Tree-structured Survival Analysis Research Articles

Personalized Medicine Using Predictive Analytics: A Machine Learning-Based Prognostic Model for Patients Undergoing Hip Arthroscopy

Background: Personalized medicine models to predict outcomes of orthopaedic surgery are scarce. Many have required data that are only available postoperatively, mitigating their usefulness in preoperative decision making. Purpose: To establish a method for predictive modeling to enable individualized prognostication and shared decision making based on preoperative patient factors using data from a prospective hip preservation registry. Study Design: Cohort study (Prognosis); Level of evidence, 2. Methods: Preoperative data of 2415 patients undergoing hip arthroscopy for femoroacetabular impingement syndrome between February 2008 and November 2017 were retrospectively analyzed. Two machine-learning analyses were evaluated: Tree-structured survival analysis (TSSA) and Cox proportional hazards modeling for predicting time to event and for computing hazard ratios for survivorship after hip arthroscopy. The Fine-Gray model was similarly used for repeat hip arthroscopy. Variables were selected for inclusion using the minimum Akaike Information Criterion (AIC). The stepwise selection was used for the Cox and Fine-Gray models. A web-based calculator was created based on the final models. Results: Prognostic models were successfully created using Cox proportional hazards modeling and Fine-Gray models for survivorship and repeat hip arthroscopy, respectively. The Harrell C-statistics of the Cox model calculators for survivorship after hip arthroscopy and the Fine-Gray model for repeat hip arthroscopy were 0.848 and 0.662, respectively. Using the AIC, 13 preoperative variables were identified as predictors of survivorship, and 6 variables were identified as predictors for repeat hip arthroscopy. In contrast, the TSSA model performed poorly, resulting in a Harrell C-statistic <0.6, rendering it inaccurate and not interpretable. A web-based calculator was created based on these models. Conclusion: This study successfully created an institution-specific machine learning–based prognostic model for predictive analytics in patients undergoing hip arthroscopy. Prognostic models to predict survivorship and the need for repeat surgeries were both adapted into web-based tools to assist the physician with shared decision making. This prognostic model may be useful at other institutions after performing external validation. Additionally, this study may serve as proof of concept for a methodology for the development of patient-specific prognostic models. This methodology may be used to create other predictive analytics models in different realms of orthopaedic surgery, contributing to the evolution from evidence-based medicine to personalized medicine.

Automatic Digital Quantification of Bone Marrow Myeloma Volume in Appendicular Skeletons — Clinical Implications and Prognostic Significance

