Author + information
- Steven E. Nissen, MD∗ ()
- ↵∗Reprint requests and correspondence:
Dr. Steven E. Nissen, Cleveland Clinic Foundation, Department of Cardiovascular Medicine, 9500 Euclid Ave, Desk F-15, Cleveland, Ohio 44195-0001.
Since its introduction in 2002, intravascular ultrasound (IVUS) “virtual histology” (VH) has generated an extraordinary number of scientific publications (1). Searching the term “intravascular ultrasound virtual histology” in PubMed yields nearly 500 citations over the last decade. A scan of these studies reveals a common theme, the assumption that VH findings, specifically the identification of a “thin cap fibroatheroma,” represent “vulnerable plaque,” with the clear inference that this VH-derived feature can predict clinical outcome for patients with coronary disease. However, more careful scrutiny of this research reveals a disturbing reality—an almost complete absence of well-controlled randomized clinical trials (RCTs) demonstrating meaningful scientific findings for VH. Essentially, an entire body of literature is based principally on anecdotal observational findings. Furthermore, most VH studies were never registered on ClinicalTrials.gov or any another trial registry. Accordingly, the findings typically represent exploratory post hoc analyses, rather than pre-specified, well-planned scientific studies.
The current trial (STABLE [Statin and Atheroma Vulnerability Evaluation] trial) by Park et al. (2), in this issue of the Journal, is a rare exception. The study was pre-planned and generally conducted using methods pre-specified through appropriate clinical trial registration (NCT00997880). The authors should be congratulated for performing one of the few prospective VH studies currently available. The findings are interesting and suggest, but do not definitively prove, that statin administration induces compositional changes in coronary plaques. The clear inference of this trial is the provocative suggestion that statins improve clinical outcomes, at least in part, by altering plaque morphology, although the study was far too small to correlate compositional changes with morbidity or mortality. Unfortunately, however, the current RCT suffers from several of the flaws that have plagued VH research for more than a decade. These limitations weaken the trial’s interpretability, resulting in the need to consider the findings as hypothesis generating rather than definitive.
High-quality clinical trials are conducted using an intent-to-treat approach because failure to collect outcome data for patients who stop study treatment can significantly bias results. In the STABLE trial, 87 of 312 randomized patients (28%) did not have final VH evaluation. Although it is typical in IVUS studies to have a sizable proportion of noncompleters, this flaw is usually a result of patients declining to undergo a repeat invasive procedure at the end of the study. Although the rate of nonretention in the current trial (28%) was not unusual, the reasons for failure to complete are problematic. A total 45 patients were withdrawn from follow-up for atypical reasons, including statin adverse effects (n = 17), noncompliance (n = 6), and withdrawal of consent before administration of statins (n = 22). It is difficult to understand why so many patients would consent to a trial involving statin administration, then withdraw consent before receiving any study drug. Noncompliance and occurrence of adverse effects are reasons for nonadherence to a study drug, but should not be used to justify incomplete follow-up for the primary endpoint.
The highly selected nature of the studied population also limits interpretation. The authors enrolled only patients with an IVUS-identified fibroatheroma within an index lesion. Because the primary endpoint was segmental change in plaque composition, the selection of this population is difficult to justify. If the main objective of the study was studying how statins modify plaque morphology, selecting a broader population would have been sensible and have allowed greater generalizability of results. The decision to randomize patients in such a small study to 10 mg versus 40 mg of rosuvastatin was also unwise and represents a significant confounder in interpreting results. Only 73 patients in the 10-mg treatment group completed the trial, too few to yield reliable insights into the relationship between statin dose and effect on VH. It is challenging to interpret a study that mixes 2 different drug doses, but was not powered to compare results between those doses.
A broader concern is the accuracy and external validity of VH IVUS. It remains uncertain exactly what VH actually measures, and there remain questions about scientific validity and clinical importance of this imaging modality. The original validation study for VH examined necropsy specimens obtained from the left anterior descending coronary arteries from just 51 patients (1). At least 1 subsequent effort to validate VH findings showed no correlation between IVUS findings obtained in vivo in an animal model and subsequent necropsy examination (3). The authors of the STABLE trial place great emphasis on an observational VH study (PROSPECT [Providing Regional Observations to Study Predictors of Events in the Coronary Tree]) that made broad claims about predicting clinical outcome based on a cluster of IVUS findings derived from post hoc analyses (4). However, that study characterized outcomes for “high-risk plaques” primarily based on revascularization outcomes because only 21 patients experienced a myocardial infarction. Additionally, the PROSPECT study combined parameters derived from VH with conventional grayscale IVUS assessment, rather than independently validating the value of VH findings. Given the low quality of existing scientific literature, VH must be viewed as an unvalidated methodology of uncertain clinical significance.
The authors of the STABLE study include some speculative findings in the current paper, several of which are not warranted on the basis of reliable statistical principles and good clinical trial practice. The absence of a difference between the 10-mg and 40-mg doses of rosuvastatin likely represents type II statistical error (lack of power) because only 73 patients were randomized to the low-dose group. Similarly, in a study this small, subgroup analyses or comparisons between stable versus unstable angina represent questionable findings, and the results are largely uninterpretable. Applying multivariate analysis to a small study is also problematic, and the results correlating biomarkers with outcome seem implausible. Although prior studies have shown that changes in high-sensitivity C-reactive protein correlate with change in plaque volume, the lack of association in the STABLE trial between low-density lipoprotein cholesterol and plaque changes is contrary to considerable scientific literature. Larger single IVUS studies and meta-analyses have consistently demonstrated that reduction in low-density lipoprotein cholesterol is strongly associated with plaque changes measured by IVUS (5).
The authors suggest that statins change plaque composition “irrespective” of effects on atheroma reduction. Again, type II statistical error is most likely responsible for this finding. It is also difficult to reconcile how quantitative coronary angiography could show worsening of arterial narrowing while IVUS shows regression. Prior studies have shown that IVUS is more sensitive at detecting regression, but directional changes are typically similar for quantitative coronary angiography and IVUS. Despite all of these limitations, the STABLE study is valuable and represents one of the few prospective RCTs to examine IVUS VH in a systematic fashion. This study is a worthy start to what we hope is a larger effort to systematically validate VH and provide the type of robust scientific analysis that has been available for a long time for volumetric grayscale IVUS analyses. If the current findings are confirmed by additional prospective trials adhering to high standards of scientific rigor, VH just might evolve from a clever gimmick to a useful scientific and clinical tool.
↵∗ Editorials published in the Journal of the American College of Cardiology reflect the views of the authors and do not necessarily represent the views of JACC or the American College of Cardiology.
Dr. Nissen has reported that he has no relationships relevant to the contents of this paper to disclose.
- 2016 American College of Cardiology Foundation
- Nair A.,
- Kuban B.D.,
- Tuzcu E.M.,
- et al.
- Park S.-J.,
- Kang S.-J.,
- Ahn J.-M.,
- et al.
- Thim T.,
- Hagensen M.K.,
- Wallace-Bradley D.,
- et al.