CAT Flashcards
What are case-control studies?
Observational and retrospective
They examine an association between an outcome and exposure to a risk factor
The studies recruit participants based on the presence or absence of an outcome. Data of exposure to a risk factor is then recorded retrospectively for each of the participants and compared between cases and controls
It asks ‘ what happened?’
What do case-control studies measure?
Odds ratios
What are cohort studies?
Observational and prospective/retrospective
They compared 2 groups, one with and one without exposures and examine them over a period of time to establish links between exposure and a health related outcome
It asks “what will happen”?
What do cohort studies measure?
Relative risk
What are cross-sectional studies?
Observational
A sample of individuals from a population is selected and data is collected on both the exposure of interest and outcome of interest. This allows estimations of prevalence and risk factors
Asks “what is happening?”
What are advantages and disadvantages of case-control studies?
Advantages:
- good for investigating rare outcomes as these are identified at the start
- quicker than cohort studies and easier - don’t require long follow ups
- relatively inexpensive
- good for long latent period
- can study multiple exposures associated with a disease
Disadvantages:
- recall bias
- not suitable when exposure to risk is rare
- subject to selection bias
- difficult to establish causality as exposure data is collected retrospectively i.e. did exposure precede disease
Whats an example of a case-control study in medical research?
a study investigating the association between smoking and lung cancer. In this study, cases would be individuals diagnosed with lung cancer, and controls would be individuals without lung cancer. Researchers would then collect information on the smoking history of both cases and controls, including the number of cigarettes smoked per day, the duration of smoking, and the age at which individuals started smoking.
The researchers would then compare the smoking history of cases and controls to determine whether there is an association between smoking and lung cancer. If the study finds that individuals who smoke have a higher risk of developing lung cancer compared to those who do not smoke, this would provide evidence for the harmful effects of smoking on lung health.
What are the advantages and disadvantages of cohort studies?
Advanatges:
- can establish temporal relationships between exposure and outcome (best information about causation)
- can study multiple outcomes associated with a particular exposure
- can study rare exposures
- can provide incidence rates
Disadvantages:
- long follow up - expensive ans time consuming
- attrition bias
- selection bias
- they rely on accurate measurement of exposure and may be subject to misclassification bias if expsoure is measured inaccurately
- not suitable for rare outcomes
- bad for long latency periods
Whats an example of a cohort study in medical research?
Framingham Heart Study, which began in 1948 and has followed several generations of participants in the town of Framingham, Massachusetts, to investigate the risk factors for cardiovascular disease. This study has been instrumental in identifying risk factors for heart disease, such as high blood pressure and cholesterol, and has contributed to the development of interventions and treatments for these conditions.
What are advantages and disadvantages of cross-sectional studies?
Advantages:
- quick
- inexpensive
- can study multiple outcomes
Disadvantages:
- cannot establish causality (can’t determine whether exposure or outcome came first)
- subject to selection bias as sample may not be representative of entire population
- cannot study rare outcomes are sample size unlikely to be large enough
What are the key differences between observational studies and experimental studies?
Observational studies are designed to observe and measure outcomes in a population without intervention. The goal of observational studies is to identify associations between exposures or risk factors and outcomes. These studies can be further divided into cross-sectional, case-control, and cohort studies.
Experimental studies, on the other hand, are designed to test hypotheses by manipulating one or more variables and measuring the effect on the outcome. Experimental studies can be further divided into randomized controlled trials (RCTs) and non-randomized controlled trials
What are the advantages and disadvantages of observational studies?
Advantages:
- cost effective
- real world settings which increases generalisability
- can study rare outcomes which would not be feasible in an experimental setting
Disadvanatges
- limited control so difficulty in drawing conclusions about causality
- prone to confounding
- cannot establish causality
What are the advantages and disadvantages of experimental studies?
Advantages:
- control - which increases the ability to establish causality
- eliminates confounding factors by methods such as blinding - increases internal validity
- can be replicated more easily which increases the reliability of findings
Disadvantages:
- expensive due to costs of intervention
- often conducted in artificial setting which may limit the generalisability
- ethical considerations e.g. use of placebo group
What are the types of experimental studies?
RCTs
Quasi-experimental studies
Single-subject experimental designs
What are RCTs?
These studies are considered the gold standard in experimental research. Participants are randomly assigned to either an experimental group or a control group, with the experimental group receiving the intervention or treatment being tested, while the control group receives a placebo or standard treatment. RCTs are designed to minimize bias and confounding variables and to ensure that any observed effects are due to the intervention being tested.
What are Quasi-experimental studies?
