No description available.

Introduction Colorectal cancer is a cancer that forms in the tissues of the colon and/ or rectum and more than 95% of colorectal cancers are adenocarcinomas. It is the third most common cancer in incidence and mortality rates, accounting for 9% of all cancer cases and for 8% of all cancer related deaths (2002). The established risk factors of colorectal cancer include personal or family history of previous colorectal cancer or adenomatous polyps, chronic bowel inflammatory disease and presence of any of the hereditary syndromes. In addition, due to the fact that the majority of colorectal cancer cases (approximately 90%) occur after the age of 50, advanced age is also considered as a risk factor. Finally, evidence for significant associations between colorectal cancer and other risk factors, including diet, body weight, physical activity, smoking, alcohol intake, NSAIDs intake and HRT in post-menopausal women, is promising and increasing. Aims and objectives The main aims of this project were: 1) to investigate the associations between colorectal cancer and specific nutrients, including flavonoids, fatty acids, folate, vitamin B2, vitamin B6, vitamin B12, alcohol, vitamin D and calcium (prior hypotheses 1-4) and 2) to conduct an overall as well as forward and backward stepwise regression analyses of demographic, lifestyle and dietary risk factors. Methods The analysis of this thesis was based on a population-based case-control study of colorectal cancer (Scottish Colorectal Cancer Study; SOCCS). In total 3,417 colorectal cancer cases and 3,396 controls were recruited in the study. Dietary and lifestyle data were collected by two questionnaires (Lifestyle & Cancer and Food Frequency Questionnaire) and were available for 2,061 cases and 2,776 controls. For the analysis of the first two hypotheses (flavonoids and fatty acids) a matched dataset of 1,489 casecontrol pairs was used and conditional logistic regression models were applied, whereas for the analysis of the last two hypotheses (folate, vitamin B2, vitamin B6, vitamin B12, alcohol, vitamin D and calcium) an unmatched dataset including 2,061 cases and 2,776 v controls was used and unconditional logistic regression models were applied. For the overall and stepwise regression analyses the unmatched dataset was used (2,061 cases and 2,776 controls). Forward and backward stepwise regression was applied on three different sets of variables and the stability of the resultant models was checked in 100 bootstrap samples. Results Regarding the first two hypotheses, statistically significant odds ratios (ORs) (matched on sex, age and health board are and adjusted for family history of cancer, BMI, physical activity, smoking, and intakes of total energy, fibre, alcohol and NSAIDs) for highest versus lowest intakes (quartiles) were observed for flavonols OR (95% CI), p-value for trend: 0.78 (0.60, 0.99), 0.08) and for the individual flavonoid compounds quercetin and catechin (OR (95% CI), p-value for trend: 0.77 (0.60, 0.99), 0.04; 0.75 (0.58-0.97), 0.02; respectively); for the 3PUFAs fatty acids (OR (95% CI), p-value for trend: 0.75 (0.59, 0.97), 0.01) and for the individual fatty acids stearic acid, EPA and DHA (OR (95% CI), p-value for trend: 1.46 (1.11, 1.91), 0.01; 0.74 (0.58, 0.95), 0.02; 0.74 (0.58, 0.95), 0.02; respectively). Regarding the last two hypotheses, statistically significant odds ratios (ORs) (adjusted for age, sex, deprivation score, family history of cancer, BMI, physical activity, smoking, and intakes of total energy, fibre, alcohol and NSAIDs) for highest versus lowest intakes (quartiles) were observed for vitamin B6, vitamin B12 and alcohol (OR (95% CI), p-value for trend: 0.86 (0.72, 1.03), 0.08; 0.80 (0.67, 0.97), 0.05; 0.83 (0.68, 1.00), 0.03); and for vitamin D (OR (95% CI), p-value for trend: 0.83 (0.69, 0.99), 0.03). Regarding the second aim of the project, several risk factors were found to be significantly associated with colorectal cancer in the overall analysis including demographic and lifestyle factors (family history of cancer, NSAIDs intake, dietary energy intake, HRT intake and physical activity), food group variables (vegetables, eggs, sweets, fruit/ vegetable juice, oily fish, coffee, fruit, savoury foods and white fish) and nutrient variables (tMUFAs, 3PUFAs, SFAs, tFAs, MUFAs, quercetin, catechin, phytoestrogen, cholesterol, fibre, protein, starch, magnesium, potassium, manganese, copper, iron, zinc, phosphorus, selenium, niacin, vitamin B6, carotenes, vitamin C, vi vitamin A, potential niacin, biotin, folate, pantothenic acid, vitamin D, vitamin B1 and vitamin B12). In addition, the variables that were selected to be included in 100% of the models after applying forward and backward stepwise regression analyses were