Multiple myeloma (MM) is a clonal plasma cell disorder originating in bone marrow. Whole body low-dose multidetector CT (MDCT) can depict bone marrow infiltration by myeloma cells in the adipose-rich fatty marrow of the appendicular skeleton. We previously reported that CT values (also known as Hounsfield Units (HU) scale) of bone marrow infiltrations in humeri and femora reflected disease stages in plasma cell dyscrasia and higher CT values predicted inferior prognosis (Nishida et al. Blood Cancer J 2015). However, automated and objective volume measurement of bone marrow infiltration has not been established, and its clinical relevance is not yet understood. To solve this problem, we developed a novel automated CT post-processing software tool (MABLE software) and evaluated the total sum of CT values (cumulative CT values, cCTv), representing mass of bone marrow infiltration by combining volume of myeloma infiltration and voxel-based CT values. cCTv was calculated for 91 patients with newly diagnosed symptomatic MM and 27 patients with smoldering MM (SMM) and 9 patients with monoclonal gammopathy of unknown significance (MGUS). Mean cCTv, expressed as log10 scale, was higher in patients with symptomatic MM than those with SMM or MGUS (mean ± SEM: 3.541 ± 0.06 vs 2.337 ± 0.1 and 3.541 ± 0.06 vs 2.582 ± 0.18, P < 0.0001 by one-way ANOVA). Among symptomatic MM patients, mean cCTv was highest in patients with revised international staging system (R-ISS) III (mean ± SEM: 3.96 ± 0.11, 3.49 ± 0.07, and 3.15 ± 0.14 in patients with R-ISS stage III, R-ISS stage II and R-ISS stage I, respectively; P < 0.0001 by one-way ANOVA). Age, serum albumin and serum or urine M-protein levels were independent predictors for cCTv by linear regression analysis. Mixed graphical model analysis revealed direct relationships between cCTv and age or R-ISS, indicating that cCTv was influenced by age and aggressive disease. Median follow-up duration of symptomatic MM patients was 30.2 months and 26 patients died during the observation period. Tree-structured survival analysis identified a group of patients with cCTv greater than or equal to 4.4 as the first split point, and another group with age above or equal to 84 years old as the second split point (Figure a, the entire cohort was defined as group 1; patients with cCTv ≥ 4.4 were defined as group 3 and patients ≥ 84 years old were defined as group 5). The patients in group 3 had shorter overall survival than those in group 4 (cCTv < 4.4 or younger than 84 years old (median survival, 8.6 months vs not reached, Hazard ratio: 0.129, P= 5.6 x 10-6, Figure b). Multivariate Cox analysis revealed that cCTv greater than or equal to 4.4 was an independent prognostic factor for overall survival (Hazard ratio: 6.09, P < 0.001). Anti-myeloma therapy reduced cCTv after treatment in 32 patients who received MDCT assessment as disease monitoring. Among patients who achieved complete response (CR), those achieved immunophenotypic CR had significantly greater difference of cCTv between before and after treatment than those achieved stringent CR or CR (mean difference ± SEM, 4.158 ± 0.107 vs 3.079 ± 0.263, P < 0.0001), raising the possibility that cCTv reduction predicts deeper response. Collectively, measuring cCTv identified a group of patients with very poor outcome irrespective of age or other clinical variables. [Display omitted] DisclosuresMizobe:Canon Inc.: Employment. Yamamichi:Canon Inc.: Employment.

The role of physical activity in the development of first cardiovascular disease event: a tree-structured survival analysis of the Danish ADDITION-PRO cohort

BackgroundAmbiguity exists in relation to the role of physical activity (PA) for cardiovascular disease (CVD) risk reduction. We examined the interplay between PA dimensions and more conventional CVD risk factors to assess which PA dimensions were associated with the first CVD event and whether subgroup differences exist.MethodsA total of 1449 individuals [median age 65.8 (IQR: 61.2, 70.7) years] with low to high risk of type 2 diabetes and free from CVD from the Danish ADDITION-PRO study were included for survival analysis. PA was measured by individually calibrated heart rate and movement sensing for 7 consecutive days. The associations of different PA dimensions (PA energy expenditure, time spent in light-, moderate- and vigorous intensity PA), sedentary time and other conventional CVD risk factors with the first CVD event were examined by tree-structured survival analysis. Baseline information was linked to data on the first CVD event (ischemic heart disease, ischemic stroke, heart failure, atrial flutter/fibrillation and atherosclerotic disease) and mortality obtained from Danish registers.ResultsDuring a median follow-up time of 5.5 (IQR: 5.1–6.1) years, a total of 201 individuals (13.9%) developed CVD. Overall CVD incidence rate was 2.6/100 person-years. PA energy expenditure above 43 kJ/kg/day was associated with lower rates of CVD events among participants ≤ 70 years and with HbA1c ≤ 5.7% (39 mmol/mol), systolic blood pressure ≤ 156 mmHg and albumin creatinine ratio ≤ 70 (incidence rates 0.0–0.8/100 person-years).ConclusionsAny type of PA resulting in increased PA energy expenditure may over time be the best prevention strategy to uphold reduced risk of CVD.

