Different types of clinical evidence and study design

Different types of clinical evidence and study design

The research question

What is a research question?

The researcher asks a very specific question and tests a specific hypothesis. Broad questions are usually broken into smaller, testable hypotheses or questions.

Often called an objective or aim, though calling it a question tends to help with focusing the hypothesis and thinking about how to find an answer

What makes a poor research question?

a question that matters to nobody, even you

hoping one emerges from routine clinical records

the records will be biased and confounded
they’ll lack information you need to answer your question reliably, because they were collected for another reason

fishing expedition/data dredging – gathering new data

and hoping a question will emerge


What makes a good question?
Feasible (answerable with a robust method)
Interesting
Novel
Ethical
Relevant

FINER criteria

Real research questions

Is five days’ treatment with

injectable ampicillin plus gentamicin
more effective than
chloramphenicol in children under 5
with very severe pneumonia in low
resource settings?

What is the prevalence of HIV

infection in India, and how many
premature deaths does it cause?

How to focus your question

brief literature search for previous evidence

discuss with colleagues

narrow down the question – time, place, group

what answer do you expect to find?

Turning a research question into a proposal

who am I collecting information from?
what kinds of information do I need?
how much information will I need? *
how will I use the information?
how will I minimise chance/bias/confounding?
how will I collect the information ethically?


* sample size – ask a statistician for help
http://www.bmj.com/collections/statsbk/13.dtl

Minimising bias and confounding

Chance - measurements are nearly always subject to random
variation. Minimise error by ensuring adequate sample size and
using statistical analysis of the play of chance

Bias - caused by systematic variation/error in selecting patients,

measuring outcomes, analysing data – take extra care

Confounding - factors that affect the interpretation of outcomes

eg people who carry matches are more likely to develop lung
cancer, but smoking is the confounding factor – so measure likely
confounders too

Ethical issues – the wider aspects

what information to give before seeking consent?
deviation from normal clinical practice?
what full burden will be imposed on participants?
what risks will participants/others be exposed to?
what benefit might participants or others receive?
how might society/future patients benefit in time?
might publication reveal patients’ identities?


Exactly what are you planning to do?
PICO

P - who are the patients or what’s the problem?

I - what is the intervention or exposure?
C – what is the comparison group?
O - what is the outcome or endpoint?

More on PICO

Patients
disease or condition
stage, severity
demographic characteristics (age, gender, etc.)
Intervention
type of intervention or exposure
dose, duration, timing, route, etc.
Comparison
risk or treatment
placebo or other active treatment
Outcome
frequency, risk, benefit, harm
dichotomous or continuous
type: mortality, morbidity, quality of life, etc.


Study designs

Population (P) Outcomes (O)

Interventions (I) or Exposures (E)

Centre for Evidence Based Medicine, Oxford, UK www.cebm.net

Are you going to observe or experiment?

observational – cross sectional, case series, case-control studies,
cohort studies
identify participants
observe and record characteristics
look for associations

experimental – before and after studies, comparative trials

(controlled or head to head), randomised trials (ditto)
identify participants
place in common context
intervene
observe/evaluate effects of intervention


Pros and cons of the RCT
An experimental comparison study where participants are allocated to
treatment/intervention or control/placebo groups using a random
mechanism. Best for studying the effect of an intervention.

Advantages:

unbiased distribution of confounders
blinding more likely
randomisation facilitates statistical analysis
Disadvantages:
expensive: time and money
volunteer bias
ethically problematic at times

Pros and cons of crossover trial

A controlled trial where each participant has both therapies
e.g is randomised to treatment A first then starts treatment B.

Advantages:

all participants serve as own controls and error variance is reduced, thus reducing sample size needed
all participants receive treatment (at least some of the time)
statistical tests assuming randomisation can be used
blinding can be maintained
Disadvantages:
all participants receive placebo or alternative treatment at some point
washout period lengthy or unknown
cannot be used for treatments with permanent effects


Pros and cons of cohort study
Data obtained from groups who have already been exposed, or not
exposed, to the factor of interest. No allocation of exposure is made by the
researcher. Best for studying effects of risk factors on an outcome.
Advantages:
ethically safe
participants can be matched
can establish timing and directionality of events
eligibility criteria and outcome assessments can be standardised
Disadvantages:
controls may be difficult to identify
exposure may be linked to a hidden confounder
blinding is difficult
for rare disease, large sample sizes or long follow-up necessary

Cohort study

Chronic kidney disease and risk
of major cardiovascular disease
and non-vascular mortality:
Prospective population based
cohort study.


Di Angelantonio E, et al.
BMJ 341:doi:10.1136/bmj.c4986

Pros and cons of case-control study

Patients with a certain outcome or disease and an appropriate group of
controls, without the outcome or disease, are selected (usually with some
matching) then information is obtained on whether the subjects have been
exposed to the factor under investigation.
Advantages:
quick and cheap as fewer people needed than cross-sectional studies
only feasible method for very rare disorders or those with long lag between exposure and outcome
Disadvantages:
reliance on recall or records to determine exposure status
confounders
selection of control groups is difficult
potential bias: recall, selection

Case-control study

Effectiveness of rotavirus
vaccination against childhood
diarrhoea in El Salvador:
case-control study.

de Palma O et al.

BMJ 340:doi:10.1136/bmj.c2825

Pros and cons of cross sectional study

Examines the relationship between 1) diseases/other health
related characteristics and 2) other variables of interest as they exist in a
defined population at one time. Exposure and outcomes both measured at
the same time. Quantifies prevalence, risk, or diagnostic test accuracy
Advantages:
cheap and simple
ethically safe
Disadvantages:
establishes association at most, not causality
recall bias, social desirability bias
researcher’s (Neyman) bias
group sizes may be unequal
confounders may be unequally distributed


Cross sectional study

Sociodemographic patterning of

non-communicable disease risk
factors in rural India: a cross
sectional study.

Kinra S et al.

BMJ 341:doi:10.1136/bmj.c4974

Reporting statements

CONSORT for randomised controlled trials
STARD for diagnostic accuracy studies
STROBE for observational studies
PRISMA for systematic reviews of trials
MOOSE for meta-analyses of observational studies

EQUATOR network

equator-network.org/resource-centre/library-of-health-research reporting/


CONSORT 2010CONsolidated Standards of Reporting Trials

CONSORT 2010 cont.

Thanks

tgroves@bmj.com