Epidemiology 2.6 Flashcards
How can sample size be determined?
several mathematical equations – depending on the type of study you’re running and the variables involved - while you don’t need to know how to calculate the sample size at this stage of training by hand, you should be aware of the factors that influence sample size
What are the three main statical factors that will push you sample size up or down?
- Difference between groups the you would be looking at
- The study power
- The study alpha
How does the difference between the groups you are looking at affect sample size?
- For example, if investigating blood pressure, we might want to look for perhaps 5 mmHg or 10 mmHg difference
- A larger difference will require a smaller sample size – just like photographing a larger object doesn’t require as good a camera (as photographing a smaller object).
How does the study power affect sample size?
- aim for 80%
2. as you increase your power, your sample size will increase.
How does the study alpha affect sample size?
- how much you want to rule out chance causing a positive finding
- it’s the equivalent of specifying the p-value and it’s also connected with one- and two-tailed testing
- Decreasing alpha from 0.05 to 0.01 will increase your sample size.
What else effects sample size?
- consider loss to follow-up
- as a rule of thumb, it’s not unusual to lose 15% of your participants – and of course longer studies will have higher loss
What is a type I error?
‘false positive’ finding
What is the p value?
the probability of obtaining a result as extreme as the observed results of a statistical test, assuming the null hypothesis is correct
What does. the p value mean?
- probability of getting your results if actually there’s no real difference
- that means it’s the probability of a false positive
When is there statical significance?
<0.05
Why are p values not so helpful?
- If you run multiple analyses, likely that one of your analyses will come back around the p = 0.05 mark - running multiple analyses on your data will eventually come back with something <0.05 when there’s no real underlying difference between your groups.
- The p-value is often mistaken as implying clinical significance. It doesn’t.
- All a p-value tells you is how likely your findings are due to chance alone
- It’s perfectly possible to have highly ‘statistically significant’ findings of a clinical difference which is entirely clinically meaningless.
When are p values helpful?
consistent findings over multiple tests <0.05 or single tests <0.01 then it gives some reassurance that chance alone is less likely to be playing a part
What is a type II error?
‘false negative’ finding
What does going form 80% to 90% poor do to the risk of deriving false negative?
Halve, but need to increase sample size which increase cost and complexity of trial
What is narrative review?
- Brings together published literature into a single article enabling the reader to rapidly understand issues
- Sometimes referred to as a literature review, scoping review or non-systematic review
What is a systematic review?
- Sets out highly structured approach to searching, sifting, including and summarising the literature
- Often presented as the basis for meta-analysis, but also exists separately
What are the strengths of a narrative review?
- Agile, normally easier and faster to write; may often be more up to date than latest systematic review
- Particularly useful when looking at areas with limited research or in higher levels or variation in research approaches
- Can be useful where work from different discipline is being brought together with a less easily-answerable research question
What are the limitations of a narrative review?
- Subject to potential bias as
1. Author free to select works (sometimes that may support their opinion): although ethical authors and robust peer review should prevent overly-speculative or unbalanced views
2. With no search being specified it is possible that important evidence is omitted by chance (rather than. intent)
What are the strengths of a systematic review?
- Aims to collate all available evidence that relates to a highly focused research question
- Implements a highly specified protocol that enables reproducibility - this will usually involved a structured search
- ‘Includes’ evidence based on pre-specified criteria: inclusion criteria
- Can take many months to design the search, review all sources (often thousands) and then synthesise the findings
What are the steps for a systematic review?
- Research questions
- Structured search
- Indices (categorise indices)
- Screening/inclusion
- Reporting
- Writing
- Submitting/revising/publishing
- Can take 18-24 months
What is a structured search?
- approach enables transparency and future researcher to reproduce the approach
- Words and heading and apply to index e.g. med and give reproducible approach to finding literature
- Develop an approach and series of searches that combine later on and justify why used some articles and why omitted some
Why are indices helpful?
It is always made clear which each indices were used and the time period involved (needs to be reproducible and transparent)
What are indices?
based on published research
What are registries?
- registration of research yet to be completed or published
- registries increasingly important to avoid duplication or omission
- interventional research register to prevent risk of publication bias (not reporting negative finding and over reporting positive findings )
What is the screening process?
- PRISMA diagram
1. Identification
2. Screening
3. Eligibility
4. included
What does the screening process entail?
- How many articles found through initial search, or removed as duplicates
Then look at abstract
Then full text review
And then how many studies included in meta-analysis - Often systematic review will identify thousand of abstracts and articles in the early stages then like 10-30 in end as
-very few may actually be revenant but applying inclusion and exclusion criteria is important to systematise the process
-A PRISMA flow diagram often used to show the ’n’ at each stage
2,500 down to 25
-Find all evidence and then screen out the to get comprehensive review
What are the limitations of a systematic review?