Automatic digital quantification of bone marrow myeloma volume in appendicular skeletons - clinical implications and prognostic significance

Multiple myeloma (MM) is a clonal plasma cell disorder originating in bone marrow. Whole body low-dose multidetector CT (MDCT) can depict bone marrow infiltration by myeloma cells into the adipose-rich fatty marrow of the appendicular skeleton. However, automated and objective volume measurement of bone marrow infiltration has not been established, and its clinical relevance remains unclear. We therefore developed novel CT post-processing software (MABLE software) and measured the total sum of CT values (cumulative CT value, cCTv) representing bone marrow infiltration, by combining volume and voxel-based CT values. The cCTv was greater in patients with symptomatic MM than in those with smouldering MM or monoclonal gammopathy of unknown significance. Patients with revised International Staging System (R-ISS) III had a higher cCTv than those with R-ISS I or II. Age, albumin, and M-protein levels independently predicted cCTv. Mixed graphical model analysis revealed direct relationships between cCTv and age or R-ISS. Tree-structured survival analysis and multivariate Cox analysis revealed that a cCTv greater than or equal to 4.4 was independently prognostic for overall survival. Anti-myeloma therapy reduced cCTv after treatment. These findings suggest that the automatically calculated cCTv reflects disease aggressiveness and is useful for accurate prognostic prediction in MM patients.

Analysis of Preoperative Metabolic Risk Factors Affecting the Prognosis of Patients with Esophageal Squamous Cell Carcinoma: The Fujian Prospective Investigation of Cancer (FIESTA) Study

Some metabolic factors have been shown to be associated with an increased risk of esophageal cancer; however the association with its prognosis is rarely reported. Here, we assessed the prediction of preoperative metabolic syndrome and its single components for esophageal cancer mortality by analyzing a subset of data from the ongoing Fujian prospective investigation of cancer (FIESTA) study. Between 2000 and 2010, patients who underwent three-field lymphadenectomy were eligible for inclusion. Blood/tissue specimens, demographic and clinicopathologic data were collected at baseline. Metabolic syndrome is defined by the criteria proposed by Chinese Diabetes Society. In this study, analysis was restricted to esophageal squamous cell carcinoma (ESCC) due to the limited number of other histological types. The median follow-up in 2396 ESCC patients (males/females: 1822/574) was 38.2months (range, 0.5–180months). The multivariate-adjusted hazard ratio (HR) of metabolic syndrome for ESCC mortality was statistically significant in males (HR, 95% confidence interval, P: 1.45, 1.14–1.83, 0.002), but not in females (1.46, 0.92–2.31, 0.107). For single metabolic components, the multivariate-adjusted HRs were significant for hyperglycemia (1.98, 1.68–2.33, <0.001) and dyslipidemia (1.41, 1.20–1.65, <0.001) in males and for hyperglycemia (1.76, 1.23–2.51, <0.001) in females, independent of clinicopathologic characteristics and obesity. In tree-structured survival analysis, the top splitting factor in both genders was tumor-node-metastasis stage, followed by regional lymph node metastasis. Taken together, our findings demonstrate that preoperative metabolic syndrome was a significant independent predictor of ESCC mortality in males, and this effect was largely mediated by glyeolipid metabolism disorder.

New tree structured survival analysis for hip fracture of SOF data

New tree structured survival analysis for hip fracture of SOF data

Tree-structured survival analysis of patients with Pseudomonas aeruginosa bacteremia: A multicenter observational cohort study

This study aimed to construct a prediction algorithm, which is readily applicable in the clinical setting, to determine the mortality rate for patients with P. aeruginosa bacteremia. A multicenter observational cohort study was performed retrospectively in seven university-affiliated hospitals in Korea from March 2012 to February 2015. In total, 264 adult patients with monomicrobial P. aeruginosa bacteremia were included in the analyses. Among the predictors independently associated with 30-day mortality in the Cox regression model, Pitt bacteremia score >2 and high-risk source of bacteremia were identified as critical nodes in the tree-structured survival analysis. Particularly, the empirical combination therapy was not associated with any survival benefit in the Cox regression model compared to the empirical monotherapy. This study suggests that determining the infection source and evaluating the clinical severity are critical to predict the clinical outcome in patients with P. aeruginosa bacteremia.

