This essay has been submitted by a student. This is not an example of the work written by our professional essay writers.
Meta-analysis is a method of comparing the results of a number of studies which aim to investigate the same or very similar issue. However, simply pooling results from the variety of studies would lead to bias and confounding, due to the potential subtle differences in the individual studies, and the loss of randomization of sampling. The precision of each individual study must be taken into account. Larger studies provide more statistically reliable measures of effect, with smaller confidence intervals, compared with smaller studies where there is a larger scope for the effect of chance in the results. This lack of homoscedasticity, (equality of variance across studies) suggests that, rather than calculating simple means from a pool of results, a weighted average should be calculated, with larger, more precise studies being given more weight within the calculation than smaller studies with wider confidence intervals.
Weightings are used in meta analysis because studies have different sample sizes. For example, smaller studies should be given less weight as they would be more subject to the play of chance.
In a meta analysis it is unlikely that each study or observation is of equal importance, or that the accuracy of the effect size is the same. Study design, sample size and the population under observation is likely to differ in each study, and therefore homoscedasticity cannot be assumed (that a common variance exists across observations - a constant variance of the measure across observations). Weightings are used in meta analysis to allow for non-constant varience across the studies.
Pooling the results together and taking a simple mean would be inappropriate and may lead to misleading results (as it is unlikely that all of the studies included will be equally precise) â€¦â€¦â€¦..results would be misleading if the measure of effect estimates were combined with no weightings and the average taken
**** The sample size of a study will â€¦.. the standard errorâ€¦â€¦measures of effect from studies based on larger samples will tend to be more precise than those based on smaller studies
Studies using larger samples will tend to have smaller variance in comparison to studies using smaller samples, therefore measures of effect from larger studies will tend to be more precise. - more weight given to these studies??
With the help of examples, describe three types of heterogeneity (3 marks)
Clinical heterogeneity pertains to health differences in the patient/participant population, the nature of the intervention being assessed, the outcome measures used, and the duration of study follow up. For example, in a meta-analysis of studies of smoking cessation treatments, patient selection, baseline disease severity/addiction, intervention method used (i.e. nicotine replacement patch, varenicline, non-medicated support etc), and duration of follow-up may all introduce clinical heterogeneity into the analysis.
Methodological heterogeneity is the difference in the methods used in each individual study included in the meta-analysis. These differences could be: how participants are randomised, the extent of loss to follow up and handling of these, and timescale of the study. For example, the methodological differences between an observational studies and a randomised control trial in terms of study population recruitment may introduce methodological heterogeneity into the meta-analysis.
Statistical heterogeneity is the effect of incompatibilities of the results of the constituent studies within the meta-analysis. These incompatibilities may be caused by clinical or methodological differences between the studies, related to another unknown study characteristic, or combinations of all three.
With the help of model notation, describe how random effect methods account for heterogeneity between the studies (5 marks)
Random effect methods modify the meta-analysis by allowing for heterogeneity within the model. They assume that an additional source of variation is present due to variances in measures of effect between the constituent studies. The random effects model works by modifying the weighting of each study effect that would be used in a fixed effect analysis, with smaller studies receiving relatively more weight within the model. This means that, whilst more weight may be given to the smaller studies, larger ones are less prone to being overweighted, which may dominate the analysis. The analysis produces a measure of effect which estimates a mean, around which each of the constituent studies measures of effect are assumed to vary. In other words, it treats the measure of effect of each study as a random sample within a distribution around the mean effect calculated through their combination within the meta-analysis. This is illustrated in the method below.
First it is useful to set out the model notation for the fixed effect method:
Where represents the measure of effect estimate from the 'th study, which equals the underlying measure of effect, , plus the random error, , related to each of the studies. This error has a mean of zero, with a variance equal to the variance of the sample in each study.
The random effect model is largely the same, however the heterogeneity of the sample studies must be built into the model, shown by below:
The inclusion of heterogeneity within the model means we treat the measure of effect of each study as a sample of a distribution, commonly a normal distribution, with a mean of and a variance of , the variance in measures of effect from each individual study.
Heterogeneity is assessed within the model using the statistic, which measures heterogeneity as a proportion of the total variability within the model:
Where equals the sum of the squared deviations of each study () from the pooled estimate mean (). Each of the squared deviations are given a weighting of the inverse variance of the study (). refers to the number of individual studies being combined within the analysis. The * denotes the use of the inverse variance method relating to a random effect model, rather than the fixed-effect model as shown below:
The result of the statistic becomes the in the overall random effect model.
Using the tabulated hypothetical data below, perform an appropriate meta-analysis to assess the association between daily aspirin use and the risk of lung cancer, describing why you have used the method you have chosen, and perform additional analyses to assess the importance of year of publication, study design and quality, and test for publication bias (10 marks)
Section B (Answer ONE question from this section)
With the help of examples, describe two situations in which you might want to undertake time series analysis (2 marks)
Changes to the licensing regulations (more drinking? Less violent crime/assault victims in A & E)
Changes in legislation. Smoking ban?