These studies are similar to RCTs but lack the random assignment of participants to groups. Participants are assigned to either an experimental or control group based on non-random methods, such as geographical location or medical history. While these studies are less rigorous than RCTs, they can still provide valuable information about the effectiveness of interventions.
What are single-subject experimental designs?
These studies involve one or a few participants who receive multiple measurements of a treatment or intervention over time. This design is particularly useful for studying rare or complex conditions and allows for the evaluation of individual responses to a treatment.
What are advantages and disadvantages of RCTs?
Advanatges:
- high internal validity
- ability to establish causation
- reproducible
- generalisable
Disadvanatges:
- expensive
- time consuming
- ethical concerns about withholding treatment from control group
- limited external validity
- attrition and non-compliance
What is bias?
Systematic (as opposed to random) deviation of the results of a study from the ‘true’ results, which is caused by the way the study is designed or conducted.
What is selection bias?
Non-random assigning individuals to groups, leading to differences in group’s qualities that may influence the outcomes
I.e. the subjects are not representative of the population
What is recall bias?
Difference in accuracy of recollection of study participants - may be due to time or influenced by motive (i.e. knowledge of presence of disorder alters recall by subjects)
Which study type is most prone to recall bias?
Case-control studyes
What is publication bias?
Failure to publish or include certain studies because they have negative results
Occurs when issues other than the quality of the study are allowed to influence the decision to publish
Which study type is publication bias most important in?
Systematic reviews and meta-analyses (where studies showing negative results may be excluded)
What is the Hawthorne effect?
The group changing their behaviour due to knowledge they are being studied
What is procedure bias?
Subjects in different groups recieving different care/not treated the same, other than just the intervention
E.g. subjects in treatment group get more attention which stimulates greater compliance
What is information bias?
when information used in a study is either measured or recorded inaccurately.
What is observer bias?
Aka pygmalion effect or expectation bias
When observers subcuncuously measure or report data in a way that favours the expected study outcome
Only a problem in non-blinded trials
What is confirmation bias?
the tendency to interpret new evidence as confirmation of one’s existing beliefs or theories.
What is attrition bias?
a type of selection bias due to systematic differences between study groups in the number and the way participants are lost from a study.
Which type of study is most prone to selection bias?
Cohort studies
What is work-up bias?
Aka verification bias
a type of measurement bias in which the results of a diagnostic test affect whether the gold standard procedure is used to verify the test result.
Seen in studies trying to validate a new diagnostic test
In clinical practice, verification bias is more likely to occur when a preliminary diagnostic test is negative. Because many gold standard tests can be invasive, expensive, and carry a higher risk (e.g. angiography, biopsy, surgery), patients and physicians may be more reluctant to undergo further work-up if a preliminary test is negative.
What is randomisation?
Randomization is the process of assigning participants to treatment and control groups, assuming that each participant has an equal chance of being assigned to any group
What are the types of randomisation?
Simple randomisation
Stratified randomisation
Block randomisation
Covariate adaptive randomisation
What is simple randomisation?
This is the most basic type of randomization, where each individual or item in a population has an equal chance of being selected.
E.g. computer random number generator
What is stratified randomisation?
This method is used when a population can be divided into subgroups or strata based on certain characteristics. The individuals are then randomly assigned to the treatment or control group within each stratum, ensuring that each group is representative of the strata.
What is blocked randomisation?
Individuals are groups based on their variable characteristics and then within these groups there will be random allocation to treatment or control group within each block
This method is used to ensure that the treatment and control groups have an equal number of individuals with certain characteristics or variables.
What is cluster randomisation?
This method is used when individuals cannot be randomized individually but rather in groups or clusters, such as schools or neighborhoods. The clusters are then randomly assigned to the treatment or control group.
What is covariate adaptive randomisation?
a type of adaptive randomization that uses pre-specified participant characteristics, or covariates, to determine the allocation of participants to treatment groups in a clinical trial. This method is used to improve the balance between treatment groups with respect to known prognostic factors that may affect the outcome of the study.
What is clinical equipoise?
This means that there is genuine uncertainty in the expert medical community over whether a treatment will be beneficial
This is lost when there is prior reason to believe 1 intervention is superior/inferior to the other
Whats the problem with inadequate randomisation?
If randomization is inadequate, then there may be systematic differences between the groups that are not related to the treatment being studied.
This can lead to biased estimates of treatment effects and reduce the internal validity of the study. Inadequate randomization can also increase the risk of type I and type II errors, where false positive or false negative results are obtained.
What is blinding?
a method used in research studies to prevent bias by ensuring that participants, researchers, and outcome assessors do not know which treatment group a participant has been assigned to.
