Diabetes does not impact the diagnostic performance of contrast-based fractional flow reserve: insights from the CONTRAST study

Background Adenosine-free coronary pressure wire metrics have been proposed to test the functional significance of coronary artery lesions, but it is unexplored whether their diagnostic performance might be altered in patients with diabetes. Methods We performed a post-hoc analysis of the CONTRAST study, which prospectively enrolled an international cohort of patients undergoing routine fractional flow reserve (FFR) assessment for standard indications. Paired, repeated measurements of all physiology metrics (Pd/Pa, iFR, contrast-based FFR, and FFR) were made. A central core laboratory analyzed blinded pressure tracings in a standardized fashion. Results Of 763 subjects enrolled at 12 international centers, 219 (29%) had diabetes. The two groups were well-balanced for age, clinical presentation (stable or unstable), coronary vessel studied, volume and type of intracoronary contrast, and volume of intracoronary adenosine. A binary threshold of cFFR ≤ 0.83 produced an accuracy superior to both Pd/Pa and iFR when compared with FFR ≤ 0.80 in the absence of significant interaction with diabetes status; indeed, accuracy in subgroups of patients with or without diabetes was similar for cFFR (86.7 vs 85.4% respectively; p = 0.76), iFR (84.2 vs 80.0%, p = 0.29) and Pd/Pa (81.3 vs 78.9%, p = 0.55). There was no significant heterogeneity between patients with or without diabetes in terms of sensitivity and specificity of all metrics. The area under the receiver operating characteristic (ROC) curve was largest for cFFR compared with Pd/Pa and iFR which were equivalent (cFFR 0.961 and 0.928; Pd/Pa 0.916 and 0.870; iFR 0.911 and 0.861 in diabetic and non-diabetic patients respectively). Conclusions cFFR provides superior diagnostic performance compared with Pd/Pa or iFR for predicting FFR irrespective of diabetes (clinicaltrials.gov identifier NCT02184117).


Background
Due to its multi-test validation and cost-effective improvement in clinical outcomes demonstrated by multiple randomized trials as well as observational data, physiological assessment of coronary stenoses by fractional flow reserve (FFR) has emerged as the gold standard for making revascularization decisions on stable lesions [1][2][3][4][5]. Importantly, FFR requires hyperemia, but that does add a small cost and risk [1,6,7]. Therefore, the CONTRAST (Can cONTrast Injection Better Approximate FFR compAred to Pure reSTing Physiology?) study recently investigated whether contrast medium could provide an easy alternative and inexpensive tool for Open Access Cardiovascular Diabetology *Correspondence: espogiov@unina.it 11 Division of Cardiology-Department of Advanced Biomedical Sciences, Federico II University of Naples, Via Pansini 5, 80131 Naples, Italy Full list of author information is available at the end of the article assessing FFR (namely contrast FFR, cFFR). It demonstrated that cFFR was superior to resting measurements (rest distal pressure [Pd]/aortic pressure [Pa], and the instantaneous wave-free ratio [iFR]) in terms of diagnostic performance to predict FFR [8].
Diabetes does not seem to significantly impact FFR accuracy or its interpretation [4,[19][20][21][22], although it can produce coexisting epicardial lesions (quantified by FFR) and microvascular dysfunction (often quantified by measures of hyperemic resistance). Because of potential alterations in microvascular reactivity, cFFR might perform differently in diabetic patients. Thus the aim of this study was to explore whether diabetes might impact the diagnostic ability of cFFR compared with Pd/Pa or iFR versus adenosine-derived FFR ≤ 0.80.

Study population
We explored the impact of diabetes in a post-hoc analysis of the CONTRAST study (clinicaltrials.gov identifier NCT02184117). The design and results of this study have been previously published [8]. Briefly summarized, CONTRAST was an investigator-initiated, prospective diagnostic accuracy study that enrolled a multicenter, international cohort of patients undergoing routine FFR assessment for standard indications. Patients were recruited from 12 centers between June 2014 and April 2015. Subsequent care proceeded as per routine from the clinical FFR measurement without further studyrelated follow-up. Each subject gave informed consent as approved by the local institutional review board of that participating center. Subjects were excluded in case of prior coronary bypass surgery, known severe cardiomyopathy (LV ejection fraction <30%) or LV hypertrophy (septal wall thickness >13 mm), contraindication to adenosine, or renal insufficiency such that an additional 12-20 ml of contrast would, in the opinion of the operator, pose an unwarranted risk. In cases of multivessel disease, only the first lesion studied using FFR was included. Culprit lesions for an acute infarction were excluded, but non-culprit lesions were permitted. Standard demographic, clinical, and catheterization parameters were collected for each subject.

