Advances in the treatment of ovarian cancer have had a limited impact on prognosis over recent decades. Alternatives to the traditional surgical and chemotherapeutic approach are being sought. Many novel therapies relate to a greater understanding of the molecular changes which occur during carcinogenesis and the development of targeted therapies to exploit these abnormalities. The aim of this thesis was to investigate the prognostic significance of markers relating to tumour immunology, angiogenesis and apoptosis, through the use of Tissue Microarray Technology. 339 cases of ovarian cancers diagnosed between 1982 and 1997 were assessed. Tumours were analysed immunohistochemically for expression of components of the IFNy (IFNGR1, STAT1, p27, caspase 1), TRAIL (DR4 and DR5) and angiogenic (VEGF) pathways. Loss of expression of IFNGR1 was an independent predictor of poor prognosis, although STAT 1 was not. High levels of cytoplasmic and nuclear p27 expression were associated with a reduced survival; cytoplasmic was independently prognostic. Tumours with reduced levels of caspase 1 had improved survival. These results suggest that only patients expressing IFNGR1 may benefit from IFNy therapy and provides evidence of immunoediting in ovarian cancer. DR4 and DR5 did not predict prognosis suggesting that the TRAIL pathway may not be significant in ovarian cancer apoptosis with implications for TRAIL-related therapy. High levels of VEGF occurred infrequently, being an independent marker of poor prognosis. This may identify a group of patients who may preferentially benefit from anti-angiogenic therapy. The thesis illustrates that ovarian cancers are heterogeneous and variations in expression of protein markers can predict tumour behaviour and stratify for therapy. Future targeted therapies may be selected on the basis of an immunohistochemical profile which predicts the pathways that are still functioning. New therapies as they arise should be trialed and targeted to tumours expressing the appropriate molecular markers.

Ovarian cystectomy is commonly performed to treat benign ovarian cysts, but might cause inadvertent damage to normal ovarian tissue, thereby influencing a woman’s ovarian reserve. Ovarian reserve is defined as the existent quantitative and qualitative supply of follicles which are found in the ovaries that can potentially develop into mature follicles which in effect determine a woman’s reproductive potential. It is commonly quantified by the levels of serum FSH and recently by total antral follicle count (2.0-10.0 mm follicles in both ovaries) and AMH levels. These tests however have inherent biological variation in relation to menstrual cycle and ageing; and are also influenced by the intra- and inter-observer variations. The aim of this thesis was to develop a reliable method of examining the effect of ovarian cystectomy on ovarian reserve. I began by examining the ultrasound markers of ovarian reserve. AFC is measured using 2D ultrasound and there is some evidence that 3D ultrasound can make more reliable counts than 2D. I examined the reliability of these two methods and compared them to a new 3D assisted method, SonoAVC which is designed to make automated AFC. I found that the intra- and inter-observer reliability of SonoAVC in counting the number of antral follicles was superior to 2D and 3D manual methods. It however required post-processing of the counts by manually clicking on the antral follicles initially missed in the automated version, thereby making it a semi-automated method. I then compared 2D ultrasound to SonoAVC in measuring the size of antral follicles as there is increasing evidence that the small antral follicles might be more predictive of ovarian reserve. I found that SonoAVC measured the size of antral follicles significantly quicker than 2D and also that the number of small follicles measured by 2D were more than SonoAVC, thereby raising the possibility that 2D might overestimate the number of small antral follicles. I then studied the ability of antral follicle counts stratified by size in prediction of ovarian response and pregnancy. I found that the small antral follicles measuring between 2.0-4.0 mm were independent predictors of clinical pregnancy and ovarian response to assisted reproduction treatment. I then examined the AFCs of different sizes made by SonoAVC and 2D in bovine ovaries and compared to the follicles obtained by manually dissecting the follicles. I found that SonoAVC with post-processing significantly underestimated and 2D overestimated the number of antral follicles measuring 4.0mm or less, but both made comparable counts