Blinding can be single-blind, where only the participants or the assessors are unaware of the treatment group, or double-blind, where both the participants and the assessors are unaware of the treatment group.
Why is blinding important?
Prevents observer bias, confirmation bias and prevents placebo effect
What is intention to treat analysis?
An assessment of the people taking part in a trial, based on the group they were initially (and randomly) allocated to. This is regardless of whether or not they dropped out, fully adhered to the treatment or switched to an alternative treatment. Intention-to-treat analysis analyses are often used to assess clinical effectiveness because they mirror actual practice, when not everyone adheres to the treatment, and the treatment people have may be changed according to how their condition responds to it.
What is power of a study?
the probability of rejecting the null hypothesis when it is in fact false
What is usually an acceptable power value?
80-90%
What is the P-value?
a statistical measure that indicates whether or not an effect is statistically significant.
p value <0.05 it is considered that there probably is a real difference between treatments.
p value is 0.001 or less the result is seen as highly significant.
What are confidence intervals?
A way of expressing how certain we are about the findings from a study, using statistics. It gives a range of results that is likely to include the ‘true’ value for the population
The CI is usually stated as ‘95% CI’, which means that the range of values has a 95 in a 100 chance of including the ‘true’ value.
What does the width of CI indicate?
. A wide confidence interval (CI) indicates a lack of certainty about the true effect of the test or treatment - often because a small group of patients has been studied. A narrow CI indicates a more precise estimate (for example, if a large number of patients have been studied).
What is an underpowered trial?
a clinical trial that has an insufficient sample size or statistical power to detect a true effect of the intervention being studied.
A study with low statistical power has a high risk of yielding a false negative result, meaning that it fails to detect a true effect even when one exists. (Type 2 statistical error)
What is subversion bias?
The deliberate manipulation of pt recruitment by clinicians in order to influence trial results e.g. shining a light through an envelope to determine the patient’s allocation group
What is standard deviation?
A value that shows how much variation there is from the mean
What is validity?
Whether a test or study actually measures what it aims to measure
what is the difference between internal and external validity?
Internal validity shows whether a study or test is appropriate for the question, for example, whether a study of exercise among gym members measures the amount of exercise people do at the gym, not simply whether people join.
External validity is the degree to which the results of a study hold true in non-study situations, for example, in routine NHS practice. It may also be referred to as the generalisability of study results to non-study populations. For example, the external validity of a study that took place in Spain may be questioned if the results were to be applied to people in Australia.
What is a type 1 error?
Aka false positive
occurs when the null hypothesis is rejected even though it is true. In other words, stating there is an effect when none exists
What is a type 2 error?
Aka false negative
occurs when the null hypothesis is not rejected even though it is false. In other words, stating there is not an effect when one exists
What is the null hypothesis?
the hypothesis that there is no significant difference between specified populations, any observed difference being due to sampling or experimental error.
It is usually denoted as H0
What is the alternative hypothesis?
Hypothesis that there is some assocation between disease and risk factor
It is usually denoted as H1 and is the hypothesis that researchers want to support by rejecting the null hypothesis
What methods are used to limit confounding in a cohort study?
Restriction
Matching
Stratification
Multiple variable regression
Propensity score matching
Sensitivity analysis
What is restriction?
Limiting participants of study who have possible confounders - means less data!
What is matching and stratification?
You make comparison groups i.e. groups with and without the confounder, adjust for confounding and then recombine the data
What is multivariable regression?
Coefficients are established for the confounder groups
This allows for better adjustment
What is a meta-analysis?
A statistical technique often used in systematic reviews to ‘pool’ the results from several studies of the same test, treatment or other intervention to estimate the overall summary effect of the treatment.
This provides a more precise estimate of treatment effect than a single study would
(In mark scheme it says you need to say pool and precise for both marks!)
What are the pros and cons of a meta-analysis?
Pros:
Increased statical power
Increases generalisability by including studies with diverse populations, settings and study designs
Highly transparent - provides a systematic and objective approach to summarising evidence
Higher precision due to higher number of pt
Cons:
Limited by heterogeneity
Publication bias
Dependant on the quality of studies included and low quality studies can bias the summary effect size
Dependant on data availability
what is a systematic review?
Its an overview that answers a specific clinical question and contains a thorough, unbiased search of the relevant literature, explicit criteria for assessing studies and a structured presentation of the results
It may or may not use statistical techniques, such as meta-analysis.
Whats are the pros and cons of systematic review?
