Integrating additional factors into the TNM staging for cutaneous melanoma by machine learning.

BACKGROUND: Integrating additional factors into the TNM staging system is needed for more accurate risk classification and survival prediction for patients with cutaneous melanoma. In the present study, we introduce machine learning as a novel tool that incorporates additional prognostic factors to improve the current TNM staging system. METHODS AND FINDINGS: Cancer-specific survival data for cutaneous melanoma with at least a 5 years follow-up were extracted from the Surveillance, Epidemiology, and End Results Program of the National Cancer Institute and split into the training set (40,781 cases) and validation set (5,390 cases). Five factors were studied: the primary tumor (T), regional lymph nodes (N), distant metastasis (M), age (A), and sex (S). The Ensemble Algorithm for Clustering Cancer Data (EACCD) was applied to the training set to generate prognostic groups. Utilizing only T, N, and M, a basic prognostic system was built where patients were stratified into 10 prognostic groups with well-separated survival curves, similar to 10 AJCC stages. These 10 groups had a significantly higher accuracy in survival prediction than 10 stages (C-index = 0.7682 vs 0.7643; increase in C-index = 0.0039, 95% CI = (0.0032, 0.0047); p-value = 7.2×10-23). Nevertheless, a positive association remained between the EACCD grouping and the AJCC staging (Spearman's rank correlation coefficient = 0.8316; p-value = 4.5×10-13). With additional information from A and S, a more advanced prognostic system was established using the training data that stratified patients into 10 groups and further improved the prediction accuracy (C-index = 0.7865 vs 0.7643; increase in C-index = 0.0222, 95% CI = (0.0191, 0.0254); p-value = 8.8×10-43). Both internal validation using the training set and temporal validation using the validation set showed good stratification and a high predictive accuracy of the prognostic systems. CONCLUSIONS: The EACCD allows additional factors to be integrated into the TNM to create a prognostic system that improves patient stratification and survival prediction for cutaneous melanoma. This integration separates favorable from unfavorable clinical outcomes for patients and improves both cohort selection for clinical trials and treatment management.

A prognostic system for epithelial ovarian carcinomas using machine learning.

INTRODUCTION: Integrating additional factors into the International Federation of Gynecology and Obstetrics (FIGO) staging system is needed for accurate patient classification and survival prediction. In this study, we tested machine learning as a novel tool for incorporating additional prognostic parameters into the conventional FIGO staging system for stratifying patients with epithelial ovarian carcinomas and evaluating their survival. MATERIAL AND METHODS: Cancer-specific survival data for epithelial ovarian carcinomas were extracted from the Surveillance, Epidemiology, and End Results (SEER) program. Two datasets were constructed based upon the year of diagnosis. Dataset 1 (39 514 cases) was limited to primary tumor (T), regional lymph nodes (N) and distant metastasis (M). Dataset 2 (25 291 cases) included additional parameters of age at diagnosis (A) and histologic type and grade (H). The Ensemble Algorithm for Clustering Cancer Data (EACCD) was applied to generate prognostic groups with depiction in dendrograms. C-indices provided dendrogram cutoffs and comparisons of prediction accuracy. RESULTS: Dataset 1 was stratified into nine epithelial ovarian carcinoma prognostic groups, contrasting with 10 groups from FIGO methodology. The EACCD grouping had a slightly higher accuracy in survival prediction than FIGO staging (C-index = 0.7391 vs 0.7371, increase in C-index = 0.0020, 95% confidence interval [CI] 0.0012-0.0027, p = 1.8 × 10-7 ). Nevertheless, there remained a strong inter-system association between EACCD and FIGO (rank correlation = 0.9480, p = 6.1 × 10-15 ). Analysis of Dataset 2 demonstrated that A and H could be smoothly integrated with the T, N and M criteria. Survival data were stratified into nine prognostic groups with an even higher prediction accuracy (C-index = 0.7605) than when using only T, N and M. CONCLUSIONS: EACCD was successfully applied to integrate A and H with T, N and M for stratification and survival prediction of epithelial ovarian carcinoma patients. Additional factors could be advantageously incorporated to test the prognostic impact of emerging diagnostic or therapeutic advances.