of follicles measuring more than 4.0mm when compared with the antral follicles dissected manually. However, the agreement with SonoAVC with post-processing was more than that with 2D. Having established that SonoAVC albeit with post-processing was the most reliable method in measuring the size of antral follicles, I began to examine the intra- and inter-cycle variation and compared to AMH. I found that the small antral follicle measuring 2.0-6.0 mm showed least intra-and inter-cycle variation and that it was comparable to AMH. The larger antral follicles showed significant intra-cycle variation but a non-significant inter-cycle variation in the early follicular phase of menstrual cycle. I also examined the inter-ovarian variation in the AFC’s and found that the small antral follicles measuring 2.0-6.0 mm again showed the least variation between ovaries within an individual. I was finally able to conclude that small antral follicles (≤6.0mm) measured using SonoAVC were the most reliable in prediction of ovarian reserve, and showed excellent correlation with AMH. Finally, I examined the effect of laparoscopic ovarian cystectomy on the ovarian reserve for up to 6 months post-operatively using AMH and small AFC measured by SonoAVC. I found that ovarian cystectomy significantly reduces ovarian reserve and that this effect may be more pronounced with cysts of endometriotic nature, followed by dermoid cysts. In summary, the effect of ovarian cystectomy on ovarian reserve is best quantified using AFC of small follicles measuring less than 6.0 mm as it provides reliable measures of ovarian reserve, has minimal biological variation and is comparable to AMH.

This thesis examines and explores women's mammography screening experiences and the factors influencing women's screening behaviours and choices. The sample consisted of Greek women who were recruited from women's organizations in Athens. 189 women completed a survey questionnaire about their values and beliefs in relation to mammography screening, 33 of whom subsequently underwent an in-depth interview to explore their experiences of mammography screening and their decision-making processes. According to the findings, women's mammography screening behaviour depended to a great extent on the quality and nature of their interactions within their social networks. Thus, the quality of information and meanings derived from these interactions was what determined women's behaviour. Family, close friends and doctors appeared to be important in the formation of women's beliefs, perceptions, emotions (fear of cancer) and behaviour in relation to breast screening. Interestingly, fear of cancer had an opposing effect towards their screening behaviour. A tentative model is presented, which attempts to explain the way these influences affect beliefs, perceptions and mammography screening behaviour, using elements of existing behavioural models. Further exploration of the influential factors and associations identified in this study is required.

The study was conducted following a systematic review of 38 articles (Collins et al 2004), which failed to find strong evidence for the optimum duration and frequency of follow-up, or the personnel best placed to provide it. Against this background of lack of evidence for ongoing follow-up, on the grounds of limited resources and an ever increasing population of breast cancer survivors; the level and nature of service provision for the patients attending a regional breast unit was changed. What was unknown was the impact this may have on the quality of the care and support and future health and well-being. Primary aim: To investigate the experience of being discharged from follow-up breast care services and evaluate the impact of the proposed changes from the patient’s perspective. This thesis reports on interviews with 20 women from the time they were discharged from the service to 2 years post discharge. In addition, 4 participants who developed recurrent disease gave interviews to discuss their experiences. 3 major themes emerged: 1. “Close shave” or “marked woman”? describes the approaches adopted by the women towards their disease, treatment and risk of recurrent disease; 2. A “blessing” or a “curse”? encompasses views of the value of follow-up care; 3. “freedom” or “isolation”? is the account of the impact of discharge from the service on future health, care and well-being. Creating a network of services and working collaboratively with other sectors may well provide the scope and quality of care that women in long-term recovery from breast cancer want and need from the survivorship care of the future.

