Regarding the potential usefulness of appropriate labels:
The beginning of wisdom is to call things by their right names. Madame
de Stael
Jim
James M. Walker, MD, FACP
Chief Health Information Officer
Geisinger Health System
...it's all very well planning what you will do in six months, what you
will do in a year, but it's no good at all if you don't have a plan for
tomorrow.
Hilary Mantel. Wolf Hall.
>>> Rod Jackson <[log in to unmask]> 6/8/2010 12:03 AM >>>
Hi David - no label is perfect but often they can be helpful. While
the
concept of experimental vs observational studies are ¼slippery* as you
mention, they are part of the ¼complete and clear textual description*
that
you indicate is important.
However as I suggest at the end of Steve*s description of the studies,
the
key issue for Steve to consider is whether there is an appropriate
comparison group.
Regards Rod
On 7/06/10 2:32 AM, "Braunholtz, David A." <[log in to unmask]>
wrote:
> The concept of experimental vs observational study can be quite
slippery, &/or
> not particularly helpful, it seems to me.
>
> Eg following Rod's definitions, a study of an intervention for the
whole
> population (comparing before / after) could be experimental, or
observational,
> depending on whether the 'investigator' is (or isn't) also the one
who is
> taking the administrative decision to introduce the intervention. ie
the
> exact same data could be labelled experimental, or observational -
clearly
> making the label irrelevant, in this case, to any assessment of
'quality' of
> the data for drawing inferences ?!
>
> More important than labels, I think, is a complete & clear textual
description
> of what is being compared with what (systematic reviewers really
appreciate
> this !) , and a thorough appreciation of the factors, apart from the
effect of
> the intervention / exposure, which may be causing any differences
found.
>
> BW
>
> David
> [log in to unmask]
>
> Immpact
> University of Aberdeen
> ________________________________________
> From: Evidence based health (EBH)
[[log in to unmask]] On
> Behalf Of Rod Jackson [[log in to unmask]]
> Sent: 06 June 2010 10:45
> To: [log in to unmask]
> Subject: Re: Observational studies
>
> Hi Steve - I have commented on the classification of each of the
studies at
> the end of each of your descriptions.
>
> However as an introduction, epidemiological studies can be classified
in
> several ways.
>
> First, studies can be classified as either experimental or
observational. In
> an experiment, the study investigator controls the allocation of the
exposure
> (or exposures) of interest to the participants. In an experiment the
exposure
> is always an ¼intervention* which is usually compared with another
> intervention or placebo. As you know, in the ideal experiment,
allocation of
> interventions to different arms of the study is done using a
randomisation
> process so each participant has an equal chance of being in any of
the study
> arms (i.e. a rct). The other type of study is an observational study
in which
> the investigator 'observes' (rather than controls) what each
participant is
> exposed to. For example the investigator observes whether the
participants are
> taking a particular drug or not or, say, whether they are smokers or
non
> smokers. The participants are then allocated to different arms of the
study
> based on their exposure status.
>
> Second, studies are then classified by whether the participants are
followed
> up over time to determine the study outcomes (a follow-up or
longitudinal
> study) or whether whether both exposure status and outcome status is
measured
> at the same time (a cross-sectional study). All experiments are
longitudinal
> because if the experimenters allocate participants to exposure, they
need to
> follow them for a period of time to determine the effect of the
exposures on
> outcomes (unless the effect of the exposure on outcomes is
instantaneous which
> is theoretically possible but unlikely). In contrast observational
studies can
> be longitudinal (i.e. Cohort studies) or cross-sectional (e.g.
Surveys) or a
> combination of both. Observational studies can be both longitudinal
and
> cross-section because the investigator only observes if participants
have been
> exposed to the exposure of interest. Therefore participants have
almost always
> been exposed (or not exposed) to the exposure of interest for some
time before
> the study starts.
>
> Third, most studies have at least 2 arms - exposure and comparison
groups -
> but both experiments and observational studies could just have one
arm (i.e.
> no comparison arm). Small groups of patients in ¼one arm* studies
are
> typically called case series. They are experiments if the
investigator
> controls the ¼exposure* and observational studies if the exposure
allocation
> is not part of a study and the investigator observes what
participants have
> been exposed to. However even most small case series can be divided
into
> smaller subgroups (as in some of the studies you describe) so the
definition
> of a case series is rather lose but generally describes one smallish
group of
> patients with a similar condition and similar treatment who are
followed for a
> period of time.
>
> While a cohort study classification is generally only given to
longitudinal
> observational studies, RCTs could be considered as cohort studies in
which
> exposure has been randomly allocated by the investigator.
>
> A case-control study is a sub-type of a observational cohort study.
You may
> have heard about nested case-control studies. The standard case
control study
> is simply nested in a virtual cohort study. None of your studies are
case
> control studies.
