IVUS-Guided Stenting Where is the Harm?

The study by Faulknier and colleagues evaluates the clinical impact of intravascular ultrasound (IVUS) on stent deployment, using bare-metal stents.¹ These investigators retrospectively evaluated the clinical outcomes of 50 stent cases where IVUS was used, and compared the subsequent clinical events with 50 randomly selected cases wherein IVUS was not employed. Clinical outcomes tended to favor the group not receiving intravascular ultrasound guidance. This led to the conclusion by the authors that “the added expense of intravascular ultrasound does not appear to be warranted.” This is a conclusion many will accept at face-value, as it is consistent with perhaps the most-commonly prevailing attitude towards the use of IVUS. However, it may be appropriate to ask if this type of research is helpful or harmful to those who read this journal. See Faulknier, et al. on pages 311–315 The authors defend the presentation of suboptimally collected data and research design because they reasoned that any more appropriately-designed research could not be performed in their specific clinical environment. What are the pros-and-cons of the publication of research under such limitations? On the positive side, the gathering together and scrutiny of data offers lessons in scientific method and critical analysis. Presenting data in a manuscript is also a useful exercise, and everyone enjoys adding a publication to their curriculum vitae and seeing their name in print. However, on the other side, misleading information can be harmful. This may be particularly true when the options include a simpler, less cumbersome approach (such as angiographic guidance of stents) versus an approach which entails up-front costs and effort. Intravascular ultrasound guidance of stent implantation, with data from several studies suggesting a reduction in restenosis (including several not referenced by the authors), is an example of a strategy with up-front inconvenience and cost, yet later possible benefits.^2–7 Some of the studies have even suggested that the use of IVUS may in fact be cost saving or cost neutral.^3,7–10 A cursory reading of this manuscript, strongly emphasized by the last line, appears to justify the simpler approach, and is likely the “take-home” message of many readers. The authors acknowledge several limitations and admit that selection bias may be playing an important role. They were unable to stratify patients as to why intravascular ultrasound was selected for particular cases. Therefore several possibilities exist to explain the findings, such as a) higher risk lesions (such as restenotic lesions) preferentially selected for intravascular ultrasound imaging, b) the operator who performed the majority of the IVUS cases was referred higher risk cases, c) the operator who performed a majority of IVUS cases did other things which perhaps contributed to the less optimal outcomes, or d) IVUS was performed but not acted upon. If higher risk cases were preferentially undergoing IVUS, it is possible that IVUS was in fact beneficial and even cost-saving — but this would be lost in the data as presented. Incomplete follow-up is also a concerning issue, which compounded another significant problem, inadequate trial size. It is notable that the authors did not present any statistical analysis of the power of the study, a common error. It is unclear why they selected 50 patients per group to analyze. Without an assessment of assumed event rates it does not appear any attempt was made to identify an appropriate number of patients to evaluate. This underpowering of data has been decried as being unethical in the context of a prospective randomized trial, by subjecting patients to studies without a reasonable expectation that the questions posed will be answered.¹¹ Although a retrospective review does not subject individual patients to a research protocol, the potential to draw misleading conclusions also has ethical implications. Clearly retrospective reviews are beneficial, even when the study sample is small. They highlight empirical observations and provide means to generate hypotheses which may be tested in larger prospective studies. Furthermore, retrospective and prospective case series complement the findings of randomized trials. They can sometimes provide additional insight into the findings of larger registry data-sets representing “real-world” clinical outcome data, by providing greater detail. Certainly, the value of IVUS guidance of stent implantation remains controversial, with some randomized trial data arguing against the benefits of IVUS.^6,12 Therefore, careful observational studies in single or multi-center registries should be encouraged. Findings in small, retrospective series must always be interpreted with caution, however, in particular when there are so many important limitations. Furthermore, comparison studies must be done even more diligently, due to the potential for creating misleading conclusions. What might these investigators have done differently? First, there might have been some attempt to risk stratify the lesions being treated for restenosis risk. These might have included angiographic measurements such as reference lumen diameter and lesion length, as well as clinical information such as diabetic status or restenotic lesion. They then might have matched the lesions using the pre-defined criteria. Second, they might have made some a priori assumptions of the likelihood of a clinical event, then assumed a potential benefit or risk of the alternative treatment (in a risk-stratified model), in order to evaluate how many patients would need to be included to draw conclusions. They could have used this information to decide how many of their patients they would need to evaluate to make sense of their data, rather than arbitrarily choosing 50 patients in each arm. The underpowering present in this study virtually ensured that the role of IVUS could not be found to be beneficial. An improperly powered study is misleading. Third, greater diligence in follow-up would have lended greater confidence in their findings. Finally, even though the data are mixed, since several more carefully constructed studies arrived at different conclusions, they might have been more cautious in their interpretations of their data. Small studies such as this are commonly performed and submitted for publication. Perhaps not enough focus has been placed on the potential for harm of submitting or publishing studies which by their structure and limitations may be misleading. This commentary is not intended to chastise these investigators, who should be commended for their diligence in seeing a study through to completion and publication. Instead, this commentary is intended to focus attention on an issue which is rife throughout the medical literature: at which point are the limitations of a study so considerable that publication of the data may in fact be harmful.