Author + information
- Received October 14, 2014
- Revision received February 10, 2015
- Accepted February 12, 2015
- Published online July 1, 2015.
- Alfred C. Burris II, MD,
- Judith A. Boura, MS,
- Gilbert L. Raff, MD and
- Kavitha M. Chinnaiyan, MD∗ ()
- ↵∗Reprint requests and correspondence:
Dr. Kavitha M. Chinnaiyan, William Beaumont Hospital, 3601 West 13 Mile Road, Royal Oak, Michigan 48073.
Objectives This study sought to evaluate the diagnostic yield of triple rule out (TRO) versus coronary computed tomography angiography (CTA) scanning in patients with acute chest pain enrolled in a large statewide registry.
Background Although TRO scans provide simultaneous evaluation of coronary artery disease (CAD), pulmonary embolism (PE), and aortic disease (AD), their use is not well defined.
Methods Patients undergoing TRO or coronary CTA at 53 Michigan institutions for acute chest pain (in the emergency department or inpatient setting) in the ACIC (Advanced Cardiovascular Imaging Consortium) were included. Demographic characteristics, scan findings, and image quality parameters were compared between coronary CTA and TRO scans. The primary outcome was diagnostic yield, defined as obstructive CAD (>50% stenosis), PE, or AD; secondary outcomes were radiation dose, contrast volume, and image quality.
Results From July 2007 to September 2013, 12,834 patients underwent computed tomography scanning (TRO, n = 1,555; coronary CTA, n = 11,279). The TRO group had more women (57.1% vs. 47.8%, p < 0.001). Diagnostic yield was similar (TRO, 17.4% vs. coronary CTA, 18.3%; p = 0.37), driven by CAD (15.5% vs. 17.2%, p = 0.093); PE and AD were 1.1% and 0.4% (p = 0.004) and 1.7% and 1.1% (p = 0.046). TRO had higher median radiation (9.1 mSv vs. 6.2 mSv; p < 0.0001) and mean contrast (113 ± 6 ml vs. 89 ± 17 ml; p < 0.0001) doses. Nondiagnostic images were frequent in TRO (9.4% vs. 6.5%; p < 0.0001). In emergency department patients, PE and AD were more often detected on TRO. Among inpatients, there were no differences in overall yield or in that of PE, AD, or CAD.
Conclusions TRO was associated with slightly higher yield of PE and AD, specifically in the emergency department. This benefit comes with higher nondiagnostic image quality, radiation, and contrast doses. Although TRO may be of value in selected patients, its indiscriminate use is not warranted. The appropriate use of TRO needs to be further defined. (Advanced Cardiovascular Imaging Consortium [ACIC]; NCT00640068).
Annually, more than 5.5 million patients present to emergency departments (EDs) throughout the United States with acute noninjury-related chest pain, the majority of whom are discharged with a diagnosis other than acute coronary syndrome (1,2). Despite this, there has been a dramatic increase in advanced imaging for acute chest pain totaling more than $10 billion annually (1,3).
Although subjective and objective data assist in the evaluation of acute chest pain, patient history has proved unreliable (4). Care providers focus much of their efforts on the evaluation of obstructive coronary artery disease (CAD), pulmonary embolism (PE), and aortic dissection (AD) (2). When left undiagnosed, these are associated with increased mortality, poor prognosis, and malpractice litigation (5,6).
In the evaluation of acute chest pain with low to intermediate risk of acute coronary syndrome, coronary computed tomography angiography (CTA) provides a time- and cost-efficient option in the evaluation of obstructive CAD, with a negative predictive value of nearly 100% (7–9). In addition to its diagnostic role in evaluating chest pain, the prognostic utility of coronary CTA is increasingly being established (10,11). Despite its improved diagnostic efficiency over traditional acute chest pain evaluation protocols, coronary CTA allows only a limited assessment of the pulmonary vasculature and aortic arch.
A triple rule out (TRO) coronary CTA protocol has thus been implemented (12,13). This protocol allows for simultaneous assessment of the thorax for CAD, PE, and AD with similar diagnostic yield, median hospital length of stay, and cost of care as with traditional coronary CTA protocols (