Risk stratification for 25-year cardiovascular disease incidence in type 1 diabetes: Tree-structured survival analysis of the Pittsburgh Epidemiology of Diabetes Complications study

The formal identification of subgroups with varying levels of risk is uncommon in observational studies of cardiovascular disease, although such insight might be useful for clinical management. Tree-structured survival analysis was utilized to determine whether there are meaningful subgroups at varying levels of cardiovascular disease risk in the Pittsburgh Epidemiology of Diabetes Complications study, a prospective cohort study of childhood-onset (<17 years old) type 1 diabetes. Of the 561 participants free of cardiovascular disease (coronary artery disease, stroke or lower extremity arterial disease) at baseline, 263 (46.9%) had an incident cardiovascular disease event over the 25-year follow-up. Tree-structured survival analysis revealed a range of risk groups, from 24% to 85%, which demonstrate that those with short diabetes duration and elevated non-high-density lipoprotein cholesterol have similar cardiovascular disease risk to those with long diabetes duration and that renal disease is a better discriminator of risk in men than in women. Our findings suggest that subgroups with major cardiovascular disease risk differences exist in this type 1 diabetes cohort. Using tree-structured survival analysis may help to identify these groups and the interrelationships between their associated risk factors. This approach may improve our understanding of various clinical pathways to cardiovascular disease and help target intervention strategies.

Prognostic classification index in Iranian colorectal cancer patients: Survival tree analysis.

Aims:The aim of this study was to determine the prognostic index for separating homogenous subgroups in colorectal cancer (CRC) patients based on clinicopathological characteristics using survival tree analysis.Methods:The current study was conducted at the Research Center of Gastroenterology and Liver Disease, Shahid Beheshti Medical University in Tehran, between January 2004 and January 2009. A total of 739 patients who already have been diagnosed with CRC based on pathologic report were enrolled. The data included demographic and clinical-pathological characteristic of patients. Tree-structured survival analysis based on a recursive partitioning algorithm was implemented to evaluate prognostic factors. The probability curves were calculated according to the Kaplan-Meier method, and the hazard ratio was estimated as an interest effect size.Result:There were 526 males (71.2%) of these patients. The mean survival time (from diagnosis time) was 42.46± (3.4). Survival tree identified three variables as main prognostic factors and based on their four prognostic subgroups was constructed. The log-rank test showed good separation of survival curves. Patients with Stage I-IIIA and treated with surgery as the first treatment showed low risk (median = 34 months) whereas patients with stage IIIB, IV, and more than 68 years have the worse survival outcome (median = 9.5 months).Conclusion:Constructing the prognostic classification index via survival tree can aid the researchers to assess interaction between clinical variables and determining the cumulative effect of these variables on survival outcome.

Time-dependent tree-structured survival analysis with unbiased variable selection through permutation tests.

Incorporating time-dependent covariates into tree-structured survival analysis (TSSA) may result in more accurate prognostic models than if only baseline values are used. Available time-dependent TSSA methods exhaustively test every binary split on every covariate; however, this approach may result in selection bias toward covariates with more observed values. We present a method that uses unbiased significance levels from newly proposed permutation tests to select the time-dependent or baseline covariate with the strongest relationship with the survival outcome. The specific splitting value is identified using only the selected covariate. Simulation results show that the proposed time-dependent TSSA method produces tree models of equal or greater accuracy as compared to baseline TSSA models, even with high censoring rates and large within-subject variability in the time-dependent covariate. To illustrate, the proposed method is applied to data from a cohort of bipolar youths to identify subgroups at risk for self-injurious behavior.

