Yong Zulina Zubairi

Estimating the Five-Year Survival of Cervical Cancer Patients Treated in Hospital Universiti Sains Malaysia

OBJECTIVE The objective of this study was to determine the five-year survival among patients with cervical cancer treated in Hospital Universiti Sains Malaysia. METHODS One hundred and twenty cervical cancer patients diagnosed between 1st July 1995 and 30th June 2007 were identified. Data were obtained from medical records. The survival probability was determined using the Kaplan-Meier method and the log-rank test was applied to compare the survival distribution between groups. RESULTS The overall five-year survival was 39.7% [95%CI (Confidence Interval): 30.7, 51.3] with a median survival time of 40.8 (95%CI: 34.0, 62.0) months. The log-rank test showed that there were survival differences between the groups for the following variables: stage at diagnosis (p=0.005); and primary treatment (p=0.0242). Patients who were diagnosed at the latest stage (III-IV) were found to have the lowest survival, 18.4% (95%CI: 6.75, 50.1), compared to stage I and II where the five-year survival was 54.7% (95%CI: 38.7, 77.2) and 40.8% (95%CI: 27.7, 60.3), respectively. The five-year survival was higher in patients who received surgery [52.6% (95%CI: 37.5, 73.6)] as a primary treatment compared to the non-surgical group [33.3% (95%CI: 22.9, 48.4)]. CONCLUSION The five-year survival of cervical cancer patients in this study was low. The survival of those diagnosed at an advanced stage was low compared to early stages. In addition, those who underwent surgery had higher survival than those who had no surgery for primary treatment.

Prognostic systems for NPC: a comparison of the multi layer perceptron model and the recurrent model

Artificial neural networks are considered good alternatives to conventional statistical methods for the prediction of survival. Neural networks have been used in medicine since the late 1980s, first, as an aid to diagnosis and treatment and then, recently, as a tool to study medical prognosis of a variety of diseases. Survival predictions at the individual level can help patients make informed decisions with regards to the quality of life and finance. We describe our research in the use of neural network to predict the prognosis of nasopharyngeal carcinoma. Two prognostic models for nasopharyngeal carcinoma were developed, namely the multi-layer perceptron model and the recurrent model and their performance compared.

Trends in evidence-based treatment and mortality for ST elevation myocardial infarction in Malaysia from 2006 to 2013: time for real change.

BACKGROUND The administration of evidence-based pharmacotherapy and timely primary percutaneous coronary intervention have been shown to improve outcome in ST elevation myocardial infarction (STEMI). However, implementation remains a challenge due to the limitations in facilities, expertise and funding. OBJECTIVES To investigate adherence to guideline-based management and mortality of STEMI patients in Malaysia. DESIGN Retrospective analysis. SETTINGS STEMI patients from 18 participating hospital across Malaysia included in the National Cardiovascular Database-Acute coronary syndrome (NCVD-ACS) registry year 2006 to 2013. PATIENTS AND METHODS Patients were categorized into four subgroups based on the year of admission (2006 to 2007, 2008 to 2009, 2010 to 2011 and 2012 to 2013). Baseline characteristics and clinical presentation, in-hospital pharmacotherapy, invasive revascularization and in-hospital/30-day mortality were analysed and compared between the subgroups. MAIN OUTCOME MEASURE(S) Rate of in-hospital catheterization/percutaneous coronary intervention. RESULTS The registry contained data on 19 483 patients. Intravenous thrombolysis was the main reperfusion therapy. Although the overall rate of in-hospital catheterisation/PCI more than doubled over the study period, while the use of primary PCI only slowly increased from 7.6% in 2006/2007 to 13.6% in 2012/2013. The use of evidence-based oral therapies increased steadily over the years except for ACE-inhibitors and angiotensin-receptor blockers. The adjusted risk ratios (RR) for in-hospital mortality for the four sub-groups have not shown any significant improvement. The 30-day adjusted risk ratios however showed a significant albeit gradual risk reduction (RR 0.773 95% CI 0.679–0.881, P<.001). CONCLUSION Adherence to evidence-based treatment in STEMI in Malaysia is still poor especially in terms of the rate of primary PCI. Although there is a general trend toward reduced 30-day mortality, the reduction was only slight over the study period. Drastic effort is needed to improve adherence and clinical outcomes. LIMITATION Retrospective registry data with inter-hospital variation.

