Clinical studies

Question 1

What is a p-value? And what can be determined about p-values in relation to significance of a result?

Answer 1

It determines the likelihood that something is down to chance. P = 0.05 suggests that the probability of a result being the result of chance is 5%. P < 0.05 (less than 0.05) is used as a test of significance, whereby if p<0.05 (less than), the probability the result is due to chance is less than 5%, and the result is significant.

Question 2

What are confidence intervals?

Answer 2

Repeated sampling will produce a spread of estimates around the true value, and confidence intervals can be used to show where 95% of sample estimates will lie; this allows us to be confident that the true value lies somewhere within this interval e.g. +/- 95mmol/mol.

Question 3

What is a clinical trial?

Answer 3

Planned experiment in humans to measure the effectiveness of an intervention. There is a comparator (placebo) and inclusion and exclusion criteria.

Question 4

What is the type of study used in a clinical trial? (x3 points)

Answer 4

NOT observational. Randomisation. Double-blind.

Question 5

What are the four phases of a clinical trial?

Answer 5

PHASE ONE – SAFETY: tests on the safety of treatment performed on healthy volunteers. PHASE TWO – BIOLOGICAL EFFECT: could it actually work? A few hundred with the condition are recruited. Safety is reviewed, and look at whether treatment is effective in the short term. PHASE THREE – OVERALL EFFECTIVENESS: will it work in the real world? Compares the treatment with a placebo: studies effectiveness and side effects. Needs several thousand patients. PHASE 4 – AFTER MARKETING: the effects are measured in the various populations. Rare side effects that were not picked up earlier are identified.

Question 6

What is allocation bias?

Answer 6

Volunteers may not be randomly allocated to placebo and treatment groups, and as such confounding variables may influence the results.

Question 7

What is measurement bias?

Answer 7

If the investigation team has pre-conceived ideas or vested interest, they may subconsciously interpret the results differently, so double-blinding is key to ensure unbiased measurement and interpretation.

Question 8

What is reporting bias?

Answer 8

Negative and neutral trials are unlikely to be reported, especially if commercially embarrassing to a pharmaceutical company.

Question 9

What is analysis bias?

Answer 9

If differential discontinuation occurs (more participants withdraw from one group than the other e.g. because of side effects), this can skew the analysis to show a larger than accurate benefit in that group.

Question 10

What are the two types of error in a clinical trial?

Answer 10

Type 1 error – false positives (alpha). Type 2 error – false negatives (beta).

Question 11

How is power measured in relation to the errors of a clinical trial?

Answer 11

1 – beta (%).

Question 12

What is the typical value for power in a clinical trial?

Answer 12

80-90%.

Question 13

What does the power of a test tell us (in relation to errors in a clinical trial)?

Answer 13

It is the probability that the test will find a significant difference between two populations.

Question 14

What is Experiment Event Rate? How is it calculated?

Answer 14

EER: incidence of a NEGATIVE OUTCOME (response to drug, adverse event, death) in the experimental group (the group getting the drug). Calculated by: look at photo.

Question 15

What is Control Event Rate? How is it calculated?

Answer 15

CER: incidence of a negative OUTCOME in the control group (the group getting the placebo). Calculated by: look at photo.

Question 16

How is relative risk calculated in relation to EER and CER?

Answer 16

EER/CER.

Question 17

How is relative risk reduction calculated in relation to EER and CER?

Answer 17

(CER-EER)/CER.

Question 18

What is relative risk reduction?

Answer 18

It is the relative decrease in risk of an adverse event in the exposed group compared to the unexposed group.

Question 19

What is absolute risk reduction?

Answer 19

It is the difference between the risk of an outcome in the exposed group and the unexposed group.

Question 20

How is absolute risk reduction calculated?

Answer 20

CER-EER.

Question 21

What is the ‘number needed to treat’ defined as, and how do you calculate it?

Answer 21

Number you need to treat in order to avoid one additional negative outcome/event (negative outcome is usually in relation to the control group). Calculated by 1/ARR (absolute risk reduction). Smaller value indicates more effective treatment. Just think of all the calculations involved.

Question 22

What would the relationship between EER and CER be if an intervention was effective?

Answer 22

EER would be LESS than CER.

Question 23

What are case-control studies?