Pros:
- Conducts a thorough search for evidence which can eminiate publication bias, language bias and selection bias
- Increases total sample size which improves power = improves certainty and precision
- replicable and transparent which increases the credibility and trustworthiness
- identifies gaps in literature
Cons:
- time consuming and resource intensive
- limited by data availability
- limited by study design
- risk of bias in selection and interpretation of studies
What is the correlation coefficient?
a number between +1 and −1
a statistical measure of the strength of a linear relationship between two variables
Whats the most appropriate study type to answer a question on diagnosis?
Cross-sectional analytic study
Whats the most appropriate study type to answer a question on aetiology?
Cohort study>Case-control study
Whats the most appropriate study type to answer a question on prognosis?
Cohort study > case control
Whats the most appropriate study type to answer a question on treatment?
RCT or systematic review of RCT ? Cohort > case control
What is treatment fidelity? Why is it important?
the degree to which an intervention or treatment is implemented as intended or designed. It ensures its conducted consistently and reliably
Treatment fidelity is important because it allows researchers to evaluate the effectiveness of an intervention and determine whether any observed effects are due to the treatment itself, rather than differences in how the treatment was delivered
What are the Bradford-Hill criteria?
Aset of nine criteria to provide epidemiologic evidence of a causal relationship between a presumed cause and an observed effect
- strength - the large the association, the more likely it is causal
- consistency - consistent findings in different studies strengthens likelihood of causality
- specificity - the more specific the association is the bigger the probability of causal effect
- temporality - the effect can to occur after the cause
- biology gradient - greater exposure should generally lead to greater incidence of effect
- plausibility - a plausible mechanism between cause and effect is helpful
- coherence - coherence between epidemiological and lab findings increases the likelihood
- experiment
- analogy - use of analogies or similarities between observed association
What is confounding?
A third factor which is associated with both the exposure and outcome that can potentially distort the association between outcome and exposure
What happens if follow-up was not complete enough in a cohort study?
Allows for selection bias and may reduce the power of the study
It can introduce selection bias because participants who are lost to follow-up may differ systematically from those who remain in the study, and this can affect the validity of the results.
What is propensity score matching?
a statistical technique that can be used to balance the distribution of confounding variables between the exposed and unexposed groups.
In an observational study, participants are not randomly assigned to treatment or control groups, which can lead to confounding factors that may affect the outcome of interest. Propensity score matching involves estimating the probability of receiving the treatment based on observed covariates or variables that may be related to both treatment assignment and the outcome. The propensity score is then used to match participants in the treatment group with participants in the control group who have similar scores, based on a specified level of similarity. This creates a comparison group that is more similar to the treatment group in terms of the distribution of potential confounding variables, which can reduce bias and improve the validity of the study results.
What is a sensitivity analysis?
assessing the robustness of the results to different assumptions about the relationship between exposure, outcome, and potential confounding variables. This can help to identify the extent to which the results are affected by confounding and can provide insights into the potential bias in the estimates.
In a sensitivity analysis, researchers test the sensitivity of the results by varying one or more aspects of the study design or analysis and assessing the impact on the outcome. This can help to identify the key factors that influence the results and provide insight into the stability and validity of the findings.
What is heterogeneity?
A term used in meta-analyses and systematic reviews to describe when the results of a test or treatmentdiffer significantly in different studies.
Such differences may occur as a result of differences in the populations studied, the outcome measures used or because of different definitions of the variables involved.
What is a case report?
An uncontrolled observational study involving an intervention and an outcome in a single patient.
What is case series?
Reports of several patients with a given condition, usually covering the course of the condition and the response to treatment. There is no control group of patients.
What is Cochrane library?
A regularly updated collection of evidence-based health databases including the Cochrane Database of Systematic Reviews, which contains reviews on a variety of health topics.
What is comparability?
Similarity of groups in terms of characteristics likely to affect study results (such as health status or age).
What is a control group?
A group of people in a study who do not have the intervention or test being studied. Instead, they may have the standard intervention (sometimes called ‘usual care’) or a dummy intervention (placebo). The results for the control group are compared with those for a group having the intervention being tested. The aim is to check for any differences.
What is a critical appraisal?
Reviewing a study to judge the quality of the method used and the reliability of the conclusions.
What is analysis and what are the 2 types?
The process of looking for patterns in information to identify cause and effect or answer specific questions, such as whether a treatment or other intervention works and what the risks are.
There are 2 types of analysis. Quantitative analysis looks for patterns in the form of numbers. Qualitative analysis looks for patterns of meaning, feeling or beliefs.
What is applicability?
How well an observation or the results of a study or review are likely to hold true in a particular setting.