Breast cancer, which is the most common cancer in women, is a complex disease characterised by multiple molecular alterations. Current routine clinical management relies on availability of robust clinical and pathologic prognostic and predictive factors, like the Nottingham Prognostic Index, to support decision making. Recent advances in highthroughput molecular technologies supported the evidence of a biologic heterogeneity of breast cancer. This thesis is a multi-disciplinary work involving both computer scientists and molecular pathologists. It focuses on the development of advanced computational models for the classification of breast cancer into sub-types of the disease based on protein expression levels of selected markers. In a previous study conducted at the University of Nottingham, it has been suggested that immunohistochemical analysis may be used to identify distinct biological classes of breast cancer. The objectives of this work were related both to the clinical and technical aspects. From a clinical point of view, the aim was to encourage a multiple techniques approach when dealing with classification and clustering. From a technical point of view, one of the goals was to verify the stability of groups obtained from different unsupervised clustering algorithms, applied to the same data, and to compare and combine the different solutions with the ones available from the previous study. These aims and objectives were considered in the attempt to fill a number of gaps in the body of knowledge. Several research questions were raised, including how to combine the results obtained by a multi-techniques approach for clustering and whether the medical decision making process could be moved in the direction of personalised healthcare. An original framework to identify core representative classes in a dataset was developed and is described in this thesis. Using different clustering algorithms and several validity indices to explore the best number of groups to split the data, a set of classes may be defined by considering those points that remain stable across different clustering techniques. This set of representative classes may be then characterised resorting to usual statistical techniques and validated using supervised learning. Each step of this framework has been studied separately, resulting in different chapters of this thesis. The whole approach has been successfully applied to a novel set of histone markers for breast cancer provided by the School of Pharmacy at the University of Nottingham. Although further tests are needed to validate and improve the proposed framework, these results make it a good candidate for being transferred to the real world of medical decision making. Other contributions to knowledge may be extracted from this work. Firstly, six breast cancer subtypes have been identified, using consensus clustering, and characterised in terms of clinical outcome. Two of these classes were new in the literature. The second contribution is related to supervised learning. A novel method, based on the naive Bayes classifier, was developed to cope with the non-normality of covariates in many real world problems. This algorithm was validated over known data sets and compared with traditional approaches, obtaining better results in two examples. All these contributions, and especially the novel framework may also have a clinical impact, as the overall medical care is gradually moving in the direction of a personalised one. By training a small number of doctors it may be possible for them to use the framework directly and find different sub-types of the disease they are investigating.

The research within this thesis concerns the monitoring of the fetus during labour, using the fetal electrocardiogram (FECG). A versatile FECG analysis system was developed for the Microsoft Windows environment, to allow various FECG parameters to be extracted. Algorithms, currently used in other FECG analysis systems, were implemented using Object Oriented Programming, thus allowing new algorithms to be easily added at a later stage. Although these current algorithms have been demonstrated by several authors, it was felt that they had been used with only partial investigation of their limitations, and with failure to fully determine their accuracy in controlled conditions. These factors are fully addressed within this thesis. By developing a FECG simulator, in which heart-rate, morphology, and noise levels could be varied, the ability of the analysis algorithms to extract the parameters, and the accuracy of these parameters under different noise conditions, were thoroughly checked. Both ability and accuracy were shown to be very good in ideal noiseless conditions; but, with the addition of noise, there exists a compromise between parameter accuracy when the morphology is static, and parameter accuracy when the morphology is changing. The accuracies of the most common indices in this field (the Conduction Index, and the T/QRS ratio) were determined for different levels of simulated noise, and their values demonstrated for data previously recorded from the fetal scalp. Errors as large as 0.3 in the CI and 0.05 in the T/QRS suggested that in the clinical environment, an indication of the accuracy of each index ought to be displayed, and this may be estimated from the measured level of noise. Furthermore, this analysis system allows the direct comparison of both indices. Finally, in order to design a more effective front-end filter, it is important to be aware of the frequency content of the underlying FECG. The Integral Pulse Frequency Modulation (IPFM) model, combined with Pulse Amplitude Modulation (PAM), was used to estimate realistic frequency components within the FECG signal. The effects of filtering could then easily be modelled to show the distortion of both the FECG and any parameters taken from it. For a FECG frontend filter, distortion was found to be insignificant provided that, above 1 Hz, both the gain remained constant and there was no phase-distortion.