Carcinomas of the Renal Pelvis, Ureters, and Urinary Bladder Share a Carcinogenic Field as Revealed in Epidemiological Analysis of Tumor Registry Data.

BACKGROUND: Urothelial carcinomas are the most common malignant tumors in the upper and lower urinary tract. Renal cell carcinomas (RCCs) have a different pathoepidemiologic incidence and characteristics. We describe a population-based approach of differentiating between urothelial and renal carcinomas as a basis to support shared morphologic phenotypes. MATERIALS AND METHODS: Data from 2000 through 2014 from the Surveillance, Epidemiology, and End Results Program of the National Cancer Institute were used to calculate the incidence rates for cancers of the kidney, renal pelvis, ureter, and urinary bladder. Graphic plots of the epidemiologic patterns were analyzed according to age frequency density and double logarithmic (log-log) plots of age-specific incidence rates and age of diagnosis. RESULTS: RCCs were initially more common than cancers of the urinary bladder, but after age 60, cancers of the bladder became more common with age-specific rates rapidly rising in all age groups. The age frequency density plot for RCC peaked earlier than for urothelial cancers indicating a different tumorigenic process. Log-log plots revealed near parallel proportional rate patterns for cancers of the renal pelvis, ureters, and urinary bladder, suggesting similar carcinogenic pathways among these tumors, whereas they were not parallel for RCCs. Similar slopes indicate that cancer incidence is increasing at similar rates regardless of the incidence of each cancer. CONCLUSION: Tumors that arise in the renal pelvis, ureters, and urinary bladder share a common carcinogenic field on the basis of pathoepidemiologic analysis. The definition of a carcinogenic field should expand to include epidemiological parameters as well as common morphologic and embryological patterns.

Creating prognostic systems for cancer patients: A demonstration using breast cancer.

Integrating additional prognostic factors into the tumor, lymph node, metastasis staging system improves the relative stratification of cancer patients and enhances the accuracy in planning their treatment options and predicting clinical outcomes. We describe a novel approach to build prognostic systems for cancer patients that can admit any number of prognostic factors. In the approach, an unsupervised learning algorithm was used to create dendrograms and the C-index was used to cut dendrograms to generate prognostic groups. Breast cancer data from the Surveillance, Epidemiology, and End Results Program of the National Cancer Institute were used for demonstration. Two relative prognostic systems were created for breast cancer. One system (7 prognostic groups with C-index = 0.7295) was based on tumor size, regional lymph nodes, and no distant metastasis. The other system (7 prognostic groups with C-index = 0.7458) was based on tumor size, regional lymph nodes, no distant metastasis, grade, estrogen receptor, progesterone receptor, and age. The dendrograms showed a relationship between survival and prognostic factors. The proposed approach is able to create prognostic systems that have a good accuracy in survival prediction and provide a manageable number of prognostic groups. The prognostic systems have the potential to permit a thorough database analysis of all information relevant to decision-making in patient management and prognosis.

The anatomy of the TNM for colon cancer.