Answer 23

Start with a source population, containing a group of cases and group of controls – where the control group are representative of the case group (e.g. both groups have same ethnicity, sex…). Looking retrospectively, compare the odds of exposure in those with the disease and those without, to see if those that have the disease were more likely to have been exposed – this will produce an ODDS RATIO.

Question 24

What are cohort studies?

Answer 24

These can be performed retrospectively (possibility of bias, cheaper) or prospectively (less bias, expensive), and involve starting with a HEALTHY cohort of volunteers (in the case of prospective), and following them up over time to study the effect of exposure on outcomes. The cohort is divided at analysis to the exposed and unexposed, and then the risk of the outcome in question is studied to provide a RELATIVE RISK.

Question 25

What are the advantages and disadvantages of case-control studies? (x3 and x4)

Answer 25

ADVANTAGES: good for rare disease with fewer cases, quick and cost effective, can investigate many exposures simultaneously. DISADVANTAGES: recall bias because performed retrospectively (and because may introduce ideas to those with disease risking explanation), poor for rare exposures, uncertainty of exposure-disease time relationship, difficult to select controls that are representative of the population of cases (can use friends/family).

Question 26

What are the advantages and disadvantages of cohort studies? (x3 and x3)

Answer 26

ADVANTAGES: can study multiple outcomes simultaneously, can follow the natural history of the disease (for prospective studies), prospective studies eliminate bias. DISADVANTAGES: inefficient for rare diseases, expensive and time consuming, loss to follow-up can introduce bias.

Question 27

What are case-control studies appropriate for?

Answer 27

Are appropriate for the determination of what causes rare disease, but are not as good evidence as an RCT because they are observational and may be strongly affected by confounders. These are only appropriate to SUGGEST AN ASSOCIATION BETWEEN EXPOSURE AND DISEASE – that must then be investigated further.

Question 28

What are the cohort studies appropriate for?

Answer 28

Are higher on the hierarchy of evidence, and a relative risk carries more than an odds ratio. These are appropriate to ANSWER RESEARCH QUESTIONS THAT SEEK TO QUANTIFT THE EFFECT THAT EXPOSURE HAS ON SPECIFIC OUTCOMES. It can be used to confirm or challenge suspected associations e.g. from case control studies.

Question 29

What carries more weight: relative risk or odds ratio?!

Answer 29

Relative risk!

Question 30

What is routine data?

Answer 30

Collected routinely without a research question in mind. Often for administrative or statutory purposes.

Question 31

What is coding in relation to collecting data?

Answer 31

Health outcomes such as deaths, adverse effects, admissions… are given unique codes under a system such as ICD10 to allow for classification and comparison.

Question 32

What are the disadvantages of coding for collecting data? (x2)

Answer 32

The codes can change. Reporting is not always standardised. May be different in different areas, so interpretation can differ - isn't this what standardisation means anyway?

Question 33

List five types of routine data in healthcare.

Answer 33

Healthcare outcomes e.g. deaths, admissions… Exposures and Health determinants e.g. pollution, crime… Demographic e.g. census. Geographical e.g. location of healthcare facilities. Health Service Provision e.g. staff/bed counts.

Question 34

What main health outcome data is routinely collected in the UK? (x9)

Answer 34

MORTALITY – sent to ONS for processing. CANCER – voluntary notification to local registry, but also noted on death certificates. NOTIFICATION OF INFECTIOUS DISEASE – reported by doctors to local authority for monitoring. TERMINATIONS – must be recorded. CONGENITAL ABNORMALITIES. HOSPITAL EPISODE STATISTICS – contains details of all admissions. QOFS – pay per performance for GPs – rewards practices for quality of care, determined by how well doctors manage chronic disease. RTAs.

Question 35

What are the three major sources of health data in the UK?

Answer 35

Census. Health Survey for England. NHS staff/patient surveys.

Question 36

What are the strengths and weaknesses of the three major sources of health data in the UK? (x4 and x5)

Answer 36

Census, Health Survey, NHS staff/patient surveys. STRENGTHS: cheap, data already collected and in a standardised fashion, relatively comprehensive with entire population coverage, wide range of recorded data available for a large number of years. WEAKNESSES: may not answer the question, incomplete ascertainment if not every case is captured, variable quality (based on the use of coding/diagnosis), variable validity, disease labelling/coding may change over time or by area.

Question 37

What are standardised mortality rates?