family history, NSAIDs, sweets and fruit/ vegetable juice. Finally according to the findings from the bootstrap analysis, the variables that were selected to be included in models for the majority of the bootstrap samples (more than 90%) were family history, NSAIDs, dietary energy, eggs, sweets, fruit/ vegetable juice and white fish. Discussion The particular dietary factors that were found to be inversely associated with colorectal cancer after applying several multivariable logistic regression models were: flavonols, quercetin, catechin, 3PUFAs, EPA, DHA, vitamin B6, vitamin B12 and vitamin D. In addition, high intakes of stearic acid were found to be positively associated with colorectal cancer. In contrast, high intakes of dietary and total folate were associated with a decreased colorectal cancer risk in the energy-adjusted model, but this inverse association was attenuated after further adjustment for several confounding factors including fibre. Regarding alcohol intake, when it was divided into quartiles, high alcohol consumption was associated with a statistically significant and dose-dependent decreased colorectal cancer risk. However, when alcohol intake was divided in categories an increased colorectal cancer risk for intakes of higher than 60 g/day was observed. Intakes of 3PUFAs, vitamin D and vitamin B12 were highly correlated due to having the same food source (oily fish) and therefore it is difficult to draw specific conclusions regarding which nutrient is truly associated with colorectal cancer and which not. Finally, it was observed that for calcium intakes to be inversely associated with colorectal cancer, a dosage of 1500mg/day or higher was necessary. The majority of these results are in accordance with results of previous epidemiological and laboratory studies; however their confirmation in further large-scale studies is required. Results from the overall and stepwise regression analysis supported previous findings of an increased colorectal cancer risk due to a high or moderate family history risk. In addition, high intakes of dietary energy were found to be positively associated with increased colorectal cancer risk in the overall analysis and in addition dietary energy was vii selected to be included in the majority of the stepwise regression models. On the other hand, regular intake of NSAIDs was found to be inversely associated with colorectal cancer risk in the overall analysis and in the majority of the stepwise regression models. Finally, the overall and stepwise regression analyses generated a few new hypotheses suggesting that low intakes of fruit/ vegetable juice, eggs, white fish and sweets (a combined variable of high-fat and high-sugar foods) and high intakes of coffee and magnesium were associated with a decreased colorectal cancer. These findings, though interesting and important for generation of new hypotheses, need further investigation (as prior hypotheses) in large-scale observational studies.

This thesis addresses the problem of information hiding in low dimensional digital data focussing on issues of privacy and security in Electronic Patient Health Records (EPHRs). The thesis proposes a new security protocol based on data hiding techniques for EPHRs. This thesis contends that embedding of sensitive patient information inside the EPHR is the most appropriate solution currently available to resolve the issues of security in EPHRs. Watermarking techniques are applied to one-dimensional time series data such as the electroencephalogram (EEG) to show that they add a level of confidence (in terms of privacy and security) in an individual’s diverse bio-profile (the digital fingerprint of an individual’s medical history), ensure belief that the data being analysed does indeed belong to the correct person, and also that it is not being accessed by unauthorised personnel. Embedding information inside single channel biomedical time series data is more difficult than the standard application for images due to the reduced redundancy. A data hiding approach which has an in built capability to protect against illegal data snooping is developed. The capability of this secure method is enhanced by embedding not just a single message but multiple messages into an example one-dimensional EEG signal. Embedding multiple messages of similar characteristics, for example identities of clinicians accessing the medical record helps in creating a log of access while embedding multiple messages of dissimilar characteristics into an EPHR enhances confidence in the use of the EPHR. The novel method of embedding multiple messages of both similar and dissimilar characteristics into a single channel EEG demonstrated in this thesis shows how this embedding of data boosts the implementation and use of the EPHR securely.

