| HOME | HELP | FEEDBACK | SUBSCRIPTIONS | ARCHIVE | SEARCH | TABLE OF CONTENTS |
Correspondence |
The recent systematic review of imaging methods for detecting hepatocellular carcinoma (HCC) in patients with cirrhosis [1] illustrates some of the reasons why the methods of evidence-based medicine (EBM) are often unhelpful and may be misleading in the evaluation of diagnostic procedures.
1. The subject material is out of date: the papers on which this review was based were published between 1995 and 2001. With the inevitable delay between acquisition of data and publication, this means that the results discussed in the review date back at least 4–5 years, possibly over 10 years in some cases. During this time there have been considerable advances in experience with multislice CT and with improved contrast-enhanced MRI techniques. Iodized oil CT is extinct and CTPA not far off being so, whilst new contrast agents have completely transformed ultrasound.
2. The approach is high selective and arbitrary: the authors detailed only 29 out of 997 papers containing relevant data. Whilst it must be accepted that much of the imaging literature is unsubstantiated and studies show major variations in diagnostic criteria and the selection of patients, an appreciative reading of the literature would seek to answer the question "what new, useful and reliable information can be gained from this study?" Using rigid and arbitrary criteria to dismiss over 97% of published material on the subject looks more like systematic neglect than systematic review, and risks throwing the baby out with the bath water.
3. The standard of reference is too variable: even amongst the 29 studies included in the review, the evidence for confirmation of the diagnosis varied from histological examination of the explanted liver – which is desirable – to "confirmatory imaging" or "rising AFP" – which is clearly inadequate. It seems perverse to exclude the vast majority of studies from consideration, but then to include some publications with inadequate verification of the true diagnosis.
4. There is no account of observer variation: it appears that in the selection of the studies for inclusion in this analysis, no account was taken of whether single or multiple observers reviewed each individual case. There is a huge amount of literature on observer variation in diagnostic imaging, all of which supports the view that studies which use a single observer for image interpretation are less reliable than those which use multiple observers. Whilst there is an argument for single observers in prospective studies – when the intention is to reflect clinical practice – retrospective studies seeking to compare different imaging methods or to measure diagnostic accuracy should certainly have multiple observers.
5. Dumbing down the detail: the Procrustean approach of bundling together studies using different patient selection, different diagnostic criteria, different modes of verification, and many variations in detail of imaging techniques, deliberately obscures the type of detail which may be critical in the proper evaluation of each study. For example, is it sensible to concatenate studies using single phase and multiphase CT, and those with different slice thicknesses? Although contrast agents in MRI do receive a brief mention, there is no discussion of the details of technique which in some cases are critical to success, and no consideration of the different virtues of tissue-specific and extracellular contrast agents.
6. The conclusion is erroneous: although the authors mentioned some of the points raised above as limitations of their review, they still came to the erroneous conclusion that "there is insufficient evidence to decide on the best imaging technique for the detection of HCC in cirrhotic patients". There is evidence – it is just that the authors' analysis is too limited and too crude to detect it. Many of the 968 studies ignored by the authors contain at least some valuable messages – for example, data on the detectability of HCCs in cirrhotic patients using dual contrast MRI with pathological correlation is given in Ward et al [2] and Bhartia et al [3], whilst a fairly recent review of the whole subject may be found in Ward and Robinson [4]. What is needed is an attempt to understand the difficulties of imaging in this context, rather than a search for a single idiot-proof method which will fit all patients.
7. It is post-modern pseudoscience: The most worrying feature of the pseudoscientific approach illustrated by this review is that it affords a cloak of apparent respectability to erroneous conclusions based on an inadequate analysis. Whilst there is a strong argument for the use of EBM methods in assessing the effects of therapeutic interventions, the case for using this approach in the assessment of diagnostic methods is very weak, largely because the outcomes are much less clearly definable. We have no diagnostic equivalent to the randomized clinical trial. In the diagnostic specialties we should really concentrate our efforts and our research on the study of patients and disease, rather than relying on an over-simplification of the literature which is closer to tabloid journalism than to erudition.
Intelligence may be defined as the ability to learn from experience, so it is unfortunate that the authors of this review – who are undoubtedly capable and dedicated professionals – give the impression that having scrutinised 997 publications relating to the diagnosis of HCC in cirrhosis, they are no nearer knowing how to go about this task than when they started.
Yours etc.,
Professor of Clinical Radiology, St James's University Hospital, Beckett Street, Leeds LS9 7TF
Received for publication November 10, 2004. Accepted for publication December 21, 2004.
References
Related articles in BJR:
| ||||||||||||||||||||||||||||||||||||||||||||||||||||
| HOME | HELP | FEEDBACK | SUBSCRIPTIONS | ARCHIVE | SEARCH | TABLE OF CONTENTS |
| BJR | DMFR | IMAGING | ALL BIR JOURNALS |
