The LLP risk model: an individual risk prediction model for lung cancer.

Using a model-based approach, we estimated the probability that an individual, with a specified combination of risk factors, would develop lung cancer within a 5-year period. Data from 579 lung cancer cases and 1157 age- and sex-matched population-based controls were available for this analysis. Significant risk factors were fitted into multivariate conditional logistic regression models. The final multivariate model was combined with age-standardised lung cancer incidence data to calculate absolute risk estimates. Combinations of lifestyle risk factors were modelled to create risk profiles. For example, a 77-year-old male non-smoker, with a family history of lung cancer (early onset) and occupational exposure to asbestos has an absolute risk of 3.17% (95% CI, 1.67-5.95). Choosing a 2.5% cutoff to trigger increased surveillance, gave a sensitivity of 0.62 and specificity of 0.70, while a 6.0% cutoff gave a sensitivity of 0.34 and specificity of 0.90. A 10-fold cross validation produced an AUC statistic of 0.70, indicating good discrimination.If independent validation studies confirm these results, the LLP risk models' application as the first stage in an early detection strategy is a logical evolution in patient care.

Family history and risk of lung cancer: age-at-diagnosis in cases and first-degree relatives.

To investigate the little known risk of lung cancer at an early age when a first-degree relative has had such a diagnosis, 579 incident cases and 1157 population controls were studied in Liverpool between 1998 and 2004 using standardised questionnaires covering demography and lifestyle. A history of lung cancer in first-degree relatives was associated with a significantly increased risk in the proband where in both individuals the cancers were diagnosed before the age of 60 years (odds ratio (OR)=4.89; 95% confidence interval (CI): 1.47-16.25). A significantly elevated risk of lung cancer was also observed in association with a relative affected before the age of 60 years, regardless of age-at-onset of the disease (OR=2.08; 95% CI: 1.20-3.59). This finding is strongly consistent with a genetic component in early-onset lung cancer risk.

A potentially useful distribution model for dietary intake data.

BACKGROUND: Conventional mixed models for the analysis of diet diary data have introduced several simplifying assumptions, such as that of a single standard deviation for within-person day-to-day variation which is common to all individuals. OBJECTIVE: We developed a model in which the within-person standard deviation was allowed to differ from person to person. DESIGN: The model was demonstrated using data on daily retinol intake from the Dietary and Nutritional Survey of British Adults. The data were from 7-day weighed dietary diaries. Estimation was performed by Markov chain Monte Carlo. Reliability of the model was assessed from the accuracy of estimation of the percentage of days on which various intakes were exceeded. For levels above the median retinol intake, estimation of percentages of days with excessive intakes was most accurate using the model with varying within-person standard deviation. SETTING: A survey of British adults aged 16-64 years. SUBJECTS: In total 2197 adults living in the UK, 1087 males and 1110 females. RESULTS: Under the traditional model, estimated daily intake ranged from 716.4 to 1421.8 microg depending on age and sex, with a within-person standard deviation of 4298.9 microg. Under the new model, estimated average daily intake ranged from 388.9 to 518.3 microg depending on age and sex, but with a within-person standard deviation varying between subjects with a 95% range of 29 to 8384 microg. The new model was shown to predict the percentage of days of exceeding large intakes more successfully than the traditional model. For example, the percentage of days of exceeding the maximum recommended intake (9000 microg for men and 7500 microg for women) was 2.4%. The traditional model predicted no excessive intakes, whereas the new model predicted 2.9%. CONCLUSIONS: This model is potentially useful in dietary research in general and for analysis of data on chemical contaminants in foods, in particular.

Probabilities of progression of aortic aneurysms: estimates and implications for screening policy.