Introduction: Early diagnosis of breast cancer can result in less radical therapy and improved survival. Current screening and diagnostic tools have limitations, as do serum marker antigens due to their low sensitivity. We hypothesised that an immune response is an early event in cancer evolution. Autoantibodies, which are the amplified signals of cancer-derived antigens, can be detected in the peripheral blood of women with early breast cancer. This thesis is a continuation of previous work at the Nottingham Breast Unit aimed at developing new panel of assays for the detection of autoantibodies in breast cancer. The goal of this thesis was to investigate the use of a potentially more reproducible ELISA assay to measure serum autoantibodies to MUC1, p53 and c-myc either singly or in combination within a panel to further clarify a role of AAbs in screening, diagnosis or prognosis of primary breast cancer. Methods: Newly expressed, biotinylated and reconfigured p53 and c-myc antigens and purified MUC1 antigen were used to establish novel in-house ELISA. These were used to measure autoantibodies to the above 3 antigens in the serum of various populations which were collected over a two year period. These populations included an at-risk population (e.g. family history and atypical ductal hyperplasia) and a population of women who had just been diagnosed with primary breast cancer, either non-invasive ductal carcinoma in situ (DCIS) or invasive cancers. Cut-off values were established for each of the autoantibodies based on 2 or 3 standard deviations from the mean of a population of control samples. The control samples were obtained from a population of women who were either deemed ‘normal’ or who had a histological diagnosis of benign breast disease. The assay was validated by assessing effect of sample age as samples were of varying age, reproducibility using Bland Altman coefficient of reproducibility and reliability by establishing the assays ability to distinguish cancer from non-cancer. Results: Eight hundred and ninety eight samples were analysed in the study. One hundred and ten were Control samples. The remaining samples included 381 that were from an at-risk population and 407 that were from a primary breast cancer population. Mean ages of Control, at-risk and primary breast cancer populations were 58.8, 50 and 62.9 years respectively. Data establishing validity of assay confirmed that sample age did not affect signal strength for MUC1 and c-myc autoantibodies. Older samples for the p53 autoantibody had lower signal than recent ones. Reproducibility data was satisfactory and was best in the samples from the group of women with benign breast disease. Using either a 2 or 3 standard deviation cut-off value the assay was also able to distinguish cancer from non-cancer for both MUC1 and p53 autoantibodies. For the c-myc autoantibody, cancer samples showed increased signal compared to non-cancer although this did not reach significance. The at-risk population were routinely followed up in an outpatient clinic dedicated for women at increased risk of breast cancer. An individual positive marker was noted in up to 10% of at-risk patients. The panel of 3 assays showed a raised marker in 18.4%. This was significantly higher than that for the Control population whose panel detection was 9.1% whilst an individual marker was noted in up to 4.5% of samples. Only the c-myc autoantibody had similar prevalence in both Control and at-risk populations. There was no correlation between risk category and autoantibody detection. The specificity for MUC1, p53 and c-myc autoantibody serum tumour markers were 92.4%, 95.2% and 95% respectively. Specificity of the assay can be further increased if two or more markers were needed to be positive before a positive result is deemed for the assay. Thirteen women in the at-risk group developed breast cancer. The panel had a higher sensitivity to detect occult tumours compared to individual markers but at reduced specificity. Two of 13 at-risk patients (15.4%) who developed breast cancer had a raised marker (MUC1 & p53 autoantibodies) within the panel with a mean lead-time of 43.5 months. Further increasing the cut-off value to Mean + 4 standard deviation of Control population increased the specificity of the panel assay to 97.2% without altering the sensitivity to detect occult tumour (15.4%). Primary breast cancer population consisted of patients who were known to have DCIS or invasive breast cancer. The latter group was further subdivided into those who were detected via screening mammogram (screen-detected) and those who presented with a lump (symptomatic). Two of the 3 markers (p53 and c-myc autoantibodies) were significantly raised in the primary breast cancer population compared to the at-risk population as well as the Control group as detailed in earlier paragraph. Individual markers were detected in up to 20.9%, 10.3% and 9.8% for p53, c-myc and MUC1 autoantibodies respectively. The panel detection rate was 35.1%. The tumour markers showed limited use as a prognostic factor. Only the c-myc autoantibody correlated with a poorer survival due to distant metastasis in symptomatic breast cancers. Data for the screen-detected breast cancer cases showed that there were no correlation between any of