[Background]
Medical technology is increasing the number of available treatments for cancer. For advanced cancer an increasing number of treatments can result in an increasing treatment sequence. Treatments are given one after another in a cycle of treatment, failure and then another treatment. Newly added treatments may not replace existing treatments. The movement of existing treatments into later lines of therapy is displacement.
Displacement poses challenges for economic evaluation. This thesis addresses three questions.
1. Does the displacement of a treatment alter its cost-effectiveness?
2. If the cost-effectiveness becomes less favourable can any resulting societal welfare loss be corrected by changing the price?
3. Can the required price change be calculated in Australia?
[Methods]
A theoretical framework is developed for displacement in cancer treatment. The implications of decision-making criteria and information gaps are assessed. Real-world data is used to estimate the number of treatments and lines of therapy received by patients, and costs of care. A systematic review and meta-analysis of randomised controlled trials which reported treatment outcomes of multiple lines of therapy is undertaken. The cost-effectiveness of displacing treatments for breast cancer, colorectal cancer and non-small cell lung cancer are modelled.
[Results]
The displacement of a treatment may result in dynamic and allocative inefficiencies. Real-world data showed 13-18% of participants received four or more lines of therapy. The mean health services cost of cancer care was approximately $4 000 per month. Displacement resulted in decreased effectiveness, an increased toxicity per unit time and reduced treatment length.
In the modelling, there was an increase in the incremental cost-effectiveness ratio with displacement. After displacement, reducing the price of cancer treatments by 32% was required to restore cost-effectiveness.
[Conclusions]
There is the potential for displacement in Australia. Displacement results in an increasing incremental cost-effectiveness ratio of a treatment. This can be corrected with price changes in most circumstances.
The Australian real-world data did not record all the treatments that were received by patients. Therefore, it is not able to be used to calculate the price changes that are required with displacement.
The addition of new treatments in Australia should consider the impact of displacement on currently subsidised treatments. A failure to do this results in biased assessments of the benefits and costs of new treatments. It will likely underestimate the cost and overestimate the benefit. Therefore, the potential for displacement should be considered in cancer treatment funding to ensure equity and cost-effectiveness.

https://opus.lib.uts.edu.au/bitstream/10453/129408/2/02whole.pdf

Increasing complexity of cancer care : how displaced treatments impact efficiency, cost-effectiveness and equity

Objective: This paper aims to present preoperative prediction of responses of locally advanced rectal cancer (LARC) patients to neoadjuvant radiochemotherapy (NRC) treatments using their gene expression profiles. Materials and Methods: Expression profiles of 24,026 genes were generated on 43 LARC samples using microarray technology. The 43 samples contained two histopathologic response groups of 14 responders and 29 non-responders to NRC treatment. Using a novel k sequential feature selection and classification (k-SS) method, a subset of gene signatures whose expression levels are correlated with the two response groups was selected for pre-therapeutic prediction of the LARC patients. Results: Five informative gene chips whose expression profiles are strongly correlated with the two histopathologic response groups of the LARC samples were selected. Some of these are protein encoding genes like SF3A1 that functions in the nucleus and helps to convert pre-messenger RNA to mRNA. The average bootstrap.632+ prediction accuracy based on these genes is about 98%. Results from cluster analysis and PCA showed good discrimination of the two clinical response groups. Although, all the five out-of-bag samples were correctly classified, the classification of the entire 43 LARC samples in a 10-fold cross-validation yielded 86% and 93% correct classifications of responders and non-responders respectively. Conclusion: Results from this study equally showed that preoperative prediction of responses of LARC patients to NRC treatment using gene expression profiles is possible. This shall immensely help in clinical diagnosis and treatment of locally advanced rectal carcinomas. Nevertheless, validation studies with larger patient groups might be desirable in future.

Microarray-based Classification of Histopathologic Responses of Locally Advanced Rectal Carcinomas to Neoadjuvant Radiochemotherapy Treatment