Coping with health stresses and remission from late‐life depression in primary care: a two‐year prospective study

Identifying the predictors of late-life depression that are amenable to change may lead to interventions that result in better and faster remission. Thus, the authors investigated the impact of two different strategies for coping with physical illness on depression in older, primary care patients. Health-oriented goal engagement strategies involve the investment of cognitive and behavioral resources to achieve health goals. Conversely, disengagement strategies involve the withdrawal of these resources from obsolete or unattainable health goals, combined with goal restructuring. The participants were 271 adults aged >59 years who took part in a two-year randomized clinical trial for treating depression in older adults (Prevention of Suicide in Primary Care Elderly: Collaborative Trial). The use of engagement and disengagement strategies, along with other risk factors for depression, were included in a tree-structured survival analysis to identify subgroups of individuals at risk for not achieving depression remission. The use of disengagement strategies predicted earlier remission of depression, particularly among more severely depressed older patients. The use of engagement strategies did not predict earlier remission. Interventions that encourage disengagement from unattainable health goals may promote remission from depression in older, primary care patients.

Developing and comparing two different prognostic indexes for predicting disease-free survival of nonmetastatic breast cancer patients

To determine 2 different prognostic indexes (PI) for the differentiation of subgroups of nonmetastatic breast cancer patients with the Cox regression analysis and survival tree (ST) methods and the additional usage of the Kaplan-Meier estimates to investigate the predictive power of these methods. Materials and methods: Prognostic factors data were collected for 410 patients. The Cox regression analysis examines the relationship of the survival distribution and covariates. The ST method is a tree-structured survival analysis based on a recursive partitioning algorithm. In this study, Harrell's concordance indexes of models for training and test sets were computed. Furthermore, survival curves were estimated by the Kaplan-Meier method. Disease-free survival (DFS) was calculated from the time of initial diagnosis (initiation of the first treatment) to the first recurrence of disease. Results: After a median follow-up of 48 months, 100 (24.4%) patients have had at least 1 of the DFS events. In Cox regression analysis, we proposed the simple PI, which is a sum of axillary nodal and HER2/neu status. In the ST method, we identified 3 variables: HER2/neu, axillary nodal, and estrogen receptor status. The axillary nodal status was the most important determining factor for recurrence. Conclusion: We found that the PI of the ST and Cox regression methods had similar performance levels in predicting DFS, and the error rates of the models were close to each other in the training and test sets. Furthermore, we determined that the axillary nodal status and HER2/neu were the most important determining factors for prediction of DFS in breast cancer patients.

A stochastic multiple imputation algorithm for missing covariate data in tree-structured survival analysis

Missing covariate data present a challenge to tree-structured methodology due to the fact that a single tree model, as opposed to an estimated parameter value, may be desired for use in a clinical setting. To address this problem, we suggest a multiple imputation algorithm that adds draws of stochastic error to a tree-based single imputation method presented by Conversano and Siciliano (Technical Report, University of Naples, 2003). Unlike previously proposed techniques for accommodating missing covariate data in tree-structured analyses, our methodology allows the modeling of complex and nonlinear covariate structures while still resulting in a single tree model. We perform a simulation study to evaluate our stochastic multiple imputation algorithm when covariate data are missing at random and compare it to other currently used methods. Our algorithm is advantageous for identifying the true underlying covariate structure when complex data and larger percentages of missing covariate observations are present. It is competitive with other current methods with respect to prediction accuracy. To illustrate our algorithm, we create a tree-structured survival model for predicting time to treatment response in older, depressed adults.

Cost-Saving Tree-Structured Survival Analysis for Hip Fracture of Study of Osteoporotic Fractures Data

It is important to predict osteoporotic fracture risk accurately in order to select high-risk patients for treatment. Previous tree-structured survival analysis (TSSA) methods focused on optimization in statistical performance in construction of survival trees. However, they did not take into account the cost of the predictive variables. Because of the high cost of some predictors, the derived algorithm may have only limited application in practice. In this article, the authors consider the cost-effectiveness in TSSA and propose a cost-saving TSSA (denoted as CSTSSA) to construct the survival tree for identifying subjects at high risk of hip fracture based on the data from Study of Osteoporotic Fractures. The new rule is compared with the optimum classification based on log-rank test statistics using the noninferiority test by Lu and others. The comparison results suggest that, for identifying patients at high risk of hip fracture, the CSTSSA is a good alternative to the optimum classification.

