Author + information
- aDepartment of Medicine (Cardiology Division), Massachusetts General Hospital, Harvard Medical School, Boston, Massachusetts
- bDepartment of Diagnostic Radiology and Nuclear Medicine, University of Maryland School of Medicine, Baltimore, Maryland
- ↵∗Reprint requests and correspondence:
Dr. Henry Gewirtz, Department of Medicine (Cardiology Division), Massachusetts General Hospital, Yawkey 5E, 55 Fruit Street, Boston, Massachusetts 02114.
In this issue of iJACC, van der Valk et al. (1) report the results of a study designed to establish an optimal threshold for distinguishing normal from abnormal positron emission tomography (PET) measured 18fluorodeoxyglucose (18FDG) arterial wall activity. The authors propose use of the cutpoint in future clinical trials requiring an objective, quantitative measure of arterial inflammation and response to intervention(s) targeted at ameliorating same. Sample size is an important issue in such trials; therefore, the authors tested several different metrics in an effort to find 1 that best separated healthy controls (n = 25) from: 1) patients with cardiovascular disease (CVD) risk factors but without overt disease (n = 23); and 2) those with manifest CVD (n = 35). Five frequently applied arterial 18FDG uptake metrics were assessed. Although all 5 18FDG uptake metrics for both the carotid artery and aorta demonstrated excellent reproducibility and differed significantly among healthy subjects and those either with risk factor for or overt atherosclerotic cardiovascular disease, there was considerable data overlap among subject categories. Ultimately, the carotid artery maximum standardized uptake value (SUV) target/venous 18FDG target to background ratio (TBRmax) metric was selected, because, in comparison with the other metrics, it allowed for the smallest sample size required to achieve statistically significant results given a range of anticipated treatment efficacy (5% to 20%). It also was best at separating putatively healthy controls from both the CVD risk factor only and manifest CVD groups. The authors acknowledge many limitations of their study, several of which have broad clinical and experimental relevance and are worth considering in more detail.
The requirement to standardize patient preparation, image acquisition, processing and analysis protocols, and calibration of PET instrumentation is a given for any clinical trial, which typically will use a core laboratory for image processing and analysis. What is problematic is that, at a given institution, any or all of these factors may differ importantly from that used in a particular study. Accordingly, it may be difficult, if not impossible, to use a cutpoint for a given parameter reported in any given trial as a basis for clinical decision making in an individual patient. As the authors note, there is a need to “harmonize” PET 18FDG arterial inflammation imaging methodology across future clinical trials. Clinical adoption of a standard methodology, which reflects that used in clinical trials, also is essential if the results of such trials are to be applied to clinical practice.
Related to this is the very important issue of which endpoint to measure and how to measure it. In the present study, the authors investigated 2 different methods for expressing arterial 18FDG SUV; either as SUV minus background activity (venous 18FDG) or as the ratio SUV/venous 18FDG, referred to as TBRmax. TBRmax reflected the mean of 3 adjacent slices with the highest arterial value of SUVmax of the most diseased segment of the target vessel. However, only those slices with TBRmax greater than or equal to a test value (e.g., 1.6) were included in the final value of TBRmax to indicate active arterial inflammation, whereas noninflamed segments were excluded. The authors indicated results were not different with either method of background correction and so used TBRmax because it has been more often used in prior clinical trials. Nevertheless, the parameter is potentially unstable because small changes in the denominator (venous 18FDG) may produce large changes in the ratio, a problem that has been reviewed in more detail by others (2). It would appear, therefore, that the simple subtraction method is likely to prove more reliable, reduce statistical noise, and thereby help to decrease sample size in clinical trials. Ultimately, the authors defined the upper limit of normal TBRmax (∼1.8) at the 90th percentile of control subjects (n = 25, 15 male; 61 ± 11 years of age; body mass index 25 ± 3 kg/m2).