What are before-and-after studies?
An approach in which dependent variables are measured before and after an intervention has been delivered. Often called a pre–post study. The people in the pre- and post-intervention stages can either be the same or different.
What is grey literature?
Literature that has not been formally published in sources such as books or journal articles.
What is the hierarchy of evidence?
Study types organised in order of priority, based on the reliability (or lack of potential bias) of the conclusions that can be drawn from each type.
What is homogeneity?
A term used in meta-analyses and systematic reviews to indicate that the results of studies are similar; the opposite of heterogeneity. Study results are also regarded as homogeneous if any differences could have occurred by chance.
What is a longitudinal study?
A study of the same group of people at different times. This contrasts with a cross-sectional study, which observes a group of people at a point in time.
What is a multi centre study?
A study in which people selected to take part come from different locations or populations. For example, from different hospitals or even different countries.
What is an observational study?
A retrospective or prospective study in which the investigator observes the natural course of events with or without control groups (for example, cohort studies and case–control studies).
What is a placebo?
A fake (or dummy) treatment given to patients in the control group of a clinical trial. It is indistinguishable from the actual treatment (which is given to patients in the experimental group). The aim is to determine what effect the experimental treatment has had - over and above any placebo effect caused because someone has had (or thinks they have had) care or attention.
What is reliability?
The ability to get the same or similar result each time a study is repeated with a different population or group.
What is reverse causation?
Where the direction of causal effect is from the outcome (or symptoms thereof) to the intervention or exposure.
What is a systematic error?
Refers to the various errors or biases inherent in a study.
consistent, repeatable error associated with faulty equipment or a flawed experiment design
What happens if follow-up was not long enough in a cohort study?
May mean relevant outcomes are not observed and associations are not detected. This leads to bias towards the null hypothesis (type 2 statistical error) and underestimation of oitcome&exposure association
Was the cohort recruited in an acceptable way?
Was the cohort representative of a defined population?
Was there something special about the cohort?
Was everybody included who should have been?
Look for selection bias which can compromise the generalisability of the findings
Was the exposure accurately mesaured in a cohort study to minimise bias?
Look for measurement or classification bias
- measurement bias - observer bias or instrument bias
- classification bias - non-differential and differential error
- did they use subjective or objective measurements
- do the measurements truly reflect what you want them to (validity)
- were all subjects classified in to exposure groups using the same procedure?
Make reference to type 1and type 2 errors!
Was the outcome of a cohort study accurately measured to minimise bias?
Look for measurement or classification bias
- did they use subjective or objective measurements
- do the measurements truly reflect what you want them to (validity)
- has a reliable system been established for detecting all cases?
- were the measurement methods similar in the different groups?
- blinding?
Make reference to type 1and type 2 errors!
Have the authors identified all important confounding factors in a cohort study?
Definition of confounding
List important confounding variables
Look for restriction in design and techniques to adjust for confounding e.g. modelling, stratified, sensitivity analysis etc
Was the follow up of a cohort study complete enough? And was it long enough?
Long enough for exposure to manifest?
Pt lost to follow up?
Were the cases in a case-control study recruited in an acceptable way?
Look for selection bias
- are the cases representative of a defined population
- was there an established reliable system for selecting all the cases
- is there something special about the cases?
- si the time frame of the study relevant to disease or exposure?
- was there a sufficient number of case selected?
- was there a power calculation?
We’re the controls in a case-control study selected in a suitable way>
Look for selection bias!
-were controls representative of a defined population
- was there something special about controls
- was the non-response high
- are they matched, population-based or randomly selected
- was there a sufficient number of controls selected
(Features needed for controls to be valid are: Selected from a defined population base of the study
and they should not have the disease or outcome being studied)
Was the exposure accurately measured to minimise bias in a case-control study?
Look for measurement, recall or classification bias
- was the exposure clearly defined and accurately measured?
- did the authors use subjective or objective measurement?
- do the measures truly reflect what they are supposed to> (validity)
- were the measurement methods similar in case and controls
- blinding?
- temporality?
Was the assignment of participants to interventions randomised in RCT?
How was randomisation done? Was the method appropriate?
Was it sufficient to eliminate systematic bias?
Was allocation sequence concealed from investigators and participants?
Were all participants who entered the RCT accounted for at its conclusion?
Were losses to follow-up and exclusions after randomisation accounted for?
ITT?
Was the study stopped early?
We’re the study groups similar at the start of the RCT?
Similar baseline characteristics?
Apart from the experimental intervention, did each study group recieve the same level of care in the RCT?
Was there a clearly defined study protocol
We’re any additional tests done?
