Objectives: Missing data represent a source of bias in randomized clinical trials (RCTs). This thesis focuses on pragmatic RCTs with missing continuous outcome data and evaluates the use and appropriateness of current methods of analysis. Methods: This thesis consists of three parts. First, a systematic review examined practices relating to missing data in published RCTs. Second, a simulation study compared the performance of various methods for handling missing data in a number of plausible trial scenarios. Finally, an empirical evaluation of two pragmatic RCTs investigated the use of a reminder process to inform whether missingness is likely to be non-ignorable. Results: The majority of 91 trials in the systematic review adopted a form of single imputation, such as last observation carried forward (LOCF) for dealing with missing data. Mixed-effects model for repeated measures (MMRM) and/or multiple imputation (MI) were limited to eight trials. Sensitivity analyses were infrequently and inappropriately used, and insufficiently reported. In the simulation study, LOCF yielded biased estimates of treatment effect in most scenarios, irrespective of missing data mechanisms. All methods, except LOCF, yielded unbiased estimates for scenarios of equal dropout rate and same direction of dropout in both treatment groups. MMRM and MI were more robust to bias than complete-case and LOCF-based analyses. In the empirical study, the evaluation using reminder responses indicated the possibility of biased MMRM estimation in one trial and unbiased MMRM estimation in the other. Conclusion: CCA and LOCF-based analysis should be disregarded in favour of methods such as MMRM and MI-based analysis. The proposed reminder approach can be used to assess the robustness of the missing at random (MAR) assumption by checking expected consistency in MAR-based estimates. If the results deviate, then analyses incorporating a range of plausible missing not at random assumptions are advisable, at least as sensitivity tests for the evaluation of treatment effect.

This study was designed to identify and address the support needs of pre-registration nursing students during practice placements in the West Midlands region. The study used an action research approach to define student needs and develop a flexible tool for modelling integrated placement support. To identify student needs, a two-round Delphi technique was used. The Delphi panel was recruited from a range of national and local stakeholders (n= 21). Round 1 used an open-ended, seven-item questionnaire to generate a range of perspectives. The round 1 response rate was 100%. Data were analysed using content analysis, resulting in the development of 46 key needs statements. The needs statements were presented back to the panel in a 46-item Likert questionnaire seeking their level of agreement, or disagreement, with each statement. The round 2 response rate was 76.2%. Round 2 generated a consensus definition of student needs across four domains: student centred; knowledge centred; assessment centred and quality centred. Consensus was not achieved in relation to ‘who’ had responsibility for supporting these needs. Using soft systems conventions, an ‘ideal model’ of support was conceptualised. To address the lack of agreement on the ‘who’ questions, the conceptual model was used to design a flexible modelling framework and tool. This enabled local stakeholders to develop and implement an enhanced, integrated model of support with clear roles and responsibilities. The tool was piloted successfully, using a modified consensus conference approach, in two localities. The transferability of the tool was validated at a national stakeholder workshop using electronic, anonymous voting handsets. The study provides new insights into the needs of student nurses during practice placements and has generated a unique modelling tool that enables stakeholders to address these needs. The findings are relevant for policy-makers, commissioners of pre-registration nurse education, HEIs and health service provider organisations.

Musculoskeletal (MSK) pain is a major reason why people consult their general practitioner. Analgesia plays a central role in its treatment but do not always work, resulting in the need to switch amongst analgesia potency levels. Stronger analgesia is however associated with increased adverse effects. The aim was to investigate the use of robust statistical approaches to determine socio-demographic and clinical factors associated with receiving and switching, prescribed analgesia in primary care management of MSK pain. The first phase reviewed statistical methods previously used in modelling medication switching, and established that Cox proportional hazards and logistic regression models were predominantly used. The second phase investigated the prevalence of prescribed analgesia, factors associated with being prescribed analgesia, and prescription patterns in the management of new MSK conditions using a general practice database. In 3236 incident consulters, 42% were prescribed analgesia, NSAIDs being most prescribed. In a 5 year follow-up period, three prescription patterns were identified: no analgesia or basic analgesia only, use of NSAIDs, and multiple-potency analgesia combinations. The main baseline factors associated with being prescribed analgesia, and stronger analgesia were increasing age and having been previously prescribed analgesia. The third phase used Cox and Weibull frailty models to identify factors associated with switching analgesia and switching to stronger analgesia. The main factors identified were age, gender and initially prescribed analgesia. The fourth phase used a prevalent cohort of 1610 patients aged 50+ with linked self-reported and medical record data. Patient-reported factors such as level of physical function and pain interference were also associated with switching of analgesia. Using a propensity score approach to modelling outcomes suggested those who switched analgesia did not have better three year outcomes, but further research is required to establish if switching analgesia is beneficial in reducing pain and improving function.