Measurements and core lab analysis
The physiology protocol and core lab analysis used for the study has been previously described [8]. Briefly, an initial period of at least 1 min provided a stable assessment of resting physiology without further contrast injection. Next, a manual or injector-based IC bolus of contrast medium was given as per local practice for diagnostic angiography. The volume and type of IC contrast medium were not mandated but varied among sites and even among subjects at a single site but with a strong recommendation for 6-10 ml. A second IC bolus of contrast medium was injected using the same technique when conditions had returned to baseline (after approximately 1 min). Next, after the return of baseline conditions, a manual IC bolus of adenosine was administered (dose at operators' discretion but with a strong recommendation for 100-200 μg). A second, identical bolus of IC adenosine was given after 1 min.
A subsequent period of at least 1 min provided a second assessment of resting physiology before starting an adenosine infusion at a standard rate of 140 μg/kg/min via either a central or wide bore peripheral IV. The duration of the infusion was approximately 2 min, but could be prolonged or abbreviated as necessary. After stopping IV adenosine and waiting approximately 2 min for conditions to return to baseline, a second, identical IV adenosine infusion was performed. At the end of the procedure, an optional but recommended drift check was performed by bringing the pressure sensor back to the guide catheter at the same location as equalization.
All pressure tracings were sent to the Cardiovascular Research Foundation physiology core lab for standardized and centralized review. The core lab carried out its post-hoc analysis without knowledge of the locally determined Pd/Pa value, IC substance (contrast medium or adenosine), enrolling site, or subject/lesion characteristics.

Statistical analysis and endpoints
The primary endpoint was accuracy against FFR ≤ 0.80 and compared using a McNemar test between metrics. Secondary endpoints included the area under the receiver operating characteristic (ROC) curve (compared using the DeLong method), sensitivity, and specificity. All the analyses were performed in diabetic and non-diabetic patients. Based on previous evidence [8], the binary thresholds were as follows: FFR ≤ 0.80; cFFR ≤ 0.83; Pd/ Pa < 0.92 and iFR < 0.90.
Continuous variables are presented as mean ± standard deviation and were compared with independent samples Student t test. Categorical variables are expressed as count and percentages and were compared with Chi square or Fisher exact tests as appropriate. A Cox proportional model was used for univariate and multivariate analysis to explore the role of diabetes in predicting the cFFR disagreement with FFR. Applicable tests were two-tailed, and p < 0.05 was considered statistically significant. Analyses were performed in SPSS, version 23.0 (SPSS Inc., Chicago, IL, USA) and MedCalc version 12.5 (MedCalc Software, Ostend, Belgium).

Results
Of 763 subjects, 219 (29%) were diabetic. Compared with non-diabetic patients, those with diabetes were more frequently women, had a higher body mass index, and more frequently had hypertension or dyslipidemia ( Table 1). The 2 groups were well-balanced for clinical presentation (stable or unstable), coronary vessel studied, volume and type of intracoronary contrast, and volume of intracoronary adenosine ( Table 1).
As a physiologic explanation for the superior diagnostic performance of cFFR compared to resting physiology (either Pd/Pa or iFR), Fig. 3 shows that the modest hyperemia induced by cFFR determines a more linear and less scattered relationship versus FFR compared to that of resting physiology, irrespective of diabetes status.
Diabetes did not predict agreement/disagreement between cFFR and FFR (p = 0.66) in a univariate analysis. Also when diabetes was forced into a multivariable model, it did not significantly influence the accuracy of cFFR (adjusted p = 0.71) ( Table 2).