In this thesis, a novel sequential genes selection and classification (k-SS) method is proposed. The method is analogous to the classical non-linear stepwise variable selection (SVS) methods but unlike any of the SVS methods, this new method uses the misclassification error rates (MERs) as its search criteria for informative marker genes in any given microarray data. Here, the importance of any selected gene is determined based on its marginal contribution at improving the prediction accuracy of the classification rule. This method ensures continuous selection of more genes in as much as the improvements brought into the decision models by the selected genes are considered to be significant enough by some established test criteria. However, further gene selection terminates when none of the remaining genes is capable at improving the prediction accuracy (lowering the MER) of the current model. Therefore, our approach only seeks to select the best combination of k marker genes that are most predictive of the biological samples in any given microarray data sets. 
An important feature of our new k-SS method is that the size α used by its test is not arbitrarily fixed by the user as common to some of the classical SVS methods. Rather, the value of α at which the best prediction accuracy is achieved (or the best combination of genes is selected) is determined by cross-validation.
The new k-SS classifier competes favourably with selected eight existing classification methods using eleven published microarray data sets. The k-SS classifier is very simple to apply and does not require any rigid assumption for its implementation. Another merit of this method lies in its ability to select only those genes that are of biological relevance to the existing cancer sub-groups in microarray data sets.
Lastly, we proposed a new preliminary feature selection procedure that employs the cross-validated area under the ROC curve (CVAUC) for gene selection. This method is capable at removing all the irrelevant genes at the preliminary selection stage before any standard classifier like the k-SS method is employed on the remaining data set for final optimum gene selection and classification of mRNA samples. Unlike some other data pruning methods, the new method employs the sub-sampling technique of the v-fold cross-validation to ensure consistency and efficiency of selections made at the preliminary selection stage.

Sequential Dimension Reduction and Prediction Methods with High-dimensional Microarray Data

In this study, the effects of some clinical variables on the survival times of patients with bladder cancer were examined. The effects of these variables on sub-distribution of the failure types were determined using the proportional sub-distribution hazards regression model described in Fine and Gray (1999). Published dataset on 294 bladder cancer patients with four clinical outcomes were analyzed using Cumulative Incidence Function approach. The four outcomes included 184 (64%) patients that experienced recurrence of bladder cancer after receiving chemotherapy treatments. Two patients died of bladder cancer while 27 patients died of other causes and the remaining 76 patients did not experienced any of these three outcomes, and as a result, were considered censored. Among the covariates considered, only the patients’ initial number of tumors and initial size of tumor were incorporated into our analysis due to high proportion of missing observations in others. Results from this work showed that, patients with tumour recurrence have highest risk of dying than those from other causes. Further results showed that, the number of tumor was positively associated with the recurrence of cancer of the bladder. However, the size of the tumor did not demonstrate a significant effect on the patients’ survival time. It can therefore, be concluded that patients with tumor recurrence have low probability of survival from bladder cancer than patients that experienced other events. Above all, number, but not size of tumor could adversely affect the survival time of bladder cancer patients, especially those with tumor recurrence after bladder cancer treatment.

Competing Risk Modelling using Cumulative Incidence Function: Application to Recurrent Bladder Cancer Data

This paper presents an efficient multiclass sequential feature selection and classification (mk-SS) method using gene expression signatures. The development of this method employs 10-fold cross-validation to ensure stability. The efficiency of this method is assessed through the misclassification error rate and some other performance measures. The performances of the mk-SS were compared with the classification results of the Support Vector Machines (SVM) over five published multiclass microarray datasets. The results showed that the mk-SS method efficiently selects the informative gene biomarkers for proper classification of the biological groups of the tissue samples. This method competes favourably with SVM in terms of prediction accuracy while it outperforms the SVM in 80% of cases considered. The quality of the features selected by mk-SS algorithm was validated by hybridizing the feature selection scheme of the mk-SS into the standard SVM algorithm which significantly improves the predictive power of the standard SVM method. This work has shown that classification of various cancer type using gene expression profiles is feasible especially when the endpoints are of multi-category. Keywords: k-SS, mk-SS, Support Vector Machines, Microarray, Misclassification error rate

/pdf/multiclass-sequential-feature-selection-and-classification-1jnj50al3u.pdf

Multiclass Sequential Feature Selection and Classification Method for Genomic Data

In this study we investigated the use of (conditional) logistic regression (CLR) techniques to model failure-time data. The widely adopted semi-parametric approach of Cox proportional hazard (CPH) model has been reported to be a successive combination of the CLR approach. This implies that the CLR method could as well be appropriate to model event history situations. The use of this regression technique to model survival data is hereby examined. It is established that this regression method, apart from sharing some features with the Cox model, is an easy alternative to model survival data. The procedure is found to be more flexible to apply especially when time-varying effects of some prognostic factors are suspected. In the two situations considered, results from the CLR models are found to be similar with those obtained from the classical Cox models. Data sets on breast and small-cell lung cancer patients were used to demonstrate our results.

Kurt Ulm

Papers

Survival analysis of Breast and Small-Cell Lung Cancer Patients using Conditional Logistic Regression Models