The aetiology of widespread musculoskeletal pain is complex. Psychological stress is a robust predictor of symptom onset and persistence but not everyone who experiences stress goes on to develop widespread pain. The aim of the studies presented in this thesis was to ascertain whether individuals with a history of trauma have an increased susceptibility to widespread pain when they experience psychological stress; to identify psychosocial mediators of the stress pain relationship, and ascertain whether these mediators differ, i.e. are moderated by, the experience of prior trauma and by sex. The trauma diathesis stress model of widespread pain, developed by the author, was assessed using structural equation modelling on data collected by two population-based prospective studies. In the General Practice Symptom Survey (GPSS), 1,443 adults aged 25–65 years provided data on the number of pain sites, psychological stress and childhood abuse. In the North Staffordshire Osteoarthritis Project (NorStOP), 6,678 adults aged 50–90 years provided data on the number of pain sites and psychological stress, whilst the occurrence of surgeries, fractures, RTAs and burns was obtained from their medical records. Higher levels of psychological stress were associated with a higher number of pain sites. The stress pain relationship was moderated by childhood abuse but not by adult physical trauma. The relationship between stress and pain was mediated by attachment style (GPSS) and by social support (NorStOP). This research explored the moderators (in whom) and mediators (how) of the stress pain relationship. Childhood abuse was identified as a susceptibility factor and adult attachment style and social support as the processes by which stress leads to pain. These findings have implications for both primary and secondary prevention; suggesting that a stratified treatment approach may be most appropriate.

Patients with chronic knee pain (CKP) frequently present to general practitioners (GPs). Exercise, a core management approach for CKP, reduces pain and improves functioning. To maximise patient outcomes, GPs should practise in line with best evidence recommendations. Using an underpinning model (developed using behavioural theory), this thesis describes the attitudes, beliefs and behaviours of GPs regarding the use of exercise for patients with CKP. A systematic review revealed a paucity of published studies specifically examining this topic. Available data suggested that GPs’ attitudes and beliefs about exercise for CKP varied widely, exercise appeared to be underused and its implementation by GPs was unclear. The need to concurrently and specifically investigate the attitudes, beliefs and behaviours of GPs regarding exercise for CKP was identified. A vignette-based pilot questionnaire survey of 800 UK GPs was undertaken to refine the survey tool and methods and to inform the required sample size for the main survey. The subsequent main survey of 5000 UK GPs revealed that exercise was used by most GPs for CKP. However, methods employed to initiate exercise within an individual patient’s management plan were variable and imperfectly aligned with evidence-based recommendations. Attitudes and beliefs about exercise for CKP were generally positive; however GPs expressed some uncertainty about safety and efficacy, particularly regarding local exercise (e.g. strengthening, range-of-movement, stretching). Although some elements of the underpinning model (e.g. role and identity) predicted GPs’ behaviour, others (e.g. beliefs about capabilities) performed less well. To maximise the clinical outcomes of patients with CKP, recommendations from this research include: development of educational, organisational change and/or behaviour change strategies to improve initiation of individualised exercise, and clarification of GPs’ role, in this context. Approaches to better understand the key influences on GPs’ behaviour are required; a greater focus on decision-making theory may be valuable.

