Duplex ultrasound velocity criteria for the stented carotid artery

Patients and treatment 

We performed 255 CAS procedures from January 1, 1996, through December 31, 2006, in an Institutional Review Board-approved program for CAS. Demographics and laboratory results were collected in a prospective registry. Risk factors that were tabulated included coronary artery disease (currently or previously symptomatic, requiring intervention), diabetes mellitus (only if treated medically), hypertension (only if treated medically), hypercholesterolemia (only if treated medically or if serum cholesterol was >180 mg/dL), and smoking (current or former smoker).

Patients with symptomatic carotid stenosis ≥50% or asymptomatic carotid stenosis ≥70% were considered for this protocol. Eligibility was further determined according to the presence of high-risk criteria.1, 2 Lesions were treated with a WallStent (Boston Scientific Corp, Natick, Mass) or Acculink stent (Abbott Vascular, Menlo Park, Calif).

Procedural details for CAS at our institution have been published by our group previously.3, 5, 10, 17 All patients received aspirin (325 mg daily) and clopidogrel (75 mg twice daily) for at least 48 hours before the procedure. Clopidogrel (75 mg daily) was continued for 30 days after the procedure, and aspirin was continued indefinitely.

Patients early in our experience underwent CAS without embolic protection. The Accunet antiembolic device (Abbott) was used in all subsequent patients. Patients were followed up with an annual clinical and DUS examination. Residual stenosis after CAS was defined as ≥20% luminal reduction,10, 18 the presence of ISR was defined as ≥50% luminal reduction,18, 19 and hemodynamically significant ISR was defined as ≥80% stenosis. Patients underwent endovascular retreatment if their ISR reached a threshold of ≥80%, regardless of neurologic symptoms5 or ≥50% in the presence of neurologic symptoms.

Measurement of stenosis by carotid angiography 

After CAS was completed, all patients underwent multiplanar cervical digital subtraction CA. During follow-up, patients with suspected high-grade ISR underwent diagnostic CA using a similar protocol to determine the degree of restenosis. The degree of stenosis on CA was measured using North Atlantic Symptomatic Carotid Endarterectomy Trial (NASCET) criteria.20 The in-stent least luminal diameter was compared with the distal nontapering portion of the internal carotid artery (ICA), which served as the reference segment. All angiograms were analyzed off-line with a computer-assisted quantitative edge-detection algorithm (MDQM; MEDCON Telemedicine Technology Inc, Livingston, NJ) by an independent observer who was blinded to the DUS and clinical findings.

Measurement of stenosis by ultrasonography 

A postprocedural DUS examination was performed ≤3 days of CAS, and patients were subsequently monitored with annual DUS evaluations. Examinations were performed with a Sequoia 512 machine (Acuson, Mountain View, Calif) in the same Intersocietal Commission on Accreditation of Vascular Laboratories (ICAVL)-approved vascular laboratory.21 An angle of insonation of 60° was maintained, and angle correction was used where this was not possible. Technologists were blinded to angiographic and clinical findings.

Velocities were determined at distal, mid, and proximal portions of the stent and in the distal ICA, and were also measured from any areas of potential narrowing in the stent identified on B-mode imaging. The highest in-stent velocity was used for further analysis. In addition, velocities were measured in the common carotid artery (CCA) proximal to the stent. Peak in-stent systolic velocity (PSV), end in-stent diastolic velocity (EDV), the in-stent PSV/EDV ratio, and the ratio of PSV within the stent to that in the CCA proximal to the stent (ICA/CCA) were recorded for each study. The velocity criteria used to identify individual categories of primary carotid artery stenoses have been validated in our laboratory through an ICAVL accreditation process and were modified from the University of Washington criteria8: 0% to 19% stenosis, PSV <130 cm/s; 20% to 49% stenosis, PSV 130 to 189 cm/s; 50% to 79% stenosis, PSV 190 to 249 cm/s with EDV 120 cm/s; 80% to 99% stenosis, PSV ≥250 cm/s and EDV ?120 cm/s, or an ICA/CCA ratio ≥3.2.