Nosocomial infections in brazilian pediatric patients: using a decision tree to identify high mortality groups

Nosocomial infections (NI) are frequent events with potentially lethal outcomes. We identified predictive factors for mortality related to NI and developed an algorithm for predicting that risk in order to improve hospital epidemiology and healthcare quality programs. We made a prospective cohort NI surveillance of all acute-care patients according to the National Nosocomial Infections Surveillance System guidelines since 1992, applying the Centers for Disease Control and Prevention 1988 definitions adapted to a Brazilian pediatric hospital. Thirty-eight deaths considered to be related to NI were analyzed as the outcome variable for 754 patients with NI, whose survival time was taken into consideration. The predictive factors for mortality related to NI (p < 0.05 in the Cox regression model) were: invasive procedures and use of two or more antibiotics. The mean survival time was significantly shorter (p < 0.05 with the Kaplan-Meier method) for patients who suffered invasive procedures and for those who received two or more antibiotics. Applying a tree-structured survival analysis (TSSA), two groups with high mortality rates were identified: one group with time from admission to the first NI less than 11 days, received two or more antibiotics and suffered invasive procedures; the other group had the first NI between 12 and 22 days after admission and was subjected to invasive procedures. The possible modifiable factors to prevent mortality involve invasive devices and antibiotics. The TSSA approach is helpful to identify combinations of predictors and to guide protective actions to be taken in continuous-quality-improvement programs.

Age and drug therapy are key prognostic factors for first clinical fracture in patients with primary osteoporosis

We evaluated the characteristics of 1,142 women and men who attended Canadian osteoporosis clinics and had T-score < or = -2.0 and no prior fractures to determine the predictors of first clinical fracture. Greater age and failure to start osteoporosis drug treatment increased the risk of first clinical fracture. Although risk factors for osteoporotic fractures are well-known, it is unclear which factors predict poor prognosis in patients with primary osteoporosis. The purpose of this study was to determine prognostic factors for first clinical fracture in patients with T-score < or = -2.0 and no previous clinical fracture. We examined prospectively collected data from 1,142 patients aged 40 and over in the Canadian Database of Osteoporosis and Osteopenia. We used prognosis methodology and performed survival analysis to determine factors that increase the risk of first clinical fracture. Our inception cohort (mean age = 60.6 years, 91% females) had a cumulative fracture incidence of 5.1% (incidence rate: 2.53/100 person-years). Age and osteoporosis drug use predicted incident clinical fractures in multivariable regression analyses. The risk of first fracture increased by 3% per year. Failing to initiate osteoporosis treatment increased fracture risk by 2.4 times. In addition, low physical activity, high body mass index and low T-scores were found to predict fracture risk in certain patient subgroups using tree-structured survival analysis. These findings were robust and did not change with most sensitivity analyses. Age and osteoporosis drug treatment are the main prognostic predictors of first clinical fracture in patients with T-score < or = -2.0.

334 POSTER The validity of tree-structured survival analysis of prognostic factors in breast cancer

334 POSTER The validity of tree-structured survival analysis of prognostic factors in breast cancer

Identifying Genetic Susceptibilities to Diabetes-related Complications among Individuals at Low Risk of Complications: An Application of Tree-Structured Survival Analysis

The authors hypothesized that genetic predisposition to diabetes complications would be more evident among low-risk individuals and aimed to identify genes related to developing complications (confirmed distal symmetric polyneuropathy, overt nephropathy, or coronary artery disease) in low-risk groups. Participants in the Pittsburgh, Pennsylvania, Epidemiology of Diabetes Complications Study of childhood-onset type 1 diabetes, first seen in 1986-1988 (mean age, 28 years; diabetes duration, 19 years), were reexamined biennially for 10 years. For each complication, subgroups with the lowest disease risk were identified by using tree-structured survival analysis, and 15 candidate genes were compared between subjects with and without complications. In the group with the lowest incidence of confirmed distal symmetric polyneuropathy (n = 123), confirmed distal symmetric polyneuropathy risk increased fivefold for those with the eNOS GG genotype (p < 0.05). In the group with the lowest risk of overt nephropathy (n = 340), the ACE D polymorphism increased overt nephropathy risk twofold (p = 0.05), whereas a protective effect was observed for the LIPC CC genotype (p < 0.05). In the group with the lowest incidence of coronary artery disease (n = 331), the MTHFR CC genotype increased coronary artery disease risk threefold (p < 0.05). Tree-structured survival analysis may help identify genetic predispositions among individuals who, despite low risk, develop diabetes-related complications.