the 3 serum marker detection and prognosis. Conclusion: Our data demonstrated the three autoantibody assays whether singly or in combination as a panel showed differences not only between cancer and non-cancer but also between Control and at-risk, as well as between at-risk and cancer. The panel showed that one or more assays were positive in 35% of breast cancers with a specificity of 83.6%. The specificity of the assay can be altered to meet clinical needs by either increasing the cut-off value or altering the markers within the panel. Current data in the literature suggests a number of markers that may be added or substituted into the panel to enhance the specificity and sensitivity. However a sensitivity of 15.4% for detection of occult tumour in the at-risk group makes any clinical application for screening in this group less cost effective using the version of the assays described in this thesis. The lead-time in the two patients who did show elevation of an autoantibody suggests that if the sensitivity and specificity can be improved that there is an in-vivo amplification signal, which might allow earlier identification of some breast cancers. Detection of c-myc autoantibodies indicates a poorer prognosis in the symptomatic group. The value of this information needs to be further determined in larger studies and within multivariate analysis. If the current results remain then there may be clinical implication to this early data. Comparison with previous data from the unit revealed that detection of cancer-associated autoantibodies in primary breast cancer and at-risk groups using this methodology appeared to be less sensitive. This may indicate that the current method has been successful in reducing background signal and hence reduce false positive results. It therefore appears that we have established a more reliable and reproducible assay compared to previous study to detect autoantibodies to tumour-associated antigens. However it is noted that this thesis reports single batches of antigens (MUC1, p53 and c-myc) used in the autoantibody assays. Investigation of differences in protein structure and immunogenicity between batches, which might also affect the sensitivity and specificity of these assays, was outside the scope of this thesis but is the subject of ongoing research by other members of the research group.

Introduction: Heavy menstrual bleeding is known to impact on health, wellbeing and social functioning. There is limited research examining women’s experiences of treatment for this complex condition. This longitudinal qualitative study explored women’s experiences of medical treatment for heavy menstrual bleeding with particular reference to women’s treatment preferences, quality of life and cultural variation. Methods: Data were generated by a series of two semi-structured interviews conducted with women who had either consented to participate in a randomised controlled trial (investigating effectiveness of levonorgestrel intra-uterine system [LNG-IUS] compared with standard medical treatments) or had declined due to an expressed treatment preference. The sample was purposefully selected to include a wide range of demographic characteristics and medical treatment options. Interviews were audiotaped and transcribed verbatim before being coded. A grounded approach to analysis was used with concepts emerging from coded data. Data generation and analysis were iterative and continued until theoretical saturation was reached. Findings: Twenty-seven women were selected and consented to the interview study, ten of whom expressed a treatment preference. A broad range of demographic characteristics and treatment options was achieved. In addition to physical aspects of heavy menstrual bleeding, most women described maintaining the societal norm of concealment of menstruation distressing. Expectations of positive treatment outcomes were high, but the experience of most treatment did not, in the initial stages, meet women’s expectations. After one year, women using LNG-IUS felt less restricted in their daily activities and described feeling less distress about their heavy menstrual bleeding. Many women who were using other treatments had discontinued them, several opting to manage their heavy menstrual bleeding by adapting coping behaviours rather than seeking alternative treatment. Treatment preferences appear to be influenced by knowledge gained from peers with LNG-IUS being the most preferred treatment. For many women in this sample the impact on quality of life was determined by their perceived ability to continue with their roles and responsibilities both at home and in the workplace. A conceptual model was developed when descriptive themes emerging from coded data were overlaid with recorded treatment outcomes. The model illustrates the complexity of heavy menstrual bleeding and highlights common feelings and experiences could be aligned to particular treatment trajectories. Conclusion: This study adds understanding to women’s experiences of medical treatment for heavy menstrual bleeding, preferences for treatment and how quality of life changes over time. The conceptual model developed highlights the complexity of this condition and might be used to improve communication between women and health professionals.

