Limits...
Observational research, randomised trials, and two views of medical science.

Vandenbroucke JP - PLoS Med. (2008)

View Article: PubMed Central - PubMed

Affiliation: Royal Netherlands Academy of Arts and Sciences, and Department of Clinical Epidemiology, Leiden University Medical Centre, Leiden, The Netherlands. j.p.vandenbroucke@lumc.nl

AUTOMATICALLY GENERATED EXCERPT
Please rate it.

Discoveries happen when things are suddenly seen in another light: the odd course of a disease in a patient, the strange results of a lab experiment, a peculiar subgroup in the analysis of data, or some juxtaposition of papers in the literature... The other view is that of medical researchers whose aim is to set up studies to evaluate whether the patient's lot is really improved by the new therapies or diagnostics that looked so wonderful initially... After the mutation was established, we looked at the data again... We found a few homozygotes for the mutation among the patients... In daily medical practice, prescribing will be guided by the prognosis of the patient: the worse the prognosis, the more therapy is given... This leads to intractable “confounding by indication”... Hence, to measure the effect of treatment, we need “concealed randomisation” to break the link between prognosis and prescription... In contrast, adverse effects are “unintended effects” of treatment, and are mostly unexpected and unpredictable, which means that they usually are not associated with the indications for treatment... If the answer to these questions is positive, that will lead to greater credibility of the results... If negative, as in the vegetarian example, we may attach no credibility to the results despite any attempts at statistical correction for confounders... Analyses are guided by clues that involve reasoning, much like in the example of factor V Leiden and oral contraceptives above... That example also shows that we did not “try to explain a subgroup” after we found it... This was already advocated for subgroups found in randomised trials, where the veracity of a surprising finding can be strongly enhanced if similar subgroup results are found across similar trials in a meta-analysis... The ideas about subgroups and prior odds of hypotheses lead to further insight in the usual hierarchy of strength of study designs with the randomised trial on top and the case report at a suspect bottom (Box 1)... The way in which prior odds might shape our views can be understood when imagining an upside-down world in which randomised trials would be started with the same prior odds of truth as individual SNPs in a genome-wide analysis, say, one in 100,000.

Show MeSH
Axis of Multiplicity
© Copyright Policy
Related In: Results  -  Collection


getmorefigures.php?uid=PMC2265762&req=5

pmed-0050067-g002: Axis of Multiplicity

Mentions: This problem can be conceptualised on an “axis of multiplicity” (Figure 2). At one extreme there are genome-wide analyses, where tens of thousands of single nucleotide polymorphisms (SNPs) are investigated for disease associations. The prior probability that some grain of explanation will come from any individual SNP is slim, say, one in 100,000 [15]. At the other extreme, there are randomised trials about a single disease, a single therapy, and a single outcome. Randomised controlled trials are started under equipoise [16,17]: the prior odds that the therapy that is tested is worthwhile are 50–50, and multiplicity of analysis is strictly not allowed. Thus, the axis of multiplicity is at the same time an axis of prior belief: the prior belief that some factor will be a causal explanation for a condition or that some therapy or treatment will work [18].


Observational research, randomised trials, and two views of medical science.

Vandenbroucke JP - PLoS Med. (2008)

Axis of Multiplicity
© Copyright Policy
Related In: Results  -  Collection

Show All Figures
getmorefigures.php?uid=PMC2265762&req=5

pmed-0050067-g002: Axis of Multiplicity
Mentions: This problem can be conceptualised on an “axis of multiplicity” (Figure 2). At one extreme there are genome-wide analyses, where tens of thousands of single nucleotide polymorphisms (SNPs) are investigated for disease associations. The prior probability that some grain of explanation will come from any individual SNP is slim, say, one in 100,000 [15]. At the other extreme, there are randomised trials about a single disease, a single therapy, and a single outcome. Randomised controlled trials are started under equipoise [16,17]: the prior odds that the therapy that is tested is worthwhile are 50–50, and multiplicity of analysis is strictly not allowed. Thus, the axis of multiplicity is at the same time an axis of prior belief: the prior belief that some factor will be a causal explanation for a condition or that some therapy or treatment will work [18].

View Article: PubMed Central - PubMed

Affiliation: Royal Netherlands Academy of Arts and Sciences, and Department of Clinical Epidemiology, Leiden University Medical Centre, Leiden, The Netherlands. j.p.vandenbroucke@lumc.nl

AUTOMATICALLY GENERATED EXCERPT
Please rate it.

Discoveries happen when things are suddenly seen in another light: the odd course of a disease in a patient, the strange results of a lab experiment, a peculiar subgroup in the analysis of data, or some juxtaposition of papers in the literature... The other view is that of medical researchers whose aim is to set up studies to evaluate whether the patient's lot is really improved by the new therapies or diagnostics that looked so wonderful initially... After the mutation was established, we looked at the data again... We found a few homozygotes for the mutation among the patients... In daily medical practice, prescribing will be guided by the prognosis of the patient: the worse the prognosis, the more therapy is given... This leads to intractable “confounding by indication”... Hence, to measure the effect of treatment, we need “concealed randomisation” to break the link between prognosis and prescription... In contrast, adverse effects are “unintended effects” of treatment, and are mostly unexpected and unpredictable, which means that they usually are not associated with the indications for treatment... If the answer to these questions is positive, that will lead to greater credibility of the results... If negative, as in the vegetarian example, we may attach no credibility to the results despite any attempts at statistical correction for confounders... Analyses are guided by clues that involve reasoning, much like in the example of factor V Leiden and oral contraceptives above... That example also shows that we did not “try to explain a subgroup” after we found it... This was already advocated for subgroups found in randomised trials, where the veracity of a surprising finding can be strongly enhanced if similar subgroup results are found across similar trials in a meta-analysis... The ideas about subgroups and prior odds of hypotheses lead to further insight in the usual hierarchy of strength of study designs with the randomised trial on top and the case report at a suspect bottom (Box 1)... The way in which prior odds might shape our views can be understood when imagining an upside-down world in which randomised trials would be started with the same prior odds of truth as individual SNPs in a genome-wide analysis, say, one in 100,000.

Show MeSH