>
> I hope these comments are of some help
>
> Rod Jackson
> Professor of Epidemiology
> Section of Epidemiology and Biostatistics
> School of Population Health, Tamaki Campus
> Faculty of Medical & Health Sciences, University of Auckland
> Private Bag 92019
> Auckland, New Zealand
>
>
> On 6/06/10 8:38 AM, "Steve Mathieu" <[log in to unmask]>
wrote:
>
>> > I would be really grateful for some help with categorising some
different
>> > types of observational studies. I have included some example
abstracts
>> below:
>> >
>> > 1) 12 patients with the need for inotropic support were studied.
One dose
>> of
>> > levosimendan (12.5 mg) was administered at a rate of 0.1-0.2
microg
>> > kg(-1).min(-1), either alone or in addition to pre-existing
inotropic
>> therapy.
>> > Haemodynamic measurements were obtained at baseline, and at 3 h, 6
h, 12 h,
>> > and 24 h after the start of the levosimendan infusion.
Levosimendan
>> > significantly increased cardiac output from (mean+/-SD)
4.3+/-0.91.min(-1)
>> to
>> > 5.2+/-1.51 min(-1) after 24h (P=0.013), by increases in stroke
volume
>> > (baseline 47+/-15 ml, after 24h 57+/-25 ml, P=0.05), as heart rate
remained
>> > unchanged. Systemic vascular resistance decreased from 1239+/-430
>> > dyn.sec.cm(-5) at baseline to 963+/-322 dyn.sec. cm(-5) at 24h
(P<0.001).
>> > Pre-existing inotropic therapy present in ten patients remained
unchanged
>> or
>> > was reduced. In postoperative critically ill patients, infusion
of
>> > levosimendan exerted favourable haemodynamic responses.
Levosimendan
>> increased
>> > cardiac output by increasing stroke volume, which might be
attributed
>> > primarily to its inotropic properties. Due to its cyclic
adenosine
>> > monophosphate independent positive inotropic effects, levosimendan
may be
>> of
>> > value as adjunctive therapy to other inotropic drugs in critically
ill
>> > patients'
>> >
>> > Is this therefore a cohort study (one arm levosimendan + other
inotropes;
>> > other arm = inotropes alone)? Prospective with 2 different
treatment arms
>> > albeit the patient groups are not matched other than all post op
and
>> > necessitating treatment for heat failure
>
> This could be described as a case series. It is an experiment as
discussed
> above. All patients were given levosimendan by the investigators and
some of
> the patients were already on pre-existing therapy. The study is
prospective as
> information was collected at the time of levosimendan administration
and the
> patients were then followed over a period of time. This study could
also be
> considered as a very small cohort study in the sense that all
follow-up
> studies are cohort studies. Almost any cohort can be divided into
different
> 'arms' as is possible with this study so you could analyse it in
these two
> 'arms.' However you would want to know why some patients received
> levosimendan while others didn*t. The reason may be a major cause of
bias.
>> >
>> >
>> > 2) This retrospective analysis covers our single centre
experience. Between
>> > July 2000 and December 2004, 41 consecutive patients were treated
for this
>> > complication. Of these, 38 patients are included in this
retrospective
>> > analysis as 3 patients died in the operating room. Levosimendan
was added
>> to
>> > the treatment protocol for the last nine patients. RESULTS: Of 29
patients
>> > treated without levosimendan, 20 could be weaned off the device, 9
survived
>> to
>> > intensive care unit discharge, 7 left hospital alive and 3
survived 180
>> days.
>> > All 9 patients treated with levosimendan could be weaned, 8 were
discharged
>> > alive from ICU and hospital, and 7 lived 180 days after surgery (p
< 0.002
>> for
>> > 180 day survival). Plasma lactate after explantation of the device
was
>> > significantly lower (p = 0.002), as were epinephrine doses.
>> >
>> > Is this a case control study? Retrospective with 2 different
groups
>> > (levosimendan + treatment protocol vs treatment protocol). Very
poorly
>> matched
>> > in terms of numbers 29 vs 9
>
> This study is not an experiment because the investigators did not
control the
> allocation of the exposure as part of a study. Instead they observed
it after
> the event. The study is described as retrospective as it was
initiated after
> the study period had in effect finished. However as information on
exposures
> (i.e. interventions) was documented in records at the time they were
given and
> as this was done before the outcomes (which were also documented in
records),
> this is actually an analysis based on prospectively collected data.
This is
> different from a retrospective study in which the data wasn*t well
documented
> at the time of exposure or outcome and the data has to be collected
> retrospectively rather than retrospectively extracted from
prospective
> records. This study could be considered a case series but it could
also be
> considered as a cohort study with two arms. The key potential cause
of bias
> related to the two groups being followed over different time periods
and both
> the severity of disease and background treatment may differ in the
two
> periods.
>> >
>> > 3) The Brazilian Evaluation of Levosimendan Infusion Efficacy
(BELIEF)
>> study
>> > was prospective, multicenter, observational and included 182
high-risk DHF
>> > patients, all of which received open-label levosimendan. Primary
end point
>> was
>> > hospital discharge without additional inotropic therapy
(responder).