Background Intimate partner violence (IPV), including physical, sexual and emotional violence, causes short and long term ill-health. Brief questions that can identify women from different ethnic groups experiencing IPV who present in clinical settings are a prerequisite for an appropriate response from health services to this substantial public health problem. Aim: To examine the evidence for the validity of questions trying to identify IPV in different ethnic groups and to determine whether their validity varies between ethnic groups. Methods Design: A systematic review and the secondary data analysis of a cross-sectional survey of four questions (HARK) identifying IPV in a primary care sample. Main outcome measures: Systematic review - for each set of index questions identified, diagnostic accuracy indices, correlation coefficients, reliability measures, validity evidence based on response processes and test content were analysed and interpreted. Secondary data analysis - diagnostic indices for IPV and its dimensions in three ethnic groups were calculated for the four HARK questions combined and for the individual HARK questions. 4 Results Systematic review – there is no evidence of questions valid for identifying IPV in specific ethnic groups, including white groups. Secondary data analysis - the optimal HARK cut off score of ≥ 1 was unaffected by the participants‟ ethnicity. The diagnostic indices generated using the HARK cut off of ≥ 1 remained at a high level, in all three ethnic groups. There were no significant ethnic differences in the diagnostic indices of the four combined and individual HARK questions‟ ability at identifying either IPV or its dimensions. Conclusion From the systematic review and secondary data analysis, there is no evidence that questions‟ validity for identifying IPV varies significantly between different ethnic groups. The secondary data analysis does provide evidence that four questions (the HARK) can identify IPV in self-classified UK census categories of African- Caribbean, south Asian, and white groups.

Trials for comparing I treatments with a control are considered, where the aim is to identify one treatment (if at least one exists) which is better than control. Tests are developed which use all of the data simultaneously, rather than combining separate tests of a single arm versus control. The null hypothesis H0 : i 0 is tested against H1 : i > 0 for at least one i, where i represents the scaled di erence in response between treatment i and the control, i = 1; : : : ; I, and, if rejected, the best treatment is selected. A likelihood ratio test (LRT) is developed using order restricted inference, a family of tests is de ned and it is shown that the LRT and Dunnett-type tests are members of this family. Tests are compared by simulation, both under normality and for binary data, an exact test being developed for the latter case. The LRT compares favourably with other tests in terms of power and a simple loss function. Proportions of subjects on the control close to ( p I 1)=(I 1) are found to maximise the power and minimise the expected loss. Two-stage adaptive designs for comparing two experimental arms with a control are developed, in which the trial is stopped early if the di erence between the best treatment and the control is less than C1; otherwise, it continues, with one arm if one experimental treatment is better than the other by at least C2, or with both arms otherwise. Values of the constants C1 and C2 are compared and the adaptive design is found to be more powerful than the xed design. The new tests can make a contribution to improving the analysis of multi-arm clinical trials and further research in their application is recommended.

Introduction It has been said that through monitoring of drug safety, pharmacovigilance (PV) systems have been instrumental in assisting regulatory decisions on product safety. However, there has been no, systematic, in-depth study of this role. This thesis reports such a study conducted in the UK. On the basis of the results, suggestions are made on how PV data might be produced and used more effectively. Methods In Phase 1, a scoping study was conducted to document all changes made to UK product labelling on safety grounds over a 10 year period (September 1st 1995 to August 31st 2005). In Phase 2, all product withdrawals and major labelling changes made during the 10 year study above, were investigated in depth to determine the therapeutic group, source of ADR data cited as the reason for the change; and product survival probability, using Kaplan-Meier modelling. Phase 3, informed by Phases 1 and 2, used a web-based survey (150 respondents) and structured interviews (13 subjects) with healthcare professionals and scientists with a PV role in the NHS, pharmaceutical companies and the UK regulator, to gain views on the current procedures for handling safety issues in the UK and how these might be improved. Inferences were drawn using interpretative henomenological analysis with NVivo 8 software. Key findings Phases 1 and 2 revealed the fragmentary nature of information in the public domain and the difficulties of obtaining unpublished information. Based on public information, Phase 1 showed that 2,630 safety notices were issued affecting 688 individual products. The two main safety notice categories were drug interactions (841;32%) and side effects (537;20%). The rank order of the four most common therapeutic areas in which safety notices occurred was: CNS (23.5%)> anti-infectives (21.6%) > cardiovascular (15.2%) > cancer chemotherapy (10.8%). The ratio of Type A : Type B side effects (ADRs) was 1:3.3. Phase 2 found that of 518 eligible products launched during the study period, 9 (1.7%) were licensed and withdrawn for safety reasons. The ten-year Kaplan-Meier probability of adverse drug reactions causing the withdrawal of a new product, postmarketingwas 2.2%. All decisions were based on more than one safety data type and all involved UK yellow cards. One decision considered prescription event monitoring (PEM) data. A total of 164 important safety notices affecting 818 individual products were identified. Of 518 products launched during the