- Only as good as method employed: a less-than-comprehensive search structure may not return all the evidence
- Only as good as indices search
- Only as good as evidence it incorporates
- Very quickly. out of. date, often exacerbated by delay in publishing, look at search date not publication date
- Can take 18-24 months to. produce and a systematic review published only every 2-10 years
What is grey information?
-Not. everything that is known is necessarily in the published literature: that has been reviewed and published within academic journals that can be searched via online indices
What is an example of grey information?
government reports of evaluations of programmes
When is grey information that makes up grey literature an issue?
In public health and global health areas, where many implementations and programmes might be reported but with varying degrees of accuracy
How do you make decisions when so many papers are published every day?
- Narrative reviews and systematic review are helpful as starting point
- Can snow ball out from these reviews to find individual literatures
- Also look at more recent citations of these reviews using online indices
What is a meta-analysis?
A quantitative, formal epidemiological study design used to systematically assess previous research studies to derive conclusions about that body of research
What is the. relative risk?
numeral presentation of the finding which provides point estimate and confidence intervals around that which is the range of values of which we are 95% confident the true value/ point estimate lies
What does meta analysis need?
requires that the pooled studies are sufficiently similar: or else you results are meaningless
What are sources of heterogeneity in meta-analysis?
- Clinical: patients, selection criteria
- Methodological: study design, blinding, intervention approach
- Statistical: reporting differences
Are there statistical ways of evaluating the heterogeneity between studies?
Yes
When can meta analysis be limited?
- By quality of the studies included
- The statistical methods employed
- Often only possible after multiple interventional or observational studies have been undertaken
What does a meta analysis need?
- Almost always necessitates a systematic review to identify the papers/studies to be included in the estimates
- Sometimes necessary to go back to original authors of included msyudieas to ask for more data, enabled the meta-analysis to be conducted
What is another consideration in meta analysis?
- Publication bias
- Publication with more positive findings are more likely to be published by journals and more likely to be submitted for publication
What is a publication funnel plot?
Can show the balance of evidence between studies assuming an overall effect size: more publications one side of lines may suggest that some studies may not have been reported
What is the best thing for scientific evidence?
meta-analysis of randomised controlled trials
What is a trial endpoint?
an outcome that usually describes a clinically meaningful outcome
e.g. cancer survival 12 months or 5. years
What are two ways that trial outcomes can be determined?
- Efficacy – how well a therapy works in achieving a desired outcome
- Safety – how well a therapy works in not causing adverse events
What is a primary endpoint in efficacy?
- the endpoint for which the study has been powered
- that is to say that the number of trial participants (sample size) will have been recruited on the basis of the pre-specified power and difference
What is the secondary endpoint in efficacy?
- common that a study will want to examine a slightly different endpoint in addition to the primary endpoint
- E.g while a study seeks to examine survival often ‘softer’ - measure such as recurrence of disease or hospital admission might also be measured
- If the secondary endpoint is proven but the primary endpoint is not, then the findings of the study may still contribute to the understanding of disease
What are safety endpoints?
- potentially measuring commonly observed adverse events (AEs) and grading them into a hierarchy of significance
- A large proportion of patients reporting AEs will require investigation.
- Eventually a judgment will need to be made as to what extent the safety profile (or lack thereof) of a therapy is offset by the efficacy
What is a composite endpoint?
- variation of the above and could potentially describe any endpoints - multiple potential endpoints have been added together
- particularly common when an outcome is uncommon
- For example, one might combine myocardial infarction and ischemic stroke to give a new composite endpoint of ‘cardiovascular event’.
What is survival analysis?
- for hazard ratio may happen at six months may not be the same at two years
2 losing statistical precision by generating arbitrary timepoints - If patient dies in 5 year (primary efficacy endpoint) then record when that death takes place so then range of survival times
- Compare likelihood of survival between two study. arms
How do you represent survival analysis?
- The way that we display these findings is using the Kaplan Meier plot.
- On the Y-axis we have overall survival (OS)
- On the X-axis we have 60 months – that’s five years.
- The two coloured lines indicate the proportion of participant alive at any point in time
- You can see that at t=0, 100% of the participants were alive in either arm
- Look at qs
What is the alpha level?
- Threshold for type 1 error
- How many type 1 error before rejecting null hypothesis
Why do you want a lower alpha?
- Harder to get false positives
- Ideal if severe consequences
What is beta level?
- Threshold our type 2 error
- Determines power of study = probability of getting a true negative
How do you calculate power?
1-beta level so if beta 0.2 then power is 0.8 or 80%
What is the relationship between beta and alpha?
Inversely proportional