A public health intervention - mass media campaign, healthy eating,
Seatbelts compulsory in cars
Healthy school campaign (take up in walking/cycling to schools across a city)
Access to confidential sexual health advice - teenage pregnancy
With the help of examples, describe the four different components of a time series (4 marks)
Describe what autocorrelation is and explain why you should not ignore it when analysing a time series (4 marks)
Critically appraise the study "Use of an interrupted time-series design to evaluate a cancer screening program" by Michielutte et al, focussing on design and methods used and the interpretation and conclusions reached (10 marks)
Why is it important for clustering to be taken account of with multilevel data? (2 marks)
Describe the features of a random intercept model, which has one random effect term, and state why this is a useful model (5 marks)
Longitudinal surveys, such as The English Longitudinal Study of Aging, collect data on several occasions from the same people as they age over time. Describe how you might use a multi-level model in analysing these type of data (3 marks)
Critically appraise the study "Smoking cessation in England: Intentionality, anticipated ease of quitting and advice provision" by Twigg et al focussing on design and methods used and the interpretation and conclusions reached (10 marks)
Describe how food composition tables are used in nutritional epidemiology (4 marks)
Food composition tables, also known as nutrient databases, contain data on the nutritional properties of a range of foodstuffs. These tables or databases are used to convert records of food consumption to measures of nutrient intake. Many countries across the world have their own Food Composition Tables, tailored to common foodstuffs available within that country. However, due to increased globalisation of foods and beverages a number of 'international' nutrient databases are becoming common.
Food Composition Tables are often used to measure the intake, and the effects of changes in intakes, of nutrients in large observational studies where direct nutrient analysis methods are impractical. Whereas direct analysis methods are considered to be more accurate and preferable in a small scale study, timescales and budgets often preclude their use in larger studies. Advances in computing power increasingly mean that analysis of large datasets can be automated and processed in far shorter timescales than through direct analysis.
Depending on the focus of the study being undertaken, bespoke tables can be constructed for quicker reference. The tables often categorise different foodstuffs based on a number of different factors such as biological characteristics, preparation methods, or nutrient concentration. For detailed studies, these food groups may be split to many sub-categories such as species or plant variety in order to provide more focussed nutrient values, whereas broader studies may just use more generic food groups such as 'vegetables' or 'dairy produce'.
Discuss two disadvantages of using food composition tables (4 marks)
Whilst useful, Food Composition Tables have disadvantages. Firstly, not all foodstuffs are the same, and therefore vary in their nutritional value. Many factors can affect the nutritional make up of a raw foodstuff. Farming methods differ across countries and continents, as do natural nutrient levels in soil and the use of artificial fertilizers rather than natural manures. Differing varieties of fruits and vegetables may have differing nutrient content, whilst packing, storage and transportation times may degrade nutritional value. In the case of animal products, differences in feeds, farming methods and welfare standards may have an impact on nutrient content. Food Composition Tables, for raw products, may only give an averaged value for types of products, along with measures of the variability of these values.
Secondly, Food Composition Tables are reliant on data for all foodstuffs and products. This is possible for raw natural products, however, many products contain a combination of different raw and pre-processed ingredients. Estimates of nutritional value therefore rely on combinations of estimates for known products, and possibly unknown products. Variations in recipes, preparation and manufacturing methods also lead to different nutritional values for what is, ostensibly, the same product.
Combinations of these drawbacks and uncertainties may lead to underestimations of an individual's nutrient intake, especially if unknown products are given zero values in the reference table being used. In most of these cases, substitute values for biologically similar foodstuffs are used, or alternative datasets may be used to provide estimated values.
Describe what 'RNI' is (2 marks)
RNI is the acronym used for Reference Nutrient Intake, a British measure used to calculate the recommended daily amount of different vitamins and minerals to be consumed each day. The Reference is age specific and, in the case of some nutrients, varies by specialist groups such as pregnant women or breastfeeding mothers. The reference, if followed, is designed to ensure that 97.5% of the population have an adequate nutrient intake to avoid deficiencies. (how to ref? Institute for Optimum Nutrition)
The RNI forms part of the wider 'Dietary Reference Values for Food Energy and Nutrients for the United Kingdom', published by the Department of Health in 1991.
How to ref this? Eatwell.gov.uk (where the doc is)
Critically appraise the study "Dietary factors associated with physician-diagnosed asthma and allergic rhinitis in teenager: analyses of the first Nutrition and Health in Taiwan" by Huang et al, focussing on design and methods used and the interpretation and conclusions reached (10 marks)
For the critiquing part of the questions, you are to use your notes from the lectures together with the knowledge you have gained over the course to critically evaluate the methods presented in the paper. You are not required to find references or examples in papers or books.
The questions asked you to critically appraise the methods used in the papers from the questions you choose. Therefore I would expect to see you talk about the method they have used, describe how they used it, discuss the limitations of the method and whether this was the best method to use for their study, and if appropriate discuss a more appropriate method they could have chosen, and discuss whether they have interpreted the findings of the method correctly. You need to use the notes we gave you from each session to be able to do this. I would say you should be able to cover all of the pertinent points with one to two sides of typed A4 for each critical appraisal.
The paper aims to investigate the association between diet and doctor diagnosed asthma and allergic rhinitis within teenagers, using the first Nutrition and Health Survey in Taiwan. The study was carried out in an attempt to identify lifestyle elements that posed risks to the population of 'more affluent' areas in terms of asthma and allergic rhinitis, following previous research. The research also followed the publication of results of previous studies involving the Second National Health and Nutritional examination Survey of the USA which suggested that lifestyle, and more specifically calorie intake and obesity, were identified as risk factors for wheezing within younger age groups.
The paper used data from the first Nutrition and Health Survey of Taiwan to
Due to the design of the study, the diet was assessed after the onset of asthma or rhinitis