We’re the follow-up intervals the same?
Whats the issue with block randomisation?
If someone knows the size of the block, they can predict which groups patients will be put in to
What is sampling bias?
Subjects are not representative relative to general population so the results are not generalisable
What is late-look bias?
When information is gathered at an inappropriate time e.g. using a survey to study a fatal disease… only pt alive will be able to answer the survey!!
How do you calculate power?
1- probability of making a type 2 error
How can you assess for publication bias in a meta-analysis?
Using a funnel plot
What is a funnel plot?
A graph that displays the studies included in the met analysis in a plot of effect size against sample size
The expected picture is one of a symmetrical inverted funnel as smaller studies have more chance variability than larger studies
If the plot is asymmetric the metanalysis may have missed some trials
Why is a sensitivity analysis important in a meta analysis?
By conducting a sensitivity analysis, the researcher can test the sensitivity of the results to changes in various factors such as study quality, inclusion criteria, and statistical methods. This helps to identify potential sources of bias or heterogeneity that may affect the conclusions of the meta-analysis and provide a more nuanced interpretation of the findings. A sensitivity analysis can increase confidence in the results of the meta-analysis and help to inform decisions about the validity and reliability of the findings.
How is data from a meta-analysis typically pictorially represented?
Using a forest plot
What does a square represent on a forest plot?
The findings from each individual study
The size of the square is proportional to the precision of the study
What does a horizontal line on each square represent on a forest plot?
The 95% confidence interval which represents the uncertainty of the estimate of the treatment effect i.e. a wider line means less certainty about the result
If the line passes through the vertical line of no effect then it means the study is not statistically significant
What does the diamond represent on a forest plot?
The aggregate effect size obtained by combining all the studies
The width of the diamond shows the 95% confidence intervals
If the diamond crosses the vertical line of no effect it means overall there is no statistical significance
What does the vertical central line represent on a forest plot?
The line of no effect
It means risk and benefit are equal
Any statistical significant study does not cross this line!
Why are confidence intervals preferable to P-values?
As they tell us a range of possible effect sizes compatible with the data whereas p-values simply provide a cut-off beyond which we assert the findings are statistically significant
What is a funnel plot used for?
To check the existence of publication bias in systematic reviews and meta-analyses
What are limitations of the funnel plot?
It can be diffiucult to detect asymmetry by eye (formal statistical methods have been developed now to test for heterogeneity)
They assume that the distribution of effect sizes should be symmetrical around the summary estimate but this assumption may not hold true!
They can only detect publication bias and cannot distinguish between this and other sources of heterogeneity
May not be informative when the number of studies included in the meta-analysis is small as there may not be enough data points to detect asymmetry
What is Egger’s test?
A regression test used to test for publication bias
It tests whether small studies tend to have larger effect sizes than would be expected i.e. implying that small studies with small effects have not been published
What is the Cox Model?
A type of regression model used in survival analysis to investigate the relationship between a set of predictor variables and time-to-event outcome
Using this may narrow the confidence intervals and therefore improve the estimate of treatment effect
What are some statistical analysis tools useful for modelling the relationship between 1 or more predictor variables and an outcome variable?
Regression analysis e.g. linear regression and logistic regression
What are some statistical analysis tools useful for analysing time-to-event data e,g. Time until a pt dies?
Survival analysis e.g. Cox hazards model
What are some statistical analysis tools useful for estimating an overall effect size of multiple stydies??
Meta-analysis
What are some statistical analysis tools useful for assessing the impact of potential sources of bias on the results of a study?
Sensitivity analysis
Subgroup analysis
What is the correlation coefficient?
Indicates how closely the points lie to a line drawn through the plotted data and gives information about how one variable will change as another is change
It’s denoted by the value r which may lie between -1 and 1
R=1 strong positive correlation
R=0 no correlation
R=-1 strong negative correlation
What is linear regression?
a linear approach for modelling the relationship between a scalar response and one or more explanatory variables
It can predict how much one variable will change if another is changes
Y= a + bx
Where y= variable being calculate, b - regression coefficient, a= intercept value when x=0 and x is the second variable!
How do you calculate standard deviation?
Square root of variance which is the average of the squared differences from the mean
What is standard error of the mean?
Standard deviation / square root of sample size
(This gets smaller as sample size increases)
How do you calculate the lower limit of the 95% CI?
Mean - (1.96 x standard error of the mean)
Note: if sample size is small then use students T critical value table to find the number to replace the 1.96
How do you calculate the upper limit of the 95% CI?