BACKGROUND: To visualize the anatomy as revealed by dendrograms of the tumor, lymph node, and metastasis (TNM) staging system for colon cancer and compare it with the Dukes' system. METHODS: A hierarchical clustering algorithm generated tree-structured dendrograms that stratified patients according to survival only. The dendrograms were constructed with the same prognostic variables used for the TNM. Because combinations of prognostic factors were stratified only on survival, additional factors of any number and type could be integrated into the TNM without changing the TNM categories. RESULTS: The algorithm provided a step-by-step visualization of the TNM and the Dukes' system for colon cancer. Dendrograms and associated 5-year survival rates were generated for the T category only, the N category only, the T, N combination, and combinations of the T, N, and M, and the T, N, M with histological grade. Dendrograms revealed visual differences between the structure of TNM and the Dukes' system of staging. Dendrograms also revealed how variations in prognostic factors changed survival. By cutting dendrograms along their dissimilarity axis, multiple prognostic subgroups could be created for colon cancer that may reflect outcomes that are more accurate to estimate. CONCLUSIONS: Dendrograms provide a new way to view cancer patient staging. They reveal a visual step-by-step hierarchical relationship between survival rates and combinations of prognostic variables. The dendrograms also revealed fundamental differences between the TNM and the Dukes system of staging. By stratifying on survival only, additional factors including molecular factors can be added to the TNM, because it classifies patients according to survival rates only and not according to pre-set rules of prognostic factors and stage groups. The clinical implications of stratifying only survival are discussed.

An Algorithm for Creating Prognostic Systems for Cancer.

The TNM staging system is universally used for classification of cancer. This system is limited since it uses only three factors (tumor size, extent of spread to lymph nodes, and status of distant metastasis) to generate stage groups. To provide a more accurate description of cancer and thus better patient care, additional factors or variables should be used to classify cancer. In this paper we propose a hierarchical clustering algorithm to develop prognostic systems that classify cancer according to multiple prognostic factors. This algorithm has many potential applications in augmenting the data currently obtained in a staging system by allowing more prognostic factors to be incorporated. The algorithm clusters combinations of prognostic factors that are formed using categories of factors. The dissimilarity between two combinations is determined by the area between two corresponding survival curves. Groups from cutting the dendrogram and survival curves of the individual groups define our prognostic systems that classify patients using survival outcomes. A demonstration of the proposed algorithm is given for patients with breast cancer from the Surveillance, Epidemiology, and End Results (SEER) Program of the National Cancer Institute.

An algorithm for expanding the TNM staging system.

AIM: We describe a new method to expand the tumor, lymph node, metastasis (TNM) staging system using a clustering algorithm. Cases of breast cancer were used for demonstration. MATERIALS & METHODS: An unsupervised ensemble-learning algorithm was used to create dendrograms. Cutting the dendrograms produced prognostic systems. RESULTS: Prognostic systems contained groups of patients with similar outcomes. The prognostic systems based on tumor size and lymph node status recapitulated the general structure of the TNM for breast cancer. The prognostic systems based on tumor size, lymph node status, histologic grade and estrogen receptor status revealed a more detailed stratification of patients when grade and estrogen receptor status were added. CONCLUSION: Prognostic systems from cutting the dendrogram have the potential to improve and expand the TNM.

Cystadenomas of the liver and extrahepatic bile ducts: Morphologic and immunohistochemical characterization of the biliary and intestinal variants.

Cystadenomas of the liver and extrahepatic bile ducts (EHBD) are uncommon but distinctive neoplasms whose terminology and epithelial phenotype have been a source of controversy. We reviewed 20 cases, 16 arising in the liver and 4 in the EHBD. Eighteen patients were women, with a mean age of 36.5 years. Eighteen tumors were multiloculated and 2 were unilocular. The tumor size ranged from 4 to 29 cm (average, 11 cm). The cyst fluid in 13 tumors was described as serous, in 2 as clear, in 2 others as hemorrhagic, and in 1 as serous and mucinous. Only in 2 tumors was the fluid described as mucinous. In 18 cystadenomas, the predominant epithelial lining consisted of a single layer of cuboidal or low-columnar nondysplastic cells similar to those of the gallbladder or bile ducts. This epithelial lining was strongly positive for cytokeratins 7 and 19, and focally positive for MUC1. Only 2 cystadenomas showed predominant intestinal differentiation characterized by mature goblet cells and columnar absorptive cells. These cells expressed CDX2, MUC2, and cytokeratin 20. Admixed with the goblet and columnar cells, there were serotonin-containing cells and Paneth cells. These 2 tumors showed extensive areas of high-grade dysplasia and invasive adenocarcinoma with intestinal phenotype. A subepithelial ovarian-like stroma was present in all tumors. None of the patients died of the tumors. We believe that the term mucinous cystic tumor recommended by the World Health Organization for all cystadenomas of the liver and EHBD is a misnomer.

