Alterations in leukocyte transcriptional control pathway activity associated with major depressive disorder and antidepressant treatment.

Major depressive disorder (MDD) is associated with a significantly elevated risk of developing serious medical illnesses such as cardiovascular disease, immune impairments, infection, dementia and premature death. Previous work has demonstrated immune dysregulation in subjects with MDD. Using genome-wide transcriptional profiling and promoter-based bioinformatic strategies, we assessed leukocyte transcription factor (TF) activity in leukocytes from 20 unmedicated MDD subjects versus 20 age-, sex- and ethnicity-matched healthy controls, before initiation of antidepressant therapy, and in 17 of the MDD subjects after 8 weeks of sertraline treatment. In leukocytes from unmedicated MDD subjects, bioinformatic analysis of transcription control pathway activity indicated an increased transcriptional activity of cAMP response element-binding/activating TF (CREB/ATF) and increased activity of TFs associated with cellular responses to oxidative stress (nuclear factor erythroid-derived 2-like 2, NFE2l2 or NRF2). Eight weeks of antidepressant therapy was associated with significant reductions in Hamilton Depression Rating Scale scores and reduced activity of NRF2, but not in CREB/ATF activity. Several other transcriptional regulation pathways, including the glucocorticoid receptor (GR), nuclear factor kappa-B cells (NF-κB), early growth response proteins 1-4 (EGR1-4) and interferon-responsive TFs, showed either no significant differences as a function of disease or treatment, or activities that were opposite to those previously hypothesized to be involved in the etiology of MDD or effective treatment. Our results suggest that CREB/ATF and NRF2 signaling may contribute to MDD by activating immune cell transcriptome dynamics that ultimately influence central nervous system (CNS) motivational and affective processes via circulating mediators.

Prediction of individual patient outcome in cancer: comparison of artificial neural networks and Kaplan--Meier methods.

BACKGROUND: There is a great need for accurate treatment and outcome prediction in cancer. Two methods for prediction, artificial neural networks and Kaplan--Meier plots, have not, to the authors' knowledge, been compared previously. METHODS: This review compares the advantages and disadvantages of the use of artificial neural networks and Kaplan--Meier curves for treatment and outcome prediction in cancer. RESULTS: Artificial neural networks are useful for prediction of outcome for individual patients with cancer because they are as accurate as the best traditional statistical methods, are able to capture complex phenomena without a priori knowledge, and can be reduced to a simpler model if the phenomena are not complex. Kaplan--Meier plots are of limited accuracy for prediction because they require partitioning of variables, require cutting continuous variables into discrete pieces, and can only handle one or two variables effectively. CONCLUSIONS: Artificial neural networks are an efficient statistical method for outcome prediction in cancer that utilizes all available powerful prognostic factors and maximizes predictive accuracy. Use of Kaplan--Meier plots for predictions is discouraged because of serious technical limitations and low accuracy.

HbA1c predicts length of stay in patients admitted for coronary artery bypass surgery.

Diabetes mellitus is associated with coronary artery disease, and diabetic patients are frequently referred for coronary bypass graft surgery. It is well known that HbA1c, which reflects long-term glycemic control, is related to diabetic morbidity and mortality. It is not known whether HbA1c is related to postoperative length of stay among patients who undergo coronary artery bypass surgery. The authors evaluated 135 patients who underwent bypass surgery at the Westchester Medical Center (Valhalla, NY). HbA1c was measured in all patients preoperatively; a value of 7% or greater was used as a threshold for uncontrolled hyperglycemia. A postoperative length of stay of 6 days or more was used as the cutoff for an extended length of stay. Linear regression was used to assess the relationship between HbA1c, adjusted for age, and length of stay in days. Logistic regression, with length of stay a binary variable <6, > or =6 days, was used to assess the accuracy of HbA1c <7%, > or =7%, adjusted for age, in predicting length of stay. An HbA1c of 7% or greater was found to be a strong predictor of a length of stay of 6 days or longer. These data suggest that HbA1c can be used as a surrogate marker for cardiac and noncardiac morbidity that prolongs hospitalization after coronary artery bypass surgery.