Measurement of stenosis by computed tomography angiography 

CTA was performed using the GE Brightspeed 16-slice system (General Electric Healthcare, Waukesha, Wis). The scan technique included a detector configuration of 64 × 0.625 mm at 40 mm coverage per rotation and pitch of 1.0, gantry speed of 0.5 seconds, and scan time of 4.2 seconds. Intravenous contrast was delivered at 4 mL/s for a total volume of 150 mL. The degree of stenosis on CTA was measured using NASCET criteria.20 The in-stent least luminal diameter was identified by reviewing all sections through the stent and comparing them with the distal nontapering portion of the ICA identified by analyzing all sections distal to the stent.22 All CTAs were analyzed on a computer workstation with a digital measurement tool (GE Advantage Windows workstation) by one independent observer who was blinded to the DUS and clinical findings.

Study design and statistical analysis 

For the purposes of this study, DUS velocities were compared with the degree of stenosis after CAS as confirmed by a CTA or CA imaging study. There were three sources of data. All patients underwent CA at the conclusion of CAS, and these measurements were compared with DUS velocities obtained ≤3 days of the procedure. Patients suspected of having high-grade ISR on follow-up according to increasing DUS velocities or persistently elevated velocities underwent diagnostic CA and similar comparisons were made. Finally, at their most recent annual follow-up, all patients prospectively underwent CTA in addition to a DUS examination, thereby allowing additional comparisons between the degree of ISR and DUS velocities. Although prior reports have confirmed the accuracy of CTA for measuring carotid stenosis with a sensitivity of 95% (91% to 97% confidence intervals [CI]) and specificity of 98% (96% to 99% CI),22 we further measured comparability between CA and CTA in a subset of our patients who underwent both tests.

Statistical analysis was performed using GraphPad Prism 3.0 (GraphPad Software Inc, San Diego, Calif), SPSS (SPSS Inc, Chicago, Ill), and MedCalc 9.3 (MedCalc Inc, Mariakerke, Belgium). Categoric data are presented as percentages, and continuous data as mean (range). Scatter graphs of in-stent PSV, EDV, and ICA/CCA ratios were plotted as a function of imaged (CA/CTA) stenosis to demonstrate the magnitude and prevalence of velocities measured in the study. Linear regression was used to compare DUS velocities with CA/CTA stenosis, and CA stenosis with CTA stenosis. Significance was considered at P ≤ .05.

ROC were generated from imaged stenosis and corresponding velocity measurements to determine optimum velocity criteria for stenoses ≥20%, ≥50%, and ≥80%. Sensitivities, specificities, positive-predictive values (PPV), and negative-predictive values (NPV) for these velocity criteria were determined. The CIs for each observation were obtained. Post hoc analyses with the sequential application of PSV and ICA/CCA ratio criteria9 were also performed to assess similar accuracy parameters.

Patients 

A total of 255 CAS procedures were performed, and baseline demographic characteristics of the patients are presented in Table I. The study excluded 39 patients who had contralateral carotid stenosis ≥50%,23, 24, 25 leaving 216 patients with unilateral carotid stenosis. Completion CA or postprocedural DUS were not available or evaluable for 27 procedures, thereby providing 189 pairs of CA–DUS measurements for analysis. During follow-up, the threshold PSV ≥250 cm/s established for native carotid arteries categorized 29 stented arteries as having recurrent stenosis ≥80%. These patients underwent diagnostic CA, which provided an additional 29 pairs of CA–DUS measurements for analysis. During a mean follow-up of 4.6 years (range 1 to 10 years) on the 216 patients, 23 died, and 64 were lost. Of the remaining patients who were invited to undergo CTA along with DUS during their most recent follow-up evaluation, 28 refused or could not undergo imaging, and two imaging studies were suboptimal, resulting in an additional 99 pairs of CTA–DUS measurements for analysis. A total of 310 pairs of observations (DUS vs CA/CTA) with varying degrees of ISR were therefore available for analysis and development of velocity threshold criteria. In addition, 19 patients undergoing CAS and two patients undergoing endovascular reintervention for high-grade ISR underwent CTA ≤ 3 days of their procedure. This enabled a comparison of CA and CTA in 19 patients.

Table I. Baseline patient characteristics

	Characteristic	Percentage, or mean (range)
	Age, (range)	71.6 (52-93)
	Male gender	54.8
	Diabetes mellitus	36.3
	Hypertension	87.5
	Coronary artery disease	68.9
	Hypercholesterolemia	65.3
	Smoking	35.9
	Symptomatic	36.3
	Lesion etiology
	Radiation	9.7
	Primary atherosclerosis	39.5
	Post-CEA restenosis	50.8

CEA, Carotid endarterectomy.