Answer 37

Ratio between the number of deaths in a study population and the number that would be expected, accounting for age and sex.

Question 38

How are standardised mortality rates calculated?

Answer 38

Observed deaths/expected deaths.

Question 39

What would standard mortality rates of less than 1, 1, and more than 1 denote?

Answer 39

LESS THAN 1: fewer than expected number of deaths. 1: same as expected. MORE THAN 1: more deaths than expected.

Question 40

What is the clinical significance of using standardised mortality rates?

Answer 40

Age standardised death rates can be given for a given condition or for the population as a whole. Can be used to assess the effectiveness of a country’s health system and compare populations e.g. whether they are dying more of one disease etc. Remember, that SMRs compare observed rates of mortality to expected rates of mortality.

Question 41

Why should standard mortality rates be treated with caution?

Answer 41

Differences in reporting, diagnostics, coding and behaviour can all affect SMRs and cause ‘alarming increases’ that do not actually translate to massively increased incidence or mortality.

Question 42

What is the purpose of a systematic review?

Answer 42

Seek to answer a very specific question; all published literature is considered, and studies that are appropriate are selected and evaluated, and the data is reported to provide a definitive answer.

Question 43

What is the purpose of a meta-analysis?

Answer 43

Use of statistical methods to summarise the results of the selected studies, weighting by how well performed and powerful the study was; generates a pooled risk estimate. Meta-analyses are conducted in the context of a systematic review.

Question 44

How can publication bias be assessed?

Answer 44

Publication bias can be assessed with a funnel plot.

Question 45

What is PICOS?

Answer 45

Research question should define a Population, Intervention, Comparison, Outcome(s), Study Design.

Question 46

Give an example of a database of systematic reviews?

Answer 46

Cochrane library.

Question 47

What is the process used to write a Systematic Review?

Answer 47

1. Plan the review using PICOS. 2. Identify the research available using databases such as Medline, looking at referenced papers, and at research that was not published; then select studies that meet the criteria for inclusion and quality assess these. 3. Use a meta-analysis to generate a pooled risk estimate, weighing the effect of size of each study to produce a forest plot. 4. Report study details in a meaningful way – use guidelines such as STROBE/STRAND.

Question 48

What is a forest plot?

Answer 48

Used to show the pooled risk estimate. Based on the weighted risk ratios of the included study.

Question 49

What are the limitations of systematic reviews/meta analyses? (x3)

Answer 49

Incompleteness – may not be able to access all data due to lack of publication or language barrier etc. e.g. some reviews just study papers written in English. Publication bias. Heterogeneity – studied differ with respect to PICOS – studies may differ based on clinical differences and unknown study characteristics – could affect results between studies.

Question 50

How can heterogeneity be measured?

Answer 50

I^2 is a measure of heterogeneity. 0% indicates something is fully homogenous. 50% indicates moderate heterogeneity.

Question 51

How can heterogeneity in systematic reviews be accounted for? (x2)

Answer 51

Can use meta-regression and sensitivity analysis.

Question 52

What is a funnel plot? What does it show?

Answer 52

Shows the relationship between sample size and risk ratios. Tends to converge to a point at the largest sample size = the true risk ratio. Any gaps lower down indication publication bias.

Question 53

How should a critical appraisal for a systematic review be carried out? (x7 stages)

Answer 53

1. Clear, predefined question – PICOS. 2. Was a comprehensive search for relevant literature carried out? 3. Was methodological quality of each study assessed appropriately? 4. Was heterogeneity explored? 5. How credible is the evidence? 6. Check guidelines for reporting.

Question 54

How should a critical appraisal for a meta-analysis be carried out?

Answer 54

1. Was heterogeneity explored? 2. Was publication bias an issue? 3. Was it appropriate to pool the studies? i.e. were they sufficiently homogenous? 4. Was the appropriate model used to pool effect estimates? Fixed versus random effects model. 5. Did different sub groups of studies give similar results? How generalisable are the findings?

Question 55

What is morbidity?

Answer 55

The number of cases of ill health, complications and side effects attributed to a particular condition over a given period i.e. the condition of being diseased. RELIEF treatment can reduce morbidity.

Question 56

What is the duration of a disease defined as?

Answer 56

Time from diagnosis to death/recovery. IMPORTANT - this differes from what others may think e.g. moment you get symptoms until pathogen presence in body is gone.