study period, 56 experienced at least one major labelling change for safety reasons. The ten-year Kaplan-Meier risk of a product experiencing at least one major labelling change on safety grounds was 13.8%. As with product withdrawals, safety decisions were based on a wide range of data sources of variable quality and quantity. Variation in dissemination of the new safety information was observed. Only one fifth of safety notices warranting a ‘Dear Healthcare Professional’ letter or a monograph in ‘Current Problems in Pharmacovigilance’, were accompanied by a boxed warning in the BNF, representing an important inconsistency in notifying prescribers. As with interview participants, respondents to the on-line questionnaire had difficulties placing the yellow card reports in a formal hierarchy of evidence whilst acknowledging that the data were valuable in the decision making process. Suggested ways of improving the quality of such reports included making the reporting more accessible and training all those eligible to report. PEM studies were cited by the majority of respondents as a means of generating credible safety data and raising the general quality of the drug safety database. In terms of dissemination and education about ADRs, Drug Safety Updates (which replaced the ‘Current Problems’ publication from the MHRA in August 2007) were highly thought of; they appeared to be more popular than ‘Dear Healthcare Professional’ letters and because they were web-based, ought to be accessible by a wider audience. Conclusions Safeguarding public health is of utmost importance when making a decision whether or not to withdraw a product or amend its labelling upon the emergence of new safety data. Labelling changes should be made only on the best evidence available at the time and appropriate risk management strategies should be instigated where feasible; not only when a safety signal arises post-marketing, but when a drug is first granted a marketing authorisation. There is no general consensus on what constitutes ‘best evidence’ and rating evidence using traditional hierarchies is problematic, The GRADE hierarchy may be an exception. Improving ADR reporting should lead to improved data bases from which to draw safety conclusions. Methods of improving reporting include early instigation and enforcement of risk management plans by the regulator, education of all those eligible to report, greater transparency of regulatory decisions and better and more rapid dissemination of safety change information.

Background: Despite public health campaigns and improvements in healthcare, socioeconomic gradients in health and life expectancy persist, and in many cases are becoming more marked – the gradient in coronary heart disease being a prime example. Classic cardiovascular risk factors (e.g. smoking, cholesterol and blood pressure) only partially explain the deprivation effect, and attempts to narrow the health gap by focussing on such risk factors do not appear to be succeeding. There also appear to be socioeconomic differences in uptake of healthy lifestyle advice. The work described in this thesis aimed to expand current understanding of the deprivation-based gap in health and life expectancy, focussing particularly on the socioeconomic gradient in cardiovascular risk. Methods: Using a cross-sectional, population-based study design based in the Greater Glasgow area, 666 participants were selected on the basis of area-level social deprivation (Scottish Index for Multiple Deprivation ranking). The study was designed to include approximately equal numbers from most deprived and least deprived areas; equal numbers of male and female participants and equal numbers of participants from each age group studied (35-44; 45-54 and 55-64 years). Participants completed an extensive questionnaire on health, lifestyle and early life experiences. Anthropometric measures (height, leg length, weight, waist, hip and thigh circumferences) were recorded. Blood pressure, heart rate and parameters of lung function (Forced Expiratory Volume in 1 second [FEV1] and Forced Vital Capacity [FVC]) were recorded. Psychological assessments (General Health Questionnaire-28, Generalised Self-Efficacy Scale, Sense of Coherence Scale, Beck Hopelessness Scale, Eysenck Personality Scale and Rosenberg Self-Esteem Scale) and assessments of cognitive function (Auditory Verbal Learning Test, Choice Reaction Time and Stroop Test) were undertaken. Fasting blood samples were obtained for classic and emerging cardiovascular risk factors including lipid profile, glucose, insulin, leptin, adiponectin, C-reactive protein, interleukin-6, soluble intercellular adhesion molecule-1, von Willebrand Factor, fibrinogen, D-dimer and tissue plasminogen activator antigen. Carotid ultrasound assessment of intima-media thickness (cIMT), plaque score and arterial stiffness was performed. Results: Total and low density lipoprotein cholesterol were significantly higher in the least deprived group (both p<0.0001). Triglycerides were higher and high density lipoprotein cholesterol lower in the most deprived group (both p<0.0001). Fasting glucose, insulin and leptin were higher in the most deprived group. C-reactive protein, interleukin-6 and soluble intercellular adhesion molecule-1 were higher in the most deprived group (all p<0.0001). Von Willebrand factor, fibrinogen and D-dimer were higher in the most deprived group. Age- and sex-adjusted cIMT was significantly higher in the most deprived group, but on subgroup analysis this difference was only apparent in the highest age tertile in males (>56.3 years). Plaque score showed a much more highly significant deprivation difference in the group as a whole (p<0.0001). No differences in parameters of arterial stiffness were found between the most deprived and least deprived groups. Neither adjustment for classic nor emerging cardiovascular risk factors, either alone or in combination, abolished the area-level deprivation-based difference in plaque presence or cIMT. Adjustment for early life markers of socioeconomic status in addition to classic cardiovascular risk factors abolished the deprivation-based difference in plaque presence. Further associations