Distribution of ultrasound velocity measurements 

Fig 1, A demonstrates the distribution of stenoses diagnosed by DUS in the 310 observations included in the study when our ICAVL-approved velocity criteria for native carotid arteries were used. Fig 1, B demonstrates the distribution of imaged (CA/CTA) stenoses in the same cohort. Of 237 arteries with normal luminal diameters, only 152 were correctly categorized as normal lumens by native DUS velocity criteria, 66 stenoses were overestimated in the 20% to 49% category, 9 were overestimated in the 50% to 79% category, and 10 were overestimated in the ≥80% category. This confirms that the use of DUS velocity criteria for native carotid arteries consistently overestimates the degree of stenosis in the stented artery across all degrees of stenosis. These differences persisted into the follow-up period, which extended to a mean of 4.6 years in the current cohort. None of the patients had neurologic symptoms in association with the development of ISR.

View Large Image Download to PowerPoint

Fig 1. Distribution of restenoses in 310 observations performed on patients with carotid artery stenting. Velocity threshold criteria developed for native unstented carotid arteries overestimates the number of patients with ≥20%, ≥50%, and ≥80% restenosis. Pie charts indicate number and percentage of observations in each category of stenosis. A, Distribution according to threshold criteria developed for native carotid arteries. B, True distribution of restenoses as confirmed by carotid angiography or computed tomography angiography.

Correlation of computed tomography angiography with carotid angiography 

A subset of 17 patients underwent CAS with completion CA and a postprocedure CTA. During follow-up, two patients had endovascular reintervention for high-grade ISR after a diagnostic CA and a preprocedural CTA. On linear regression analysis, the degree of stenosis measured by CTA correlated significantly (r2 = 0.88) with our CA findings in these patients (Fig 2); therefore, CTA was effective and accurate in measuring the degree of stenosis after CAS and could be used to develop DUS velocity criteria in patients being followed up.

View Large Image Download to PowerPoint

Fig 2. Computed tomography (CT) angiography correlates with carotid angiography in the identification of carotid luminal stenosis in this linear regression analysis (solid line).

Correlation of ultrasound velocities with imaged stenosis 

Scatter graphs of the in-stent PSV, the ICA/CCA ratio, and the EDV were plotted as a function of imaged (CA/CTA) stenosis (Fig 3), which confirmed that our cohort included a complete spectrum of degrees of ISR. Linear regression showed that the postprocedural in-stent PSV, ICA/CCA ratios, and EDV correlated with the imaged degree of stenosis. The correlation was strongest for PSV (P < .0001, r2 = 0.85, Fig 3, A) and ICA/CCA ratios (P < .0001, r2 = .76, Fig 3, B), and weaker for EDV (P < .01, r2 = 0.51, Fig 3, C) and PSV/EDV ratios (P = .4, r2 = .04, data not shown).

View Large Image Download to PowerPoint

Fig 3. Scatterplot shows 310 pairs of ultrasound velocity measurements plotted against percentage of stenosis as determined by carotid angiography or computed tomography angiography. The solid line shows the linear regression analysis. Data-points were available across the entire spectrum of degrees of stenosis. A, Peak in-stent systolic velocity (PSV). B, In-stent carotid artery/common carotid artery ratio (ICA/CCA). C, End in-stent diastolic velocity (EDV).

Receiver operating characteristic analysis 

ROC curve analysis was performed to obtain the sensitivity, specificity, PPV, and NPV of DUS velocity thresholds for three clinically relevant degrees of stenosis (Fig 4). A stenosis threshold of ≥80% identifies high-grade, hemodynamically significant ISR and is generally considered an indication for reintervention, a threshold of ≥50% defines patients with ISR and necessitates more frequent monitoring; and a threshold of ≥20% defines patients with suboptimal results after CAS, which also necessitates more frequent monitoring.

View Large Image Download to PowerPoint

Fig 4. Receiver operating characteristic (ROC) curve analysis was used to determine the accuracy of indicators of threshold velocities appropriate for the identification of (A) high-grade in-stent restenosis (ISR) ≥80%, (B) ISR ≥50%, and (C) residual stenosis ≥20%. ROC curves were developed for peak systolic velocity (PSV) (black line), internal carotid artery/common carotid artery ratios (red line), end diastolic velocity (EDV) (green line), and PSV/EDV ratios (blue line) for each threshold stenosis.