Over recent decades, translational ovarian cancer research has been impeded by its underappreciated molecular heterogeneity and five-year survival has remained poor. One strategy for addressing this problem is to search for molecular biomarkers that can better inform the development and targeting of novel treatments. The aim of this thesis was to construct and validate a tissue microarray of ovarian cancer cases and to survey the expression and prognostic capabilities of immunological molecular markers: specifically HLA class I and the membrane bound complement regulatory proteins CD46, CD55 and CD59. These are central to the efficacy of certain immunotherapies and while they have been shown to have prognostic power in breast and colorectal cancer, they have been investigated less in ovarian cancer. Five copies of a tissue microarray representing 339 cases of ovarian cancer which presented to Derby City General Hospital between 1982 and 1997 were made. The array was stained for CK7, CK20, CA125, CEA, p53 and Bcl-2 following a standard immunohistochemical protocol. A linked clinical database was adapted and assessed for data consistency and subsequently used to analyse the prognostic and clinicopathological associations of the expression data. The array was then stained for HLA class I, B2microglobulin and CD59 using commercial antibodies and for CD55 and CD46 using in-house antibodies. Retained expression of HLA class I molecules independently predicted improved prognosis. High expression of CD55 and CD59 were associated with worse prognosis, though not independently of other factors. CD55 expression was more widespread than previously appreciated. This thesis describes the discovery of a new independent marker of prognosis which suggests that immunoediting occurs in ovarian cancer, describes the distribution of markers known to have a negative impact on immunotherapy in ovarian cancer in a large series for the first time and documents the production of a valuable resource for future studies.

Computational and mathematical models have become increasingly important and have contributed to significant advances in our understanding of complex biological systems. We developed a mathematical model to characterise the expression profile of transmembrane electrogenic proteins of excitable cells. The cell of interest is the myometrium smooth muscle cell, which is the principal unit of electrical activity in the uterus. These cells remain quiescent throughout most of gestation, whereas just prior to and during labour they are able to generate spontaneous action potentials. A more detailed and comprehensive characterisation of these cells, in comparison to previous models, would furnish an appropriate tool for the development of therapeutics to manage preterm birth and other perinatal problems associated with uterine contractility, such as postpartum haemorrhage. The "conductome" can be defined as the totality of ion channels and ion transporters expressed by an electrically active cell, i.e., a list specifying the cell surface density and oligomeric composition of each of these species. Gene expression techniques can accurately survey the complete set of all mRNA species encoding electrogenic proteins (e.g., subunits of channels). The conductome is constrained by this transcriptome, but the link between the two is complicated by the facts that (i) presence of an mRNA species does not necessarily imply the presence at the transmembrane proteomics level; and (ii) subunits can combine in various ways to give rise to conducting channels with different properties. Every individual potential oligomeric channel complex was represented as a mathematical model on the basis of biophysical data taken from the literature; these data were obtained mainly using heterologous expression systems. We investigate the possibility of combining the behavioural information of the action potential with the detailed molecular data of the transcriptome. The general problem is that electrical behaviour does not necessarily lead to a unique solution. The question addressed here is to what extent the additional information provided by transcriptomics helps to constrain the solution space. We develop and apply a method to characterise the functional redundancy of electrically active cells. We use mRNA sequencing to determine which electrogenic species the cell is capable of expressing, combined with a least-squares parameter estimation procedure to determine the conductome from electrophysiological data. Moreover, we estimate the parameters associated with the gating kinetics from published data, so that the only remaining free parameters are the surface densities of the species on the list defined by the transcriptomics analysis.