Pretreatment cytogenetics add to other prognostic factors predicting complete remission and long-term outcome in patients 60 years of age or older with acute myeloid leukemia: results from Cancer and Leukemia Group B 8461

We investigated the relative prognostic significance of cytogenetics in 635 adult acute myeloid leukemia (AML) patients 60 years of age or older treated on front-line protocols. Classification trees and tree-structured survival analysis (TSSA) were used to identify important cytogenetic groups, and their prognostic significance was then assessed in multivariable analysis (MVA). Overall, 48.5% achieved complete remission (CR); 6.6% survived at 5 years. Complex karyotypes with at least 3 abnormalities (complex > or = 3) and a group including "rare aberrations" predicted lower CR rates (25% and 30%) versus other patients (56%). Compared with complex > or = 3, the odds of CR were significantly higher for noncomplex karyotypes without rare aberrations on MVA. Cytogenetically, complex > or = 5 predicted inferior disease-free survival on TSSA, remaining significant on MVA together with white blood cell count (WBC), sex, and age. For survival, complex > or = 5, rare aberrations, and core-binding factor (CBF) abnormalities were prognostic (P < .001), with 5-year survivals of 0%, 0%, and 19.4%, respectively, and 7.5% for remaining patients. Together with WBC, marrow blasts, sex, and age, the cytogenetic groups remained significant on MVA. In conclusion, pretreatment cytogenetics adds to other prognostic factors in older AML patients. Patients with complex > or = 5 appear to benefit minimally from current treatment and are better suited for investigational therapy or supportive care.

The Clinical Relevance of Stromal Matrix Metalloproteinase Expression in Ovarian Cancer

Matrix metalloproteinases (MMP) are proteolytic enzymes implicated in cancer progression and metastasis. We sought to determine the role of epithelial (tumor cell-derived) and stromal (host-derived) expression of MMPs in predicting the clinical outcome of patients with epithelial ovarian cancer (EOC). MMP-2, MMP-9, and membrane type 1 (MT1)-MMP expression was evaluated using immunohistochemistry in 90 invasive EOCs, and samples were scored for epithelial and stromal staining. Results were correlated with clinicopathologic characteristics using univariate and multivariate analyses. High expression of MMP-2, MMP-9, and MT1-MMP in tumor epithelium was detected in 54%, 97%, and 100% of cases, and in stromal compartments, in 38%, 70%, and 38% of cases, respectively. High stromal expression of MMP-2, MMP-9, and MT1-MMP was significantly associated with aggressive features such as high stage, high grade ascites, and positive lymph node status. Kaplan-Meier analysis showed that high epithelial and stromal expression of MMP-2, MMP-9, and MT1-MMP were each significantly associated with shorter disease-specific survival (DSS; P < 0.01). On tree-structured survival analysis, patients with strong epithelial MT1-MMP expression had the shortest DSS, whereas patients with moderate epithelial MT1-MMP and low stromal MMP-9 expression had the longest DSS (P < 0.01). On multivariate analysis, high stromal expression of MMP-9 (P = 0.01) and MT1-MMP (P = 0.04), strong epithelial MT1-MMP (P = 0.01) and high stage (P = 0.04) were independent predictors of poor DSS. Overexpression of stromal MMP-9 and MT1-MMP is independently associated with shorter DSS in EOC. Thus, host-derived MMPs are valuable predictors of clinical outcome in EOC.