Knee pain in older adults is common and often disabling, with the majority of knee pain in adults over the age of 45 being attributed to osteoarthritis (OA). Regular physical activity and exercise are recommended for all older adults with knee pain and are associated with reduced pain and improved function. However, physical activity levels are low in this population and there is uncertainty regarding its long-term safety, whether change in physical activity level is associated with future pain and function, and the relationship between attitudes and beliefs about physical activity and physical activity level. This thesis addressed these research questions. A systematic review of safety outcomes from 49 published studies found exercise was safe for the majority of older adults with knee pain, although most evidence related to low impact, moderate cardiovascular intensity exercise. Secondary data analysis of an exercise randomised controlled trial for older adults with knee pain (n=514) did not find an association between change in physical activity level between baseline and three months and clinical outcome at either three or six months. Secondary cross-sectional data analysis, using baseline data from the same trial and a community survey of older adults with knee pain (n=611), found that a number of scales measuring attitudes and beliefs about physical activity were associated with physical activity level in multivariable models. Positive outcome expectations, self-efficacy for exercise, kinesiophobia and a composite scale measuring physical activity attitude themes were associated with physical activity level. Further longitudinal analysis from the trial showed that positive outcome expectations and self-efficacy for exercise remained associated with future physical activity level at three and six months within multivariable models whilst negative outcome expectations were not. The original thesis findings have contributed to a better understanding of attitudes, beliefs and physical activity in older adults with knee pain.

This PhD project was nested within the Psychological Workstream of the Spinal Pain Programme,a National Institute for Health Research (NIHR) funded programme of work (Grant code RPPG-0707-10131) obtained at the Research Institute for Primary Care and Health Sciences by Elaine Hay and colleagues. The idea for this project was devised following discussion of the development of an intervention to target illness perceptions to improve functional outcomes for patients with low back pain. Previous Research Institute work had focused on prognostic studies and evidence had been gathered on what factors influenced patient outcome, but little work had been carried out to investigate whether these same factors could also be effectively targeted during treatment. This PhD project was devised to address this gap in knowledge. Throughout the course of this PhD project I have developed the ideas in this thesis under the guidance of my supervisors (Jonathan Hill, Daniëlle van der Windt and Chris Main (Chris took over from Kevin Vowles, who supervised me until August 2012)). My supervisors advised on the planning of all the analyses presented and on the writing and structure of the included chapters. I designed the analysis plans, conducted all analyses and wrote all chapters. I received guidance on search strategies from Joanne Jordan and Nadia Corp (Chapter 4), and the critical appraisal of the studies included in Chapter 4 was aided by Jemma Cowen. Elaine Thomas provided statistical guidance on Chapters 5 and 6 and Kelvin Jordan and Anne Smith provided guidance on the Latent Growth Modelling presented in Chapter 7. All of the data presented in this thesis was collected prior to my appointment at the Research Institute. I was provided with cleaned datasets from Nadine Foster (BeBACK data presented in Chapter 5), Jonathan Hill (STarT Back data presented in Chapter 6) and Michael Von Korff (Back In Action data presented in Chapter 7). Data from the IMPACT service (Chapter 8) was provided by Julie Ashworth, which did require cleaning by myself prior to use.

Damaged or diseased mature articular cartilage cannot undergo effective tissue repair and due to its avascular, hypocellular nature defects become widespread and painful. No ‘gold standard’ treatment exists for this indication and the ultimate recourse is prosthetic joint replacement. Cartilage is therefore an ideal target for regenerative medicine therapies aiming to recapitulate native cartilage. Despite over fifty years of research and encouraging outcomes, re-creation of the hyaline tissue has yet to be consistently achieved, possibly as a result of the application of a sub-optimal cell type. Chondrocytes and bone marrow mesenchymal stem cells (MSCs) have been used clinically, with future prospects for other alternative MSC sources and human embryonic stem cell (hESC)-derived cells. Further in vitro study of cellular chondrogenic capacity is desirable but hampered by cell changes and senescence. This work examines the hypothesis that the re-introduction of the catalytic sub-unit human telomerase reverse transcriptase (hTERT) can extend the proliferative cell capacity of cells whilst concomitantly bypassing changes associated with cell aging and senescence. The utility of umbilical cord blood (UCB) as a possible alternative source of more naive MSCs was also investigated. Human bone marrow MSCs, chondrocytes, and hESC-derived cells were transduced with hTERT and their resulting chondrogenic capacity, assessed principally by extracellular matrix (ECM) production and gene expression, examined and compared to that of the three non-transduced, parental cell sources. UCB was not found to be a viable alternative MSC source due to a very low cell number and colony recovery; however, foetal bovine serum (FBS) batch and atmospheric oxygen tension were identified as key to influencing recovery outcomes. Of the three parental cell types examined for chondrogenic potential MSCs and chondrocytes produced similar amounts of sGAG but chondrocytes produced a more homogeneous ECM with persistent chondrogenesis, whereas MSCs became hypertrophic. hESC derived cells had a more muted chondrogenic response with similarities to both chondrocytes and MSCs. TERT extended the proliferative capacity of all three cell types, two extensively but was also associated with changes in cell phenotype and a reduction, although not complete ablation, in the subsequent chondrogenic capacity. Taken together the results demonstrate that with current differentiation techniques primary articular chondrocytes provide the most optimal result, supporting their continued use for clinical therapies, and this capacity may not be preserved by the application of hTERT transduction strategies.