BACKGROUND: Screening for abdominal aortic aneurysm, and intervention with elective repair, can reduce the incidence of aneurysmal rupture by a half. If a screening programme is implemented, it is essential to determine appropriate follow up intervals for rescreening. This paper estimates probabilities of progression growth of aortic diameter to provide evidence for this. METHODS: Data were taken from 2342 men aged 65-80 screened in the Chichester randomised control trial, who have been followed up for an average of 11 years. Aortic diameter was modelled as a Markov process with four categories: <30 mm (normal), 30-44 mm, 45-54 mm, and > or =55 mm. Estimates of the probabilities of progressing to each higher category were obtained. RESULTS: The probabilities of progression increased with greater initial aortic diameter. The estimated rates/year were 0.018 (95% confidence interval 0.014 to 0.023), 0.16 (0.12 to 0.20), and 0.49 (0.35 to 0.70) respectively for moving up one category. The probabilities of moving from <30 mm to > or =55 mm were estimated as 1% in 5 years and 12% in 15 years, while the corresponding figures for moving from 45-54 mm to > or =55 mm were 91% and 99%. There were differences in rates of progression according to age, with men over 70 years having rates about three times those of men under 70. CONCLUSIONS: It seems unnecessary to follow up men with normal aortic diameter as they experience a low probability of reaching criteria for surgery even within 15 years. However, follow up intervals should be progressively shorter for those with greater aortic diameter, especially in those aged over 70. Active follow up, for example every 3 months, is appropriate for men with an aortic diameter of 45-54 mm.

Bayesian methods in health technology assessment: a review.

BACKGROUND: Bayesian methods may be defined as the explicit quantitative use of external evidence in the design, monitoring, analysis, interpretation and reporting of a health technology assessment. In outline, the methods involve formal combination through the use of Bayes's theorem of: 1. a prior distribution or belief about the value of a quantity of interest (for example, a treatment effect) based on evidence not derived from the study under analysis, with 2. a summary of the information concerning the same quantity available from the data collected in the study (known as the likelihood), to yield 3. an updated or posterior distribution of the quantity of interest. These methods thus directly address the question of how new evidence should change what we currently believe. They extend naturally into making predictions, synthesising evidence from multiple sources, and designing studies: in addition, if we are willing to quantify the value of different consequences as a 'loss function', Bayesian methods extend into a full decision-theoretic approach to study design, monitoring and eventual policy decision-making. Nonetheless, Bayesian methods are a controversial topic in that they may involve the explicit use of subjective judgements in what is conventionally supposed to be a rigorous scientific exercise. OBJECTIVES: This report is intended to provide: 1. a brief review of the essential ideas of Bayesian analysis 2. a full structured review of applications of Bayesian methods to randomised controlled trials, observational studies, and the synthesis of evidence, in a form which should be reasonably straightforward to update 3. a critical commentary on similarities and differences between Bayesian and conventional approaches 4. criteria for assessing the reporting of a Bayesian analysis 5. a comprehensive list of published 'three-star' examples, in which a proper prior distribution has been used for the quantity of primary interest 6. tutorial case studies of a variety of types 7. recommendations on how Bayesian methods and approaches may be assimilated into health technology assessments in a variety of contexts and by a variety of participants in the research process. METHODS: The BIDS ISI database was searched using the terms 'Bayes' or 'Bayesian'. This yielded almost 4000 papers published in the period 1990-98. All resultant abstracts were reviewed for relevance to health technology assessment; about 250 were so identified, and used as the basis for forward and backward searches. In addition EMBASE and MEDLINE databases were searched, along with websites of prominent authors, and available personal collections of references, finally yielding nearly 500 relevant references. A comprehensive review of all references describing use of 'proper' Bayesian methods in health technology assessment (those which update an informative prior distribution through the use of Bayes's theorem) has been attempted, and around 30 such papers are reported in structured form. There has been very limited use of proper Bayesian methods in practice, and relevant studies appear to be relatively easily identified. RESULTS: Bayesian methods in the health technology assessment context 1. Different contexts may demand different statistical approaches. Prior opinions are most valuable when the assessment forms part of a series of similar studies. A decision-theoretic approach may be appropriate where the consequences of a study are reasonably predictable. 2. The prior distribution is important and not unique, and so a range of options should be examined in a sensitivity analysis. Bayesian methods are best seen as a transformation from initial to final opinion, rather than providing a single 'correct' inference. 3. The use of a prior is based on judgement, and hence a degree of subjectivity cannot be avoided. However, subjective priors tend to show predictable biases, and archetypal priors may be useful for identifying a reasonable range of prior opinion.