Mean + (1.96 x standard error of the mean)
Note: if sample size is small then use students T critical value table to find the number to replace the 1.96
What is parametric data?
Data that is normally distributed into a bell-shape curve
What is paired data?
Data that comes from a. Single group of patients e.g. measurement before and after an intervention
What significant test would you do for paired or unpaired parametric data?
Student’s T test
What is Pearson’s product moment coefficient?
a statistical measure that indicates the strength and direction of the linear relationship between two continuous variables. It is denoted by the symbol “r” and ranges from -1 to +1, where -1 represents a perfectly negative correlation, 0 represents no correlation, and +1 represents a perfectly positive correlation.
What is the student’s T test?
a statistical test used to determine if there is a significant difference between the means of two independent groups
It assumes the data is normal distributes and that the variances of the 2 groups are equal
It provides a P value
What significant test would you do for unpaired non-parametric data?
Mann-Whitney U test
What is the Mann-Whitney U test?
a non-parametric statistical test used to compare the medians of two independent groups. It is a non-parametric alternative to the independent samples t-test
It works by ranking the combined data from both groups, assigning ranks from smallest to largest and then calculating the sum of the ranks for each group and comparing the sums to see if there is significant difference between the groups
It generates a U statistic
What significant test would you do for paired non-parametric data?
Wilcoxon signed-rank test
What is the Wilcoxon signed-rank test
A non-parametric statistical test that is used to compare the medians when data does not meet the assumptions required for parametric tests
It works by calculating the differences between the paired observations, assigning ranks to the absolute values of these differences, and then calculating the sum of the positive ranks and the sum of the negative ranks. The test then compares the sums to see if there is a significant difference between the two related samples.
It generates a signed-rank statistic
Which significance test is used to compare proportions or percentages of non-parametric data?
Chi-squared test
What is the chi-squared test?
A statistical test used to determine if there is a significant association between two categorical variables
Which significance test is used to measure correlation in non-parametric data?
Spearman’s/Kendall rank test
What is a box-and-whisker plot?
A graphical representation of the sample minimum, lower quartile, median, upper quartile and sample maximum values
What is a histogram?
a graphical display of continuous data where values have been categorised
commonly used in statistical analysis to visualize the distribution of continuous or discrete data
What is a Kaplan-Meier survival plot?
the probability of surviving in a given length of time while considering time in many small intervals
The curve is constructed by plotting the survival function against time.
What are the 2 mechanisms that produce misclassification errors?
Differential misclassification (non-random) - erros occur with greater frequency in one of the study groups e.g. recall bias and surveillance bias (expecting a particular outcome, you may be more likely to diagnose it)
Non-differential misclassification (random) - frequency of errors is approximately the same in groups being compared - i..e exposure is not measured properly causing a misclassification. This dilutes the effect so the effect estimates will be closer to the null
Give a general example of how the way in which a cohort is recruited might influence the validity of the study?
Selection of the cohort will have particular importance for external validity as observed associations may not be generalisable in different cohorts
Cohort should be representative of the population being studies
Explain why in general inaccuracy in exposure measurements could lead to biased results?
Non-differential error (random misclassification) - bias towards null and hides true relationships - type 2 errors
Differential error (non-random misclassification i.e. affects one group more) - could bias in any direction creating spurious associations or obscuring true ones
Explain why inaccuracy in outcome measurements could lead to biased results?
If outcome measurement reporting was uniformly poor e.g. missing most disease outcomes - trend towards null = type 2 error
If outcome measurement is more accurate in one exposure group you my get spurious positive relationship (type 1 error)
What are some features of a study’s methods that you might look for which could enhance the accuracy of outcome measurement?
Standard procedures for outcome measurement
Validation of algorithms used
Comprehensive population coverage
Blinding - Outcome established blind to exposure category to minimise observer bias
Multiple and independent assessors
What would be the consequences for a research study if all important confounders were not identified?
The association between exposure and outcome could be biased - (it could over or underestimate the true effect size, reduce statistical power and limit external validity )
What evidence can you look for to check confounding factors were taken into account in a cohort study?
Adjustment
Statistical analysis
Explain why the completeness of a cohort study might influence the validity of a study?
Incomplete follow up allows for selection bias as differential loss to follow up amongst different groups can lead to systematic differences between characteristics of study groups
Incomplete follow up may reduce the power of the study as it can reduce the effective sample size
Explain why the length of follow up of a cohort study might influence the validity of a study?
Insufficient length of follow-up may mean that relevant outcomes are not observed and associations are not detected = type 2 error
What information do you need to judge if a study is sufficiently long enough?