Discussion
This sub-analysis of the CONTRAST study shows that there is no significant interaction between diabetes and the diagnostic performance of cFFR, therefore, the findings of the main analysis can be extended to patients with diabetes. Contrast FFR provides superior diagnostic performance compared with resting metrics (iFR and Pd/Pa) irrespective of diabetic status.
The CONTRAST study is the largest study to investigate the diagnostic performance of cFFR and the present sub-analysis is the first to explore the impact of diabetes on cFFR performance compared with resting methods [23][24][25].
Diabetes mellitus is a well-known risk factor for cardiovascular disease [9][10][11]. The presence of diabetes accelerates the natural course of atherosclerosis, involving both proximal epicardial and distal small vessel coronary arteries with greater plaque ulceration and thrombosis and microvascular disease. Patients with diabetes have a variety of structural and functional microvascular coronary abnormalities [12][13][14][15][16][17][18] that may translate into a reduction of hyperemic blood flow due to an impairment of microcirculatory vasodilation.
Some studies investigated the accuracy of FFR in diabetic patients. Yanagisawa et al. compared FFR performance in 96 and 149 patients with or without diabetes respectively and observed that there was no significant difference, although within diabetic patients, those with a poor glycemic control (glycated hemoglobin >7%) had a lower FFR specificity [19]. Domínguez-Franco et al. found that deferring PCI in diabetic patients (29% of all patients) with a moderately severe coronary artery stenosis based on the FFR value was as safe as in patients without diabetes [20]. Sahinarslan  lesions of 224 patients (113 non-diabetics and 111 diabetics) with stable CAD, also exploring the role of adequate glycemic control (defined as glycated hemoglobin <7%). They found that FFR accuracy was not affected by diabetic status and its glycemic control [22]. In the FAME-2 trial, the FFR guided PCI was superior to medical therapy for the primary end point (composite of death from any cause, nonfatal myocardial infarction, or urgent revascularization within 2 years) in patients with or without diabetes (28 and 72% respectively) with an absence of significant interaction between the two sub-groups interaction p = 0.50) [4]. Interestingly, some data have questioned the accuracy of FFR is some cases. Although FFR and coronary flow velocity reserve (CFVR) have an equivalent diagnostic accuracy for inducible myocardial ischemia, it has been shown that they provide discordant results in 30-40% of cases, which was suggested to have a significant impact on clinical outcomes, particularly when FFR is normal and CFVR abnormal that generally indicates predominant microvascular involvement in coronary artery disease [26,27]. Therefore it cannot be excluded that the accuracy of FFR, and thus also of cFFR, may be affected in case of microvascular disease as occurs in patients with diabetes. Indeed, a recent study has showed that in patients with diabetes, particularly those with previous MI, FFR-based deferred revascularization was associated with poor medium-term outcomes suggesting that Fig. 1 Diagnostic performance of cFFR, iFR and Pd/Pa in patients with or without diabetes. Accuracy, specificity, and sensitivity of each metric are reported for patients with (red bars) and without (blue bars) diabetes, demonstrating absence of heterogeneity in these subgroups the combining of FFR with imaging techniques may be useful to guide our treatment strategy in these patients with high-risk, fast-progressing disease [28]. Comparing FFR-guided deferred revascularization in patients with and without diabetes, Kennedy et al. also showed that those with diabetes were associated with a significantly higher rate of target lesion failure [29]. Accordingly, a recent study by Liu et al. showed that a higher FFR was associated with lower rates of death, myocardial infarction and revascularization among non-diabetic patients with deferred PCI, but in diabetic patients with deferred revascularization, FFR was not able to differentiate the risk of cardiovascular events [30].
Whether intravascular imaging, may impact on this higher failure rate remains unexplored, but the COM-BINE study, the first prospective multicenter combining FFR and optical coherence tomography, will contribute to clarify this issue investigating the hypothesis that adding the study of plaque morphology to FFR in intermediate lesions may better predict events in patients with diabetes [31].
Although few data exist on the clinical value of FFR in diabetic patients, the impact of diabetes and any potentially associated microvascular dysfunction was previously unknown for cFFR performance. Indeed, whether the submaximal vasodilatation induced by contrast medium compared with adenosine might affect the accuracy of cFFR measurement in patients with diabetes had Fig. 3 Scatterplots of each metric with FFR in patients with or without diabetes. Resting physiology (either Pd/Pa or iFR) displays a less linear and more scattered relationship with FFR than does modest hyperemia (contrast FFR) in both diabetic (red dots) and non-diabetic patients (blue dots), as shown visually by the raw data and its local regression (thick blue line with dashed 95% confidence intervals) and quantified by correlation coefficients (r from Pearson, ICC from intraclass correlation) never been explored before. Our novel results found no significant heterogeneity between subgroups stratified by diabetes, and that the overall results of the CONTRAST study can be extended to all patients irrespective of diabetic status. Therefore, cFFR has been proven accurate and superior to resting metrics in patients with diabetes as well as in those without diabetes.

Limitations
This is an unplanned post-hoc analysis of the CON-TRAST study, which was not designed or powered to investigate the diabetic subgroup specifically. Therefore, our exploratory analysis should be considered hypothesis-generating and needs to be confirmed in further trials. Additionally, our sub-analysis shares the limitations of the main study [8], namely the short-acting nature of contrast-induced hyperemia and lack of clinical data regarding contrast-induced nephropathy. Finally, being cFFR measured during submaximal hyperemia, it shares the same FFR limitations related to hyperemia in the setting of microvascular dysfunction.

Clinical relevance and conclusions
When performing physiological lesion assessment, some operators prefer to avoid adenosine due to rare side effects and minimal but nonzero added time and expense. Moreover, adenosine is still expensive or unavailable in some areas of the world, and sporadic patients have contraindications.
Contrast FFR provides superior diagnostic performance than Pd/Pa or iFR for predicting FFR irrespective of diabetes. Although FFR remains the reference standard for diagnostic certainty, for clinical scenarios or healthcare systems in which adenosine is contraindicated or prohibitively expensive, cFFR may represent a simple alternative technique to adenosine-FFR for invasive coronary physiological assessment in diabetic as well as nondiabetic patients.