Association of urothelial carcinoma of the renal pelvis with papillary and medullary thyroid carcinomas. A new sporadic neoplastic syndrome?

We describe 2 adult women (72 and 54 years), 1 with a low-grade noninvasive papillary urothelial carcinoma of the renal pelvis, who 14 years later developed a papillary carcinoma in 1 thyroid lobe and a medullary carcinoma in the contralateral lobe. Both neoplasms were similar in size and appeared symmetrical. Despite its small size, the medullary carcinoma metastasized in multiple cervical lymph nodes. The second patient had a high-grade invasive papillary urothelial carcinoma of the renal pelvis that infiltrated the renal parenchyma and metastasized in one of the lungs. Five months later, a papillary carcinoma was discovered in the thyroid gland. The 2 papillary thyroid carcinomas were of the follicular variant. Adjacent to 1 papillary carcinoma, there was a dominant nodule of a colloid and adenomatous goiter. The medullary carcinoma contained stromal amyloid and was immunoreactive for calcitonin and carcinoembryonic antigen. There was no C-cell hyperplasia (medullary carcinoma in situ). The 2 patients are alive, 1 is living with pulmonary metastasis from the high-grade urothelial carcinoma. Twelve cases of this neoplastic association were registered in the Survey, Epidemiology, and End Results Program from 1980 to 2009. We believe that the combination of these unusual neoplasms in the same patient may represent a new sporadic neoplastic syndrome.

Endometrial stromal sarcomas: immunoprofile with emphasis on HMB45 reactivity.

OBJECTIVES: We describe the morphologic and immunohistochemical features of 17 endometrial stromal neoplasms, 16 sarcomas, and one stromal nodule. METHODS: We reviewed 35 cases interpreted as endometrial stromal neoplasms, but 17 high-grade endometrial stromal sarcomas (ESS) and one case of mixed endometrial sarcoma and leiomyosarcoma were excluded from the study. Data from the Surveillance Epidemiology and End Results program on low- and high-grade ESS for 1973 through 2003 were obtained. RESULTS: One uterine primary ESS had collections of clear cells (20%), while a metastatic ESS contained predominantly clear cells (90%). CD10 (88.2%) and smooth muscle actin (70.5%) were the most common positive immunohistochemical markers. The latter marker was located in the cytoplasm in 47% of the ESS and in the nucleus in 23.5%, a previously unreported feature. HMB45 was detected in 23.5% of the ESS, which contrasts with the 2% reported by other authors. CONCLUSIONS: The presence of clear cells and HMB45 reactivity does not justify the term perivascular epithelioid cell tumors for these neoplasms. Two of 17 patients with ESS died of metastatic disease. However, among 274 cases of ESS (all stages included) collected by the Surveillance Epidemiology and End Results Program of the National Cancer Institute during a 30-year period, the 10-year survival rate was 94%.

Rokitansky-Aschoff sinuses mimicking adenocarcinoma of the gallbladder: a study of 8 cases.