An indication of the likely time from exposure to development of clinical outcomes
What are some examples of information you might look for in a cohort study research paper to help you answer “can results be applied to your local population”?
Description of study cohort to allow comparison with local population
Study setting and location
Consideration of whether exposure pattern was likely to be similar to local population
Information to allow assessment of internal validity as if it is not valid then its not possible to apply externally
Identify some advantages of systematic reviews over other research designs?
- larger sample size which increases the power of the study
- Rigorous summary of all research evidence that related to a specific question
- Disadvantages of single studies can be guarded against e.g. inadequate sample size, results varying due to chance, differences in designs of studies
Why can excluding publications not in English bias the findings of a systematic review?
As relevant publications might be published in non-English journals
The quality of RCT and observational studies were assessed by 2 authors and disparities resolved by discuss. Outline 3 reasons why this is a good practice
- 2 authors - reduces the chance that errors will be made
- Assessing quality is important as studies with weaker designs will be less valid and overestimate effects
Why are RCTs less subject to bias than observational studies?
- Groups are likely to be similar so we can be more confident that any observed differences in outcome are due to the intervention
- Randomisation - similar groups at baseline - minimises selection bias and confounders
- RCTs usually have a more standardized and controlled study design, including standardized interventions and outcome measurements, which further reduces the risk of bias.
- blinding of participants, investigators, and outcome assessors can help to minimize the risk of measurement bias.
What is meta-analysis?
A statistical technique for quantitatively combining the results of multiple studies that measure the same outcome into a summary estimate
What is 95% CI?
The range between 2 values in which wthere is a 95% probability that the true value lies for the whole population of pt from which the study pt are selected
Define the term odds ratio?
A ratio of the odds of an event in the exposed group to the odds of the same event in a group that is not exposed
Give an explanation of the term confounding?
An apparent relationship between exposure and outcome that is false; it implies a causal relationship when one does not exist
What is the primary purpose of controls in a case-control study?
To provide a comparison group that is similar to the cases with respect to certain characteristics but do not have the outcome.
They serve as a reference groups against which the cases are compared so that researchers can assess association between exposure and outcome
They minimise bias and confounding which improves the validity
What is the hazards of simply performing a quick MEDLINE search?
Complex and sensitive search strategies are needed to identify all potential trials
Some clinical trials will be missed if they are published in an journal not listed by MEDLINE
What are some potential sources for finding studies?
MEDLINE
Cochrane Controlled Trials Registry
EMBASE
Bibliographies of studies
Review articles and textbooks
Symposia proceedings
Pharmaceutical companies
Direct communication with experts in the field
What is the largest and most complete database of clinical trials in the world?
Cochrane Controlled Trials Registry
What is language bias?
Studies with positive findings are more likely to be published in an English-language journal and also more quickly than those with inconclusive or negative findings
How many reviewers should check a systemic review and how should differences be settled?
At least 2 reviewers
Differences should be settled by consensus or by a third-person arbitrator
If large areas of diasagreement then the reader should question the validity of the review
What are the 2 main statistical models used to combine results for a meta-analysis?
Random effects model
Fixed effects model
What is that random effects model?
This assumes that the different studies’ results may come from different populations with varying responses to treatment
Used for heterogeneous studies
What is that fixed effects model?
This assumes that each trial represents a random sample of a single population with a single response to treatment
Used for homogeneous studies
What is statistical heterogeneity?
Variability in the intervention effects being evaluated in the different studies
This is a result of clinical or methodological heterogeneity
What is clinical heterogeneity?
Variability on the participants, interventions and outcomes studied
What is methodological heterogeneity?
Variability in study design and risk of bias
what test tests for statistical heterogeneity in systematic review?
Chi-squared test
What symbol represents the Chi-squared test?
Q
How should you interpret chi squared results for heterogeneity?
Use I squared:
0% to 40%: might not be important;
30% to 60%: may represent moderate heterogeneity;
50% to 90%: may represent substantial heterogeneity;
75% to 100%: considerable heterogeneity*.
What can be used to incorporate heterogeneity among studies?
Perform a random-effects meta-analysis
Strategies for addressing heterogeneity?
Check data is correct
Dont do a meta analysis
Explore the heterogeneity via subgroup analyses or meta regression
Ignore heterogeneity and used a fixed-effects meta-analyses
Perform a random-effects meta-analysis
Change the effect measure
Exclude studies
In a systematic review, what condition must be met before individual study results can be combined in a meta-analysis?
The studies must have no or limited heterogeneity
what statistics can be used to quantify heterogeneity in studies
Cochran’s Q test which is based on chi-square test
I squared index