COPD is a debilitating chronic respiratory disease with a systemic footprint. COPD is a highly heterogeneous disease but invariably its management involves a one-size fits all approach. This fails to address individual variations in disease progression, symptom burden and functional decline. There remains a need for sensitive monitoring tools that could provide personalised care based on patients’ particular phenotypes and informative self-management. This thesis has standardised collection protocols and processing for saliva, a complex bodyfluid which is readily accessible and user-friendly for near-patient testing. I have modified immunoassays to work in saliva with demonstrable reproducible results for quantification of C-reactive protein, Procalcitonin and Neutrophil Elastase. Symptom assessment is crucial in longitudinal self-monitoring of COPD. I designed a novel patient wellbeing scale incorporated into an electronic self-assessment diary, which was embraced by patients as improving symptom change recognition, education and self-management. Using sophisticated analytical tools, I have attempted to cluster/phenotype disease trajectory paths driven by a compilation of symptom scores, spirometric volumes and saliva biomarker levels and produced novel patientspecific multidimensional composite scores with significant correlation to COPD disease severity. Prodromal changes in FEV1, salivary biomarkers and self-assessment scores were reproducibly demonstrated, with potential to predict exacerbation onset. These results could be exploited for the development of a much-needed personalised COPD monitoring eco-system, which isolates early deteriorations and prompts timely interventions, leading to beneficial disease outcomes. Patient-researcher iterative co-design has been key throughout this thesis. One outcome of this relationship is the design and production of a bespoke integral saliva collector prototype which could substitute laboratory-based processing of saliva samples in readiness for analyte testing. In conclusion, this thesis has created the necessary tools to improve the classification and monitoring of COPD, opening new avenues for proactive patient self-management and providing the basis for future personalised and stratified care.

Mycoplasmas represent some of the smallest and simplest free-living organisms known. Mycoplasma hominis and Mycoplasma pneumoniae are two human pathogens that colonise the urogenital and respiratory tracts, respectively, causing a diverse range of disease. Detection of Mycoplasma hominis is hampered by the fastidious nature and genetic heterogeneity of this organism. Characterisation of mycoplasmas is becoming more important due to increasing antibiotic resistance, particularly in M. pneumoniae, and the need for more discriminatory methods to enhance epidemiology and examine transmission chains. Firstly, this thesis develops a quantitative, multiplex, real-time PCR assay to simultaneously detect M. hominis and Ureaplasma species in neonatal clinical specimens, where infection is associated with chronic lung disease, bacteremia and other clinical signs. Results showed that the PCR method was clinically more sensitive than culture and has applications for monitoring bacterial load in clinical specimens and characterising bacterial response to antibiotics. Secondly, genetic characterisation of M. hominis was undertaken by the examination of the variable adherence-associated antigen and the development of sequence based typing. Due to the genetic heterogeneity of M. hominis, bioinformatics analysis of genomic sequence was used as a novel method to develop a minimum multi-locus sequence typing (MLST) scheme that accurately represented genomic phylogeny of this species. Finally, an MLST scheme was developed for M. pneumoniae, to aid the analysis of epidemic periods and clusters of infection. A successful scheme was developed based on eight housekeeping genes which had increased discrimination of M. pneumoniae compared to established typing methods for this organism. Furthermore, the MLST scheme was found to be representative of genomic sequence-derived phylogeny, with two distinct genetic clades identified. Application of this MLST to UK epidemics revealed that no predominant sequence type was responsible for the epidemic periods studied, indicating a polyclonal population, supporting the hypothesis that epidemics are driven by population immunity.