We describe 8 cases of cholecystectomy specimens (6 laparoscopic and 2 open cholecystectomies) with Rokitansky-Aschoff (R-A) sinuses that were misinterpreted as adenocarcinomas. They were compared with 8 examples of classical R-A sinuses and 6 cases of R-A sinuses containing foci of adenocarcinoma. Five cases misinterpreted as adenocarcinomas consisted of densely packed, closely opposed R-A sinuses with little intervening stroma or surrounded by a desmoplastic stroma. They were lined by a single layer of cuboidal or columnar cells. There were also pseudostratified columnar cells with mucin-containing cytoplasm and hyperchromatic or vesicular nuclei but without mitotic figures. In 2 cases, the columnar cells had subnuclear vacuoles. Small papillary projections into R-A sinuses were seen in 4 cases, and in 3 others collections of metaplastic pyloric glands, some connected to the epithelium of the sinuses, were recognized. There was focal reactive atypia in both the epithelium of the surface and that of the sinuses. The R-A sinuses resembling gland-like structures had a laminar distribution rather than a disorderly haphazard distribution seen in well-differentiated adenocarcinoma. The remaining 3 cases misinterpreted as adenocarcinomas consisted of numerous deeply penetrating long and short R-A sinuses that branched in different directions and which reach the subserosal or perimuscular connective tissue mimicking invasion. The sinuses were surrounded by hyperplastic smooth muscle bundles and lined by pseudostratified columnar cells mixed with a few goblet cells showing reactive atypia and no mitotic figures. There was focal reactive atypia in both the epithelium of the surface and that of the sinuses. The 2 types of R-A sinuses did not label with carcinoembryonic antigen or p53 and had very low proliferative activity as measured by the MIB1-labeling index. All patients are alive and disease free from 8 months to 17 years (mean follow-up 7 y). In contrast, the foci of invasive adenocarcinoma that arose in R-A sinuses consisted of glands lined by atypical cuboidal or columnar cells with loss of polarity, large hyperchromatic or vesicular nuclei, prominent nucleoli, and mitotic figures, quite different from the cells lining the R-A sinuses. Because of increasing number of laparoscopic cholecystectomies performed annually in the United States, pathologists should become familiar with these gallbladder lesions that are usually incidental findings but can simulate malignant epithelial neoplasms.

Nonmucinous cystadenomas of the pancreas with pancreatobiliary phenotype and ovarian-like stroma.

Among 31 benign cystic neoplasms of the pancreas diagnosed as mucinous cystadenomas, we identified 9 (29%) cases of nonmucinous cystadenomas with a pancreatobiliary phenotype and an ovarian-like stroma. Although both cystic tumors belong to the same family, they should be separated because their epithelial lining and cyst fluid are different. The lining cells of the nonmucinous cystadenomas consisted of a single layer of cuboidal cells, similar to the epithelial cells of the normal pancreatic ducts, and were not dysplastic (90%-100% of the lining cells). The cyst fluid was described as serous or clear. The remaining 22 classical mucinous cystadenomas, lined predominantly by mucinous and foveolar epithelium, revealed focal pancreatobiliary epithelium in 86% of the cases, and 6 pancreatic invasive mucinous cystadenocarcinomas failed to show pancreatobiliary differentiation. We believe that these nonmucinous cystadenomas of the pancreas represent a distinctive subset of cystic neoplasms of the pancreas that probably have no malignant potential.

Mucinous cystadenoma of the pyelocaliceal system: a report of 3 examples and an analysis of 17 previously published cases.

We report 3 patients all men between 45 and 64 years of age with unilocular or multilocular mucinous cystadenomas of the kidney. One tumor arose from the renal pelvis, and 2 involved the entire pyelocaliceal system. The tumors measured between 2.4 and 37 cm in greatest dimension. Two patients were asymptomatic, and 1 had recurrent attack of acute pyelonephritis. Microscopically, the morphology and immunophenotype (CK20, MUC2, and CDX2 positive) of the tumors were similar to the colonic adenomas. Two patients were asymptomatic 24 and 64 months after surgery, including the patient with mucinous cystadenoma and intramucosal carcinoma. One patient died of acute myocardial infarction, and his tumor was an autopsy finding. Only 17 cases of mucinous cystadenomas and 5 cases of mucinous cystadenocarcinomas have been reported. Of the 17 mucinous cystadenomas, 2 arose in horseshoe kidneys. The mean size of these neoplasms was 15 cm (2.4-37 cm). Despite their large size, some patients with mucinous cystadenomas were asymptomatic. Sixty percent were associated with renal lithiasis. Thirty percent progressed to mucinous adenocarcinomas, and only 2 cases showed areas of intramucosal carcinomas. Two cases were associated with carcinoid tumors, similar to those reported in the appendix. Most patients were asymptomatic after surgery, and only 1 patient died by abdominal sepsis related to adenomucinosis. The 3 examples of mucinous cystadenomas of the pyelocaliceal system reported here, and those previously published indicate that they are very uncommon neoplasms with morphology and intestinal immunophenotype similar to the colonic adenomas.

