p53 pathway play a role in Hürthle cell tumorigenesis, but other unidentified molecular changes seem to be required to induce the malignant phenotype.

Older age is the strongest predictor of postoperative atrial fibrillation.

BACKGROUND: Identification of patients vulnerable for atrial fibrillation (AF) after major thoracic surgery will allow targeting those most likely to benefit from prophylactic therapy. The goal of the current study was to evaluate the accuracy of easily available clinical characteristics for the prediction of this complication. METHODS: Patients undergoing major elective thoracic surgery were chosen from an ongoing prospective database. RESULTS: Postoperative in-hospital AF occurred in 79 (15%) of 527 patients Using cut-point methodology and logistic regression, the authors identified two preoperative risk factors independently associated with AF: age 60 yr or older (P = 0.0007) and heart rate 74 beats/min or greater on preadmission electrocardiogram (P = 0.005). The odds of developing AF increased by a factor of 2.5 (95% confidence interval, 1.7-3.4; P < 0.0001) between incremental age categories (< 60 yr, 60-69 yr, > or = 70 yr) and by a factor of 2.3 (95% confidence interval, 1.4-3.8; P < 0.0007) between heart rate categories (< 74 beats/min, > or = 74 beats/min). The combination of age 60 yr or older and preoperative heart rate 74 beats/min or greater predicted AF with a sensitivity of 73% and specificity of 57%. Maximum P-wave duration as measured from standard electrocardiogram did not differentiate patients who did or did not develop AF. Patients who developed AF had a higher incidence of postoperative pneumonia (14 vs. 4%; P = 0.001), acute respiratory failure (8 vs. 1.6%; P = 0.01), greater hospital stay (17 +/- 17 vs. 10 +/- 9 days; P = 0.001) and 30-day mortality (11 vs. 3%; P = 0.001) when compared with those who did not develop AF, respectively. CONCLUSIONS: Advanced age and preoperative heart rate identify patients at high risk for development of AF after thoracic surgery. Postoperative AF occurs more frequently in patients with greater postoperative morbidity and length of hospitalization.