Adherence to prescription medications among medical professionals.

BACKGROUND: We evaluated adherence to medication usage by health care professionals to estimate the expected upper limit of adherence among the general population. METHODS: In a self-administered survey, physicians and nurses were asked about their use of prescribed medications for acute and chronic illnesses. The settings were a teaching hospital, employee health service, medical college, and educational conferences. RESULTS: Among 435 respondents, 301 physicians and nurses had medications prescribed for acute and/or chronic illnesses within 2 years of the survey. Of 610 prescribed medications, > or =80% were taken as prescribed, with a 77% compliance rate for short-term medications and 84% for long-term medications. Older age was associated with better adherence, whereas a greater number of doses per day was associated with poorer adherence. CONCLUSIONS: Approximately 80% of respondents reported properly taking prescription medications > or =80% of the time. Given the nature of the study population, it is unlikely that a nonclinical trial population will consistently achieve better adherence without specific interventions.

Clinical and laboratory evaluation of all-trans retinoic acid modulation of chemotherapy in patients with acute myelogenous leukaemia.

All-trans retinoic acid (ATRA) is synergistic with chemotherapy in leukaemia cell lines. We treated 53 patients with newly diagnosed acute myelogenous leukaemia (AML) with high-dose cytarabine-based chemotherapy followed by ATRA. Peripheral blood and bone marrow samples were obtained to study the effect of in vitro exposure to ATRA and to measure apoptosis and bcl-2. The response rate was 72% for patients under age 60 years and 46% for patients aged 60 years or above. There was no difference in the percentage of responding patients, time to recurrence or overall survival for patients receiving chemotherapy with ATRA vs. historical controls receiving chemotherapy without ATRA. After in vitro exposure of day 3 bone marrow samples to ATRA, there was an increase in apoptotic cells in 25% of patient samples compared with samples not exposed to ATRA. Later date of peak apoptosis in peripheral blood and higher percentage of apoptotic cells in bone marrow on day 3 of treatment were associated with lack of clinical response to treatment. Increased bcl-2 in patient samples was associated with shorter time to recurrence and poor cytogenetic risk. The addition of ATRA to chemotherapy did not improve patient outcome. However, evidence of in vitro response to ATRA in 25% of patients suggests that retinoid pathways should be studied further in patients with AML.

Discovering patterns in microarray data.

The human genome is a complex system characterized by gene interactions and nonlinear behaviors. Complex systems cannot be viewed as the aggregate of their isolated pieces but must be studied as an integrated whole. Microarray technologies offer the opportunity to see the entire biological system as it existed at one moment in time. It is tempting to try to analyze the entire microarray at once to immediately discover the pattern being sought, for example, the pattern of a breast cancer. However, such an analysis would be a mistake because microarrays provide massively parallel information, the analysis of which is a nondeterministic polynomial time (NP)-hard problem. Current statistical methods are not sufficiently powerful to solve this NP-hard problem. The best approach to microarray analysis is to begin with a small number of the elements in the microarray known to be a pattern and ask questions of the other elements in the microarray; i.e., perform instantaneous scientific experiments regarding whether each of the other elements in the microarray are related to the known pattern.

Artificial neural networks applied to survival prediction in breast cancer.

In this study, we evaluated the accuracy of a neural network in predicting 5-, 10- and 15-year breast-cancer-specific survival. A series of 951 breast cancer patients was divided into a training set of 651 and a validation set of 300 patients. Eight variables were entered as input to the network: tumor size, axillary nodal status, histological type, mitotic count, nuclear pleomorphism, tubule formation, tumor necrosis and age. The area under the ROC curve (AUC) was used as a measure of accuracy of the prediction models in generating survival estimates for the patients in the independent validation set. The AUC values of the neural network models for 5-, 10- and 15-year breast-cancer-specific survival were 0.909, 0.886 and 0.883, respectively. The corresponding AUC values for logistic regression were 0.897, 0.862 and 0.858. Axillary lymph node status (N0 vs. N+) predicted 5-year survival with a specificity of 71% and a sensitivity of 77%. The sensitivity of the neural network model was 91% at this specificity level. The rate of false predictions at 5 years was 82/300 for nodal status and 40/300 for the neural network. When nodal status was excluded from the neural network model, the rate of false predictions increased only to 49/300 (AUC 0. 877). An artificial neural network is very accurate in the 5-, 10- and 15-year breast-cancer-specific survival prediction. The consistently high accuracy over time and the good predictive performance of a network trained without information on nodal status demonstrate that neural networks can be important tools for cancer survival prediction.