A larger area under the ROC curve (AUC) is a measure of improved discrimination, with 1.0 being the best. According to this criterion, PSV and ICA/CCA ratios were best able to discriminate the three selected stenosis thresholds. The AUC for detecting ISR ≥80% (Fig 4, A) by PSV was 0.99 (95% CI, 0.98 to 1.0) and by ICA/CCA was 0.98 (95% CI, 0.97 to 0.99). The AUC for detecting ISR ≥ 50% (Fig 4, B) by PSV was 0.99 (95% CI, 0.98 to 1.0) and by ICA/CCA was 0.99 (95% CI, 0.97 to 0.99); and for detecting ISR ≥20% (Fig 4, C) by PSV was 0.98 (95% CI, 0.95 to 0.99) and by ICA/CCA was 0.96 (95% CI, 0.93 to 0.98).

ROC curve analysis was used to calculate the parameters of accuracy for PSVs and ICA/CCA ratios through a complete range of values to determine optimal velocities. Sensitivity (± 95% CI), specificity (± 95% CI), PPV, and NPV values to detect ≥20%, ≥50%, and ≥80% stenoses for selected velocities are summarized in Table II, Table III, Table IV, respectively. Two important considerations went into selecting optimal velocity criteria: the first was to maximize overall sensitivity, specificity, PPV, and NPV; and the second was to emphasize that post-CAS DUS is primarily a screening tool. The clinical relevance of identifying ≥50% and ≥20% stenoses is to initiate a more intense monitoring program. Similarly, the clinical implication of identifying ≥80% stenosis includes confirmatory CA, or at least another confirmatory imaging study (CTA); therefore, a higher sensitivity and NPV are to be preferred when selecting these post-CAS velocity criteria.

Table II. Sensitivity, specificity, and positive and negative predictive value of ultrasound velocity measurements for the identification of ≥80% stenosis in the stented carotid artery

	Measurement/criterion	Sensitivity	95% CI	Specificity	95% CI	PPV	NPV
	PSV ≥
	200	100.00	82.2-100.0	86.25	81.8-90.0	32.2	100.0
	255	100.00	82.2-100.0	92.44	88.8-95.2	46.3	100.0
	300	100.00	82.2-100.0	96.56	93.8-98.3	65.5	100.0
	308	100.00	82.2-100.0	96.91	94.2-98.6	67.9	100.0
	322	100.00	82.2-100.0	97.25	94.7-98.8	70.4	100.0
	331	100.00	82.2-100.0	97.94	95.6-99.2	76.0	100.0
	335	100.00	82.2-100.0	98.28	96.0-99.4	79.2	100.0
	340⁎	100.00	82.2-100.0	98.63	96.5-99.6	82.6	100.0
	352	94.74	73.9-99.1	98.63	96.5-99.6	81.8	99.7
	358	94.74	73.9-99.1	99.31	97.5-99.9	90.0	99.7
	388	94.74	73.9-99.1	100.00	98.7-100.0	100.0	99.7
	ICA/CCA≥
	2.5	100.00	82.2-100.0	85.57	81.0-89.4	31.1	100.0
	2.77	100.00	82.2-100.0	90.03	86.0-93.2	39.6	100.0
	3.1	100.00	82.2-100.0	91.75	88.0-94.6	44.2	100.0
	3.31	100.00	82.2-100.0	93.47	90.0-96.0	50.0	100.0
	3.71	100.00	82.2-100.0	95.88	92.9-97.8	61.3	100.0
	3.73	100.00	82.2-100.0	96.22	93.3-98.1	63.3	100.0
	3.81	100.00	82.2-100.0	96.56	93.8-98.3	65.5	100.0
	4.15⁎	100.00	82.2-100.0	97.25	94.7-98.8	70.4	100.0
	4.21	94.74	73.9-99.1	97.59	95.1-99.0	72.0	99.6
	4.41	78.95	54.4-93.8	98.28	96.0-99.4	75.0	98.6
	4.87	57.89	33.5-79.7	100.00	98.7-100.0	100.0	97.3

CI, Confidence interval; PPV, positive predictive value; NPV, negative predictive value; PSV, peak in-stent systolic velocity in cm/s; ICA/CCA, peak in-stent systolic internal carotid artery/common carotid artery velocity ratio.