between early life factors and health outcomes were noted: lung function (FEV1) and cognitive performance appeared to be influenced by father’s occupation, whether the parents/guardians were owner-occupiers or tenants, and by degree of overcrowding; cIMT was modestly related to father’s occupation and carotid plaque was related strongly to father’s occupation and parental home status. Socioeconomic differences were noted in the impact of personality in determining mental wellbeing, and also in relation to the health behaviours of fruit and vegetable consumption and smoking cessation. Conclusions: The relationship between social deprivation and health is complex and multifactorial and appears to involve the interplay of early life factors, biological mediators, psychological parameters such as personality and cognitive function, health behaviours and outcomes such as atherosclerosis. Approaches aiming to narrow the deprivation gap in health will need to be designed to take into account this complexity, addressing factors such as early life experiences and personality, as well as the more classically recognised factors such as smoking, cholesterol and blood pressure, if they are to have a chance of succeeding in improving the health of those most in need.

Socioeconomic deprivation (SED) is inversely associated with mortality. The most deprived are at a higher risk of all cause mortality and cardiovascular mortality. However, only limited study of the relationship between SED and non-fatal cardiovascular disease has been previously undertaken. In those studies that have examined the relationship between SED and non-fatal cardiovascular disease, analyses have been limited to one form of cardiovascular disease (CVD), such as myocardial infarction or stroke and often prevalent disease. Furthermore, these studies have often failed to examine the association between SED and CVD whilst adjusting analyses for cardiovascular risk factors which are more prevalent in the most deprived. The aim of this work was to examine the association between SED and a number of cardiovascular outcomes after adjusting for the traditional cardiovascular risk factors of age, sex, smoking, blood pressure, diabetes mellitus and cholesterol. To determine is SED is in fact a risk factor for CVD after adjustment for these other risk factors, the relationship between SED and a number of fatal and non-fatal cardiovascular outcomes was examined. A number of forms of CVD were examined, including all coronary heart disease, myocardial infarction, stroke and heart failure A cohort of over 15,000 men and women who participated in the Renfrew Paisley cohort study was examined. These individuals were enrolled between 1974 and 1976 and underwent comprehensive screening for cardiorespiratory risk factors. They have since been followed for hospitalisations and deaths for 28 years. SED was measured using the Registrar General’s social class system and the Carstairs Morris index of deprivation. Rates of fatal and non-fatal outcomes were calculated, as were a number of composite outcomes. Adjusted analyses using multivariable regression were conducted to account for the risk factors of age, sex, smoking, blood pressure, diabetes and cholesterol. Further adjustment for the risk factors of lung function as measured by forced expiratory volume in 1 second, cardiomegaly on chest x-ray, body mass index, and a history of bronchitis was also made. The association between SED and the risk of recurrent cardiovascular hospitalisations, the burden of cardiovascular disease, as well as mortality and premature mortality was assessed for SED. I found that SED was associated with higher rates of hospitalisation for CVD disease in men and women irrespective of the measure of SED, either social class or the area based score of the Carstairs Morris index. This association persisted after adjustment for the traditional cardiovascular risk factors of age, sex, smoking, systolic blood pressure and diabetes and cholesterol. Further adjustment for lung function, the presence of bronchitis, body mass index and cardiomegaly on a chest x-ray did not explain the relationship between SED and each outcome. This risk was long lasting and persisted to the end of follow up. The strength of association of SED with coronary heart disease, myocardial infarction and stroke and all cause mortality was similar. The risk of a recurrent CVD hospitalisation was not higher in the most deprived after adjustment for CVD risk factors. However, I observed that SED was associated with higher mortality following an admission to hospital with CVD, before and after adjustment for cardiovascular risk factors of age, sex, smoking, systolic blood pressure, cholesterol and diabetes and after adjusting for the year of first developing cardiovascular disease. All cause mortality and cardiovascular mortality was highest in the most deprived. Again this association persisted after adjustment for cardiovascular risk factors. The most deprived also experienced longer hospital stays than the least deprived for a number of cardiovascular diseases including myocardial infarction and stroke. As a result the costs associated with cardiovascular disease admissions to hospital were highest in the most deprived despite their higher risk of dying during follow up. The cost differential was also explained by the finding that the most deprived experienced a higher number of admissions per person. Finally, the population attributable risk associated with SED is comparable to that of other traditional cardiovascular risk factors. In conclusion, I have found that the risk of CVD in the most deprived is higher even after adjustment for a number of cardiovascular risk factors. The numbers of hospitalisations, costs and mortality are also highest in the most deprived. Efforts are required to redress this imbalance. This can be achieved at the level of the individual through health care interventions to reduce the absolute burden of cardiovascular risk factors and to treat disease. However, societal level interventions are also required to tackle this problem as SED exerts complex effects on health that seem to also be independent of risk factors.