Specimen banks for cancer prognostic factor research.

Prognostic factors are necessary for determining whether a patient will require therapy, for selecting the optimal therapy, and for evaluating the effectiveness of the therapy chosen. Research in prognostic factors has been hampered by long waiting times and a paucity of outcomes. Specimen banks can solve these problems, but their implementation and use give rise to many important and complex issues. This paper presents an overview of some of the issues related to the use of specimen banks in prognostic factor research.

Predicting response to adjuvant and radiation therapy in patients with early stage breast carcinoma.

BACKGROUND: Screening and surveillance is increasing the detection of early stage breast carcinoma. The ability to predict accurately the response to adjuvant therapy (chemotherapy or tamoxifen therapy) or postlumpectomy radiation therapy in these patients can be vital to their survival, because this prediction determines the best postsurgical therapy for each patient. METHODS: This study evaluated data from 226 patients with TNM Stage I and early Stage II breast carcinoma and included the variables p53 and c-erbB-2 (HER-2/neu). The area under the receiver operating characteristic curve (Az) was the measure of predictive accuracy. The prediction endpoints were 5- and 10-year overall survival. RESULTS: For Stage I and early Stage II patients, the 5- and 10-year predictive accuracy of the TNM staging system were at chance level, i.e., no better than flipping a coin. Both the 5- and 10-year artificial neural networks (ANNs) were very accurate--significantly more so than the TNM staging system (Az 5-year survival, TNM = 0.567, ANN = 0.758; P < 0.001; Az 10-year survival, TNM = 0.508, ANN = 0.894; P < 0.0001). For patients not receiving postsurgical therapy and for either chemotherapy or tamoxifen therapy, the ANNs containing p53 and c-erbB-2 and the number of positive lymph nodes were accurate predictors of survival (Az 5-year survival, 0.781, 0.789, and 0.720, respectively). CONCLUSIONS: The molecular genetic variables p53 and c-erbB-2 and the number of positive lymph nodes are powerful predictors of survival, and using ANN statistical models is a powerful method for predicting responses to adjuvant therapy or radiation therapy in patients with breast carcinoma. ANNs with molecular genetic prognostic factors may improve therapy selection for women with early stage breast carcinoma.

Artificial neural networks improve the accuracy of cancer survival prediction.

BACKGROUND: The TNM staging system originated as a response to the need for an accurate, consistent, universal cancer outcome prediction system. Since the TNM staging system was introduced in the 1950s, new prognostic factors have been identified and new methods for integrating prognostic factors have been developed. This study compares the prediction accuracy of the TNM staging system with that of artificial neural network statistical models. METHODS: For 5-year survival of patients with breast or colorectal carcinoma, the authors compared the TNM staging system's predictive accuracy with that of artificial neural networks (ANN). The area under the receiver operating characteristic curve, as applied to an independent validation data set, was the measure of accuracy. RESULTS: For the American College of Surgeons' Patient Care Evaluation (PCE) data set, using only the TNM variables (tumor size, number of positive regional lymph nodes, and distant metastasis), the artificial neural network's predictions of the 5-year survival of patients with breast carcinoma were significantly more accurate than those of the TNM staging system (TNM, 0.720; ANN, 0.770; P < 0.001). For the National Cancer Institute's Surveillance, Epidemiology, and End Results breast carcinoma data set, using only the TNM variables, the artificial neural network's predictions of 10-year survival were significantly more accurate than those of the TNM staging system (TNM, 0.692; ANN, 0.730; P < 0.01). For the PCE colorectal data set, using only the TNM variables, the artificial neural network's predictions of the 5-year survival of patients with colorectal carcinoma were significantly more accurate than those of the TNM staging system (TNM, 0.737; ANN, 0.815; P < 0.001). Adding commonly collected demographic and anatomic variables to the TNM variables further increased the accuracy of the artificial neural network's predictions of breast carcinoma survival (0.784) and colorectal carcinoma survival (0.869). CONCLUSIONS: Artificial neural networks are significantly more accurate than the TNM staging system when both use the TNM prognostic factors alone. New prognostic factors can be added to artificial neural networks to increase prognostic accuracy further. These results are robust across different data sets and cancer sites.