Thyroid-like cholangiocarcinoma of the liver: an unusual morphologic variant with follicular, trabecular and insular patterns.

We report the case of a 26-year-old woman with a 19 cm malignant hepatic neoplasm with morphological features that closely resembled a follicular thyroid carcinoma. Despite this, it was interpreted as a cholangiocarcinoma due to the absence of a primary thyroid tumor and the lack of thyroglobulin and TTF-1 immunoreactivity by the hepatic tumor. The left hepatic lobectomy specimen showed an encapsulated and multinodular gray-white mass with cystic and hemorrhagic areas. Microscopically, it displayed predominant macro and microfolicullar patterns with focal solid, trabecular and insular areas. The small and distended follicles contained a colloid-like secretion and were lined by low cuboidal cells with scant cytoplasm, round or oval hyperchromatic nuclei with fine chromatin. The solid areas, trabecular and insular structures were similar to those of follicular or papillary thyroid carcinomas. In addition, some of the neoplastic cells had clear nuclei with occasional grooves. The tumor was positive for cytokeratin (CK) 7, CK 19 and CD138, and negative for TTF-1, thyroglobulin, Hepar-1, Glypican-3, alpha-fetoprotein and neuroendocrine markers. A thyroid neoplasm was excluded clinically and by ultrasound and computed tomography. Although, the residual hepatic parenchyma was initially not cirrhotic, the patient eventually developed cryptogenic cirrhosis. The patient received adjuvant chemotherapy and died of metastatic disease 18 months after surgery. The thyroid-like pattern broadens the morphologic spectrum of cholangiocarcinoma.

Adenocarcinoma with pyloric gland phenotype of the extrahepatic bile ducts: a previously unrecognized and distinctive morphologic variant of extrahepatic bile duct carcinoma.

Carcinomas of the extrahepatic bile ducts are uncommon and morphologically heterogeneous. We report 3 unique examples of adenocarcinomas that show predominantly pyloric gland differentiation (80%-100%) and compare their immunohistochemical profile with that of pyloric gland adenomas of the gallbladder, foveolar, and intestinal-type adenocarcinomas of the extrahepatic bile duct. The 3 patients with pyloric gland adenocarcinomas were younger than those with conventional adenocarcinomas. The 3 tumors were very well differentiated but showed extensive perineural invasion. They consisted of a variable proportion of small, medium-sized, and cystically dilated glands separated by abundant desmoplastic stroma. The glands were lined by columnar cells with abundant mucin-containing cytoplasm and small hyperchromatic basally placed nuclei with inconspicuous nucleoli. A characteristic feature of these pyloric gland adenocarcinomas was that the glands had a stellar pattern that was not seen in foveolar-, intestinal-, or biliary-type adenocarcinomas. Two pyloric gland adenocarcinomas coexpressed MUC6 and MUC5AC. The diffuse pattern of reactivity of MUC5AC and MUC6 was similar to that of 10 pyloric gland adenomas of the gallbladder and 2 foveolar adenocarcinomas of the extrahepatic bile duct. In contrast, 5 intestinal adenocarcinomas of the extrahepatic bile duct labeled with the intestinal marker CDX2 and 3 with the colonic MUC2 but were negative for MUC6 and MUC5AC. We believe that these pyloric gland adenocarcinomas represent a previously unrecognized distinct clinicopathologic entity. Despite their deceptively benign microscopic appearance, 1 patient died with local recurrence and liver metastasis, another patient is living with tumor, and the third patient is asymptomatic but only 5 months after surgery.