The authors also investigated 2 other related metrics that are worthy of consideration. They cite another report that indicates that in approximately 80% of CVD patients, the diseased arteries arise from the aorta (3) and so TBRmax for the aorta was investigated as a “read out” metric as were: 1) the percent of active carotid and ascending aorta slices above a given TBR threshold (several were tested); and 2) the mean TBR of the active slices above that threshold. In brief, they found that separation of the 3 groups based on TBRmax for the aorta was not as good as that based on the carotid and that there was considerable overlap among the 3 groups for both metrics, although more so for aorta. Further, the authors reported for aorta that the percent of active slices (TBRactive slices cutpoint ≥2.4) and TBRactive (mean of maximum TBR of all slices greater than test cutpoint [i.e., 2.4]) did not differ significantly among the 3 groups with 88% of controls having 74 ± 30% active slices. Separation of controls from the CVD risk factor and manifest CVD groups based on carotid active segment analysis was possible, although almost 50% of controls had active segments (32 ± 40%) and CVD risk factor, and manifest CVD groups were very similar to one another for both metrics. Thus, using the present population-based approach, substantial overlap exists among the 18FDG metrics of apparently healthy subjects and those either at risk for or with manifest atherosclerotic cardiovascular disease. Accordingly, not only, as the authors note, does choice of read-out vessel matter in terms of study sample size, but more important, the data raise the question of just what arterial uptake of FDG really signifies and how best to define what is abnormal. In future studies, perhaps 18FDG metrics should be defined by outcome data to enable the assessment of truly pathological reference ranges of 18FDG uptake indicative of carotid artery and aortic inflammation.
It is possible that, because “healthy” subjects in this study were predominately overweight, middle-aged males, subclinical arterial inflammation was present in many of them and, based on prior work, was more likely to involve the aorta and its major (3) branches than the carotids, although the latter certainly were not immune. Further, consideration should be given to what is truly pathological FDG uptake. It is apparent from results of the current study that a purely statistical definition may not be adequate (or clinically meaningful), especially because the frequency of aortic involvement, where atherosclerotic vascular disease presents in the vast majority of patients, is so prevalent in seemingly healthy, albeit overweight, middle-aged (60 ± 11 years of age) individuals, more often males. Moreover, it is well known that healthy young males killed in military combat as well as civilian trauma commonly have evidence of early atherosclerotic lesions in their coronary vessels (4–7) and aorta (8) as well. Indeed, in the abdominal aorta of subjects 30 to 34 years of age, the prevalence of fatty streaks, fibrous plaques, and complicated and calcified lesions, respectively, was found to be 100%, 55% to 65%, 1% to 4%, and 8% to16%, depending on race and gender (8). The authors recognize that outcome data from 2 ongoing large clinical trials (9,10) are likely to be more useful in defining how best to define pathological FDG uptake. Such data may also prove useful clinically for purposes of individual decision making, something the current statistical approach is unsuitable for.
In summary, the authors are to be congratulated for performing a carefully done, provocative study that raises several important issues. These include: 1) the need to standardize all technical aspects of patient preparation, image timing, data acquisition, and analysis for PET 18FDG arterial wall inflammation investigations; 2) determining the appropriate target arteries for analysis, diseased vessels rather than surrogates would be preferable; 3) determining the optimal method for quantization of extent and severity of inflammation; a volumetric estimate likely would be relatively easy to implement given the digital nature of the data and co-registration of FDG signal with computed tomography angiograms; 4) determining the optimal method for correcting lesion FDG SUVs, which based on both numerical and biological considerations, appears to be simple subtraction of blood pool activity (2) rather than the more commonly used but potentially unstable target/blood ratio approach; and 5) determining the most meaningful definition of what constitutes abnormal levels of arterial wall FDG uptake, which may best be tied to results of clinical outcomes studies such as the 2 referenced by the authors (9,10) rather than purely statistical considerations.
↵∗ Editorials published in JACC: Cardiovascular Imaging reflect the views of the authors and do not necessarily represent the views of JACC: Cardiovascular Imaging or the American College of Cardiology.
Both authors have reported that they have no relationships relevant to the contents of this paper to disclose.
- American College of Cardiology Foundation
- van der Valk F.M.,
- Verweij S.L.,
- Zwinderman K.A.H.,
- et al.
- Chen W.,
- Dilsizian V.
- Joseph A.,
- Ackerman D.,
- Talley J.D.,
- Johnstone J.,
- Kupersmith J.
- Tuzcu E.M.,
- Kapadia S.R.,
- Tutar E.,
- et al.