An Alternative Analysis of Two Circular Variables via Graphical Representation

The relationship between variables is vital in data analysis. The scatter plot, for instance, gives an easy preliminary exploratory analysis for finding relationship between two variables, if any. Statistical method such as correlation and linear relationship are standard tools in most statistical packages. For circular variables that take value on the circumference of a circle, the analysis however is different from those of the Euclidean type variables because circumference is a bounded closed space. Unlike linear variable, standard statistical packages for circular variables are limited. This paper proposes a graphical representation of two circular variables as a preliminary analysis using the MATLAB environment. A plot called Spoke plot is developed to visually display relationship between two circular variables and linear correlation. As an illustration, the Malaysian wind data is used in the analysis. This new type of representation promises an alternative approach in the preliminary analysis of circular data.

ANN as a tool for medical prognosis

The analysis of cancer survival is used to determine the efficiency of treatment programmes and protocols; it is also used to determine the type of treatment. At the individual level a prediction of cancer survival can help patients make informed decisions with regards to their quality of life and future finances. Currently available prediction methods apply to groups of people, and may not be adequate to predict treatment outcome for individual patients. This paper presents a conceptual model of a cancer knowledge base incorporating a computer-based predictor for survival. It proposes the use of an artificial neural network (ANN) as an alternative tool for investigating cancer survival. The focus of this research will be on nasopharyngeal carcinoma survival data collected in Malaysia.

A Dynamic Model of Optimal Share Repurchase

We propose a model of dynamic share repurchase. The model highlights the central importance of payout for corporate decisions. Our two main results are: (1) free cash flows depends on the operating cash flows changes; (2) optimal share repurchase timing is decided by the relative changes between the free cash flow and dividends.

A clustering approach to detect multiple outliers in linear functional relationship model for circular data

ABSTRACT Outlier detection has been used extensively in data analysis to detect anomalous observation in data. It has important applications such as in fraud detection and robust analysis, among others. In this paper, we propose a method in detecting multiple outliers in linear functional relationship model for circular variables. Using the residual values of the Caires and Wyatt model, we applied the hierarchical clustering approach. With the use of a tree diagram, we illustrate the detection of outliers graphically. A Monte Carlo simulation study is done to verify the accuracy of the proposed method. Low probability of masking and swamping effects indicate the validity of the proposed approach. Also, the illustrations to two sets of real data are given to show its practical applicability.

Detecting multiple outliers in linear functional relationship model for circular variables using clustering technique

Outlier detection has been used extensively in data analysis to detect anomalous observation in data and has important application in fraud detection and robust analysis. In this paper, we propose a method in detecting multiple outliers for circular variables in linear functional relationship model. Using the residual values of the Caires and Wyatt model, we applied the hierarchical clustering procedure. With the use of tree diagram, we illustrate the graphical approach of the detection of outlier. A simulation study is done to verify the accuracy of the proposed method. Also, an illustration to a real data set is given to show its practical applicability.

A new technique to solve the unidentifiablity problem of linear structural relationship model

Several techniques have been used to solve the unidentifiability problem of linear structural relationship model. Most of them assumed either the error variance σδ2 or σe2 is known or both are known or the ratio of them is known and then can be used to estimate the rest of parameters. In this study, we assume the slope parameter, β is known and then derive the maximum likelihood estimate (MLE) for rest of the parameters. In fact, the slope is estimated separately by a nonparametric method and assumed to be known when the rest of the parameters are estimated by maximum likelihood method. We obtain closed-form estimates of parameters and their variances and covariances. Using a simulation study, we showed that the estimated values of the parameters are unbiased and consistent. Finally, this method is illustrated using real data set.