⁎ Optimal value.

Table III. Sensitivity, specificity, and positive and negative predictive value of ultrasound velocity measurements for the identification of ≥50% stenosis in the stented carotid artery

	Measurement/criterion	Sensitivity	95% CI	Specificity	95% CI	PPV	NPV
	PSV≥
	100	100.00	91.9-100.0	32.71	27.1-38.7	19.7	100.0
	150	100.00	91.9-100.0	85.71	80.9-89.7	53.7	100.0
	176	100.00	91.9-100.0	90.23	86.0-93.5	62.9	100.0
	200	100.00	91.9-100.0	94.36	90.9-96.8	74.6	100.0
	208	100.00	91.9-100.0	95.86	92.7-97.9	80.0	100.0
	220⁎	100.00	91.9-100.0	96.24	93.2-98.2	81.5	100.0
	223	97.73	87.9-99.6	96.24	93.2-98.2	81.1	99.6
	226	97.73	87.9-99.6	96.99	94.2-98.7	84.3	99.6
	230	95.45	84.5-99.3	97.74	95.2-99.2	87.5	99.2
	235	90.91	78.3-97.4	99.25	97.3-99.9	95.2	98.5
	260	84.09	69.9-93.3	99.25	97.3-99.9	94.9	97.4
	ICA/CCA≥
	2.0	100.00	91.9-100.0	78.57	73.1-83.3	43.6	100.0
	2.27	100.00	91.9-100.0	88.72	84.3-92.3	59.5	100.0
	2.5	97.73	87.9-99.6	93.23	89.5-95.9	70.5	99.6
	2.58	97.73	87.9-99.6	94.36	90.9-96.8	74.1	99.6
	2.63	97.73	87.9-99.6	95.11	91.8-97.4	76.8	99.6
	2.67	97.73	87.9-99.6	95.49	92.3-97.6	78.2	99.6
	2.70⁎	97.73	87.9-99.6	95.86	92.7-97.9	79.6	99.6
	2.75	93.18	81.3-98.5	96.62	93.7-98.4	82.0	98.8
	2.83	93.18	81.3-98.5	97.74	95.2-99.2	87.2	98.9
	3.19	79.55	64.7-90.2	98.50	96.2-99.6	89.7	96.7
	3.68	72.73	57.2-85.0	100.00	98.6-100.0	100.0	95.7

CI, Confidence interval; PPV, positive predictive value; NPV, negative predictive value; PSV, peak in-stent systolic velocity in cm/s; ICA/CCA, peak in-stent systolic internal carotid artery /common carotid artery velocity ratio.

⁎ Optimal value.

Table IV. Sensitivity, specificity, and positive and negative predictive value of ultrasound velocity measurements for the identification of ≥20% stenosis in the stented carotid artery

	Measurement	Sensitivity	95% CI	Specificity	95% CI	PPV	NPV
	PSV≥
	100	100.00	95.0-100	36.71	30.6-43.2	32.7	100.0
	140	95.89	88.4-99.1	84.81	79.6-89.1	66.0	98.5
	145	95.89	88.4-99.1	91.98	87.8-95.1	78.7	98.6
	146	95.89	88.4-99.1	93.67	89.8-96.4	82.4	98.7
	148	95.89	88.4-99.1	94.94	91.3-97.4	85.4	98.7
	150⁎	95.89	88.4-99.1	95.78	92.4-98.0	87.5	98.7
	159	94.52	86.5-98.5	95.78	92.4-98.0	87.3	98.3
	165	91.78	83.0-96.9	95.78	92.4-98.0	87.0	97.4
	171	90.41	81.2-96.0	96.62	93.5-98.5	89.2	97.0
	181	84.93	74.6-92.2	97.89	95.1-99.3	92.5	95.5
	200	78.08	66.9-86.9	99.16	97.0-99.9	96.6	93.6
	226	69.86	58.0-80.1	100.00	98.4-100	100.0	91.5
	ICA/CCA≥
	1.5	98.63	92.6-99.8	54.43	47.9-60.9	40.0	99.2
	1.75	95.89	88.4-99.1	69.62	63.3-75.4	49.3	98.2
	1.9	95.89	88.4-99.1	79.32	73.6-84.3	58.8	98.4
	2	93.15	84.7-97.7	86.08	81.0-90.2	67.3	97.6
	2.1	89.04	79.5-95.1	90.30	85.8-93.7	73.9	96.4
	2.13	89.04	79.5-95.1	91.98	87.8-95.1	77.4	96.5
	2.14	89.04	79.5-95.1	92.41	88.3-95.4	78.3	96.5
	2.15⁎	89.04	79.5-95.1	93.67	89.8-96.4	81.3	96.5
	2.2	86.30	76.2-93.2	93.67	89.8-96.4	80.8	95.7
	2.29	82.19	71.5-90.2	94.51	90.8-97.0	82.2	94.5
	2.4	79.45	68.4-88.0	94.94	91.3-97.4	82.9	93.8
	2.5	73.97	62.4-83.5	97.05	94.0-98.8	88.5	92.4