Insulin like growth factor-binding protein-1 as a marker for hyperinsulinemia in obese menopausal women.

Hyperinsulinemia, a manifestation of insulin resistance, precursor of non-insulin dependent diabetes mellitus (NIDDM) and the hallmark of Syndrome X was assessed in 27 obese post-menopausal women. Insulin-like growth factor binding protein-1 (IGFBP-1), which had been shown previously to correlate inversely with insulin in animal and human studies, was evaluated as a diagnostic marker for abnormal glucose stimulated area under the curve (AUC) insulin (defined a priori as > or = 100 microU/ml). We performed analysis of variance and logistic regression to assess IGFBP-1 and other study covariates, including body mass index, blood pressure, lipids and measures of glucose and insulin in hyperinsulinemic vs. normal women and evaluated performance characteristics (sensitivity, specificity, positive and negative predictive values and accuracy rates). The mean IGFBP-1 was 6.1 ng/ml (95% confidence interval (CI) 3.1 to 8.9) for the hyper-insulinemic women compared to 33.5 ng/ml (CI 15.8 to 51.2) for normal women (P = .0027). At a cutoff point of 15ng/ml, which was selected to correspond to the lower 95% confidence limit for the normal study population, IGFBP-1 was abnormal in all 13 women with hyperinsulinemia and 4 women with normal insulin levels (sensitivity 100%, specificity 69%; positive predictive value 76%, negative predictive value 100%, diagnostic accuracy rate 85%). Logistic regression models indicated that, of all study covariates, IGFBP-1 was the best predictor variable for AUC-insulin as a binary dependent variable. These results suggest that IGFBP-1 may be a simple serum marker for hyperinsulinemia in a subpopulation of obese menopausal women.

Increasing the power of surrogate endpoint biomarkers: the aggregation of predictive factors.

A variable that predicts an outcome with sufficient accuracy is called a predictive factor. Predictive factors can be divided into three types based on the outcomes to be predicted and on the accuracy with which they can be predicted. These three types include risk factors, where the main outcome of interest is incidence and the predictive accuracy is less than 100%; diagnostic factors, where the main outcome of interest is also incidence but the predictive accuracy is almost 100%; and prognostic factors, where the main outcome of interest is death and the predictive accuracy is variable. Surrogate outcomes are predictive factors that are used for a purpose beyond the prediction of an outcome--surrogate outcomes are predictive factors that are substituted for the true outcome in order to determine the effectiveness of an intervention. Surrogate outcomes used in clinical trials are called intermediate endpoints and surrogate endpoints. Predictive factors used as surrogate outcomes have a poor accuracy rate in predicting the true outcome; aggregating risk factors increases predictive accuracy. Artificial neural networks effectively combine predictive factors. Aggregating predictive factors increases the degree of linkage of the surrogate outcome to the true outcome. The resulting increase in predictive accuracy allows enrollment of people most likely to benefit from intervention. This increases the trial's efficiency, reducing the number of people required to assess a chemopreventive agent.

Artificial neural networks for cancer research: outcome prediction.

The use of artificial neural networks in biological and medical research has increased tremendously in the last few years. Artificial neural networks are being used in cancer research for image processing, the analysis of laboratory data for breast cancer diagnosis, the discovery of chemotherapeutic agents, and for cancer outcome prediction. A neural network generalizes from the input data to patterns inherent in the data, and its uses these patterns to make predictions or to classify. This paper explains how neural networks work, and it shows that a neural network is more accurate at predicting breast cancer patient outcome than the current staging system.