This thesis presents the use of pattern recognition and data mining techniques into risk prediction models in the clinical domain of cardiovascular medicine. The data is modelled and classified by using a number of alternative pattern recognition and data mining techniques in both supervised and unsupervised learning methods. Specific investigated techniques include multilayer perceptrons, radial basis functions, and support vector machines for supervised classification, and self organizing maps, KMIX and WKMIX algorithms for unsupervised clustering. The Physiological and Operative Severity Score for enUmeration of Mortality and morbidity (POSSUM), and Portsmouth POSSUM (PPOSSUM) are introduced as the risk scoring systems used in British surgery, which provide a tool for predicting risk adjustment and comparative audit. These systems could not detect all possible interactions between predictor variables whereas these may be possible through the use of pattern recognition techniques. The thesis presents KMIX and WKMIX as an improvement of the K-means algorithm; both use Euclidean and Hamming distances to measure the dissimilarity between patterns and their centres. The WKMIX is improved over the KMIX algorithm, and utilises attribute weights derived from mutual information values calculated based on a combination of Baye’s theorem, the entropy, and Kullback Leibler divergence. The research in this thesis suggests that a decision support system, for cardiovascular medicine, can be built utilising the studied risk prediction models and pattern recognition techniques. The same may be true for other medical domains.

Background: There is increasing evidence that psychological factors such as stress and depression might have an influence in the onset of many physical illnesses, but less is known about their effect from early life. This study is an epidemiological life-course analysis to test: (1) the hypotheses that early-life psychological factors are linked to later development of chronic disease (cancer, diabetes, and asthma) in mid-life, (2) whether such associations can be explained by pre-existing confounding factors, and (3) whether such links are mediated by other biological, behavioural, social, and cognitive factors. Methods: The data were from two ongoing prospective longitudinal studies following the lives of about 17,000 people born in Great Britain in one particular week in 1958 and 1970. Outcomes included diagnosis of cancer, asthma, and diabetes up to the year 2000. Psychological measures taken from ages 5 to 16 years were the main predictors. Associations were examined using discrete-time survival analysis and structural equation models, adjusting for potential confounders and mediators. Results: In the 1958 cohort, a standard deviation increase in the scores of conduct problems at ages 11 and 16 years, indicating severe behavioural problems, was associated with 2 to 34% increase in the odds of being diagnosed with cervical or all cancers after adjusting for childhood confounders. These effects were completely mediated by adulthood psychological distress and health behaviours. Only the teacher-reported behavioural problems significantly predicted the risk of diabetes after adjusting for family history of diabetes and sex (odds ratios of 1.05 to 1.08, p<O.Ol). These associations were partly mediated by mid-life psychological distress and adiposity. Significant associations were observed between most of the childhood psychological factors and adult-onset asthma even after adjusting for possible confounders and mediators. Conclusions: Childhood problem behaviours may predict chronic disease risk over the life-course either directly or mediated through adulthood factors. A consistent pathway among the disease groups was through adulthood psychological distress. Such continuities from childhood to adulthood psychological distress to the disease may be explained by the substantial biological plausibility of the association between psychological factors and physical health, primarily via alteration of the endocrine and the immune systems. The importance of promoting positive emotional and behavioural development in early life is stressed.