CI, Confidence interval; PPV, positive predictive value; NPV, negative predictive value; PSV, peak in-stent systolic velocity in cm/s; ICA/CCA, peak in-stent internal carotid artery/common carotid artery velocity ratio.

⁎ Optimal value.

For the determination of ISR ≥80% (Table II), a PSV ≥340 cm/s (sensitivity, 100; specificity, 98.6; PPV, 82.6; NPV, 100), and an ICA/CCA ratio ≥4.15 (sensitivity, 100; specificity, 97.2; PPV, 70.4; NPV, 100) were found to be optimal thresholds. Increasing either the PSV or the ICA/CCA ratio thresholds resulted in improved specificity and PPV, but at the expense of NPV and sensitivity.

The identification of ISR ≥50% (Table III) was best achieved by a PSV ≥220 cm/s (sensitivity, 100; specificity, 96.2; PPV, 81.5; NPV, 100) or an ICA/CCA ratio ≥2.7 (sensitivity, 97.7; specificity, 95.8; PPV, 79.6; NPV, 99.6). Increasing either of the thresholds resulted in a loss in sensitivity and NPV without significant gains in specificity or PPV. Finally, a residual stenosis ≥20% (Table IV) was best identified by a PSV ≥150 cm/s (sensitivity, 95.9; specificity, 95.8; PPV, 87.5; NPV, 98.7) or an ICA/CCA ratio ’2.15 (sensitivity, 89.0; specificity, 93.7; PPV, 81.3; NPV, 96.5). Accuracy parameters were modestly improved by the use of serial post hoc analysis.9 Combining the PSV and ICA/CCA ratios resulted in small increases in the PPV of each threshold: ISR ≥80% (PPV, 83.1), ISR ≥50% (PPV, 82.3), and residual stenosis ≥20% (PPV, 88.2).

Limitations 

This study includes retrospectively collected data in conjunction with prospectively collected CTA data. It is therefore subject to the limitations associated with retrospective analyses. A prospective study using the proposed thresholds will serve to further validate our findings and is currently underway. DUS criteria are influenced by the equipment used, laboratories, and the technologist performing the test. Although our results can be used as guidelines, individual laboratories must develop threshold criteria that are accurate for their own environment.

Contralateral carotid occlusion25 and stenosis ≥50%23 have been associated with increased carotid volume flow,24 resulting in an overestimation of the severity of ipsilateral disease; however, we excluded all patients with known contralateral stenoses ≥50%.

Because the observed velocity alterations appear to result from altered stent-artery biomechanics, it is possible that future alterations in stent composition and design, with consequent changes in the mechanical properties, may result in altered velocity profiles. It is not known whether these changes will be significant enough to warrant further revisions in the threshold criteria. Our own results were consistent across the two stent types used in the cohort (WallStents and Acculink stents, data not shown).

This study compared velocity with anatomic measurements performed on patients at varied periods of time after CAS. It is possible that velocity profiles are altered as a function of time after CAS, which may influence results of this study. We have reported, however, that velocity elevations are consistent ≤3 days of stenting vs >90 days of follow-up.10

The unique design of our study supplementing CA with CTA measurements to augment statistical power also introduces a limitation. It creates a nonhomogenous group, some of whom had CTA and some had CA. Although we have demonstrated that CTA correlates closely with CA in a subset of patients, our results remain subject to the assumption that this correlation holds true for the entire cohort.