>> Secondary
>> > end points were changes in hemodynamics, clinical parameters, and
brain
>> > natriuretic peptide (BNP). RESULTS: Mortality rate was 14.8%, and
139 of >>
182
>> > patients were responders. In non responders it was 62.8%. Systolic
blood
>> > pressure was a predictor of response. In b-adrenergic agonist
resistant
>> group,
>> > 55.8% were responders. Overall, 54 patients experienced at least
one
>> adverse
>> > event; most of them resolved either spontaneously or after
levosimendan
>> dose
>> > reduction. A significant improvement in quality of life was
verified at 2-6
>> > months of follow-up (p<0.0001). CONCLUSION: Our results suggest
>> levosimendan
>> > infusion as an alternative therapy in the short term management of
DHF
>> > patients. HF severity can influence the response to levosimendan
treatment.
>> > Prospective studies are warranted in a Brazilian cohort including
Chagas
>> heart
>> > disease.
>> >
>> > This has been recorded in the journal as 'prospective,
multicentre,
>> > observational' and later 'open-label and non randomised'. There is
no
>> > comparator group identified at the start i.e the control group who
did not
>> > receive treatment. However, they have found during the analysis
that some
>> > patients 'respond' and others are 'non responders' to treatment
with
>> > levosimendan. Does this now become a cohort study because of
subsequent
>> > statistical analyses of these groups, or is it a (very extensive)
case
>> series?
>
> This study has some features of the two previous studies. If the
investigators
> were responsible for giving the participants the infusion, then it is
a non
> randomised experiment with two groups, one receiving only
levosimendan and
> the other receiving levosimendan plus additional inotropic therapy.
It is
> prospective so could be considered a cohort study if the
investigators did not
> control the intervention. It could also be considered as an extensive
case
> series as you suggest.
>> >
>> > 4) Twenty-six (21 male) consecutive patients were studied,
corresponding to
>> 32
>> > levosimendan administrations (bolus + 24h infusion), aged
56.7+/-13.0
>> years,
>> > with decompensated chronic heart failure, in NYHA functional class
III-IV
>> > (78.1% in class IV), and cardiac index (CI) <2.5 l/min/m2.
Clinical (NYHA
>> > class), non-invasive hemodynamic (echocardiography) and
neurohormonal
>> (Elecsys
>> > ECLIA NT-ProBNP) evaluations were performed before levosimendan
>> administration
>> > and on days 1, 4, 10 and 30. RESULTS: 1) Until day 10, there was
a
>> progressive
>> > decrease in NT-ProBNP values and weight (p<0.001), with an
increase in CI
>> > (p<0.001); 2) NYHA functional class improved progressively, with
76% of the
>> > patients in NYHA class II at day 30; 3) NT-ProBNP values at day 1
>> correlated
>> > inversely (r=-0.414; p=0.024) with CI at day 4; and 4) the
absolute
>> decrease
>> > in NT-ProBNP values at day 4 (relative to baseline values)
correlated with
>> > weight loss at day 4 (r=0.495, p=0.005), day 10 (r=0.424, p=0.031)
and day
>> 30
>> > (r=0.486, p=0.030). CONCLUSION: Levosimendan therapy in patients
with
>> > decompensated chronic heart failure contributes to progressive
NYHA class
>> > improvement. The variations seen in NYHA class and hemodynamics
was
>> reflected
>> > in changes in NT-ProBNP.
>> >
>> > This presumably is a case series? Does a consecutive case series
such as
>> this
>> > (if agree it is indeed a case series) carry any more validiity
than one
>> where
>> > single cases are grouped together over a wider period?
>
> This is the most typical case series as there appears to be no
comparison
> group.
>
> The usefulness of these studies depends on your question. If it it
something
> along the lines of: ¼is levosimendan alone, or in addition to other
therapies
> more effective than other therapies alone in reducing morbidity and
mortality*
> then these studies all suffer from the same fundamental flaw - there
is no
> appropriate comparison group. You own retrospective analyses (study
2)
> probably comes closest to having a meaningful comparison group, if
the only
> reason the previous patients did not receive levosimendan was because
it was
> unavailable. But not only are the numbers very small but the two
groups were
> from different time periods.
>> >
>> > Thank you very much in advance for your help with this. I'm afraid
I am
>> very
>> > new to data analyses and whilst I have done some background
reading
>> continue
>> > to be baffled with some of the wording/categories when trying to
critically
>> > appraise papers.
>> >
>> > Best wishes
>> >
>> > Steve
>
>
> The University of Aberdeen is a charity registered in Scotland, No
SC013683.
>
IMPORTANT WARNING: The information in this message (and the documents attached to it, if any) is confidential and may be legally privileged. It is intended solely for the addressee. Access to this message by anyone else is unauthorized. If you are not the intended recipient, any disclosure, copying, distribution or any action taken, or omitted to be taken, in reliance on it is prohibited and may be unlawful. If you have received this message in error, please delete all electronic copies of this message (and the documents attached to it, if any), destroy any hard copies you may have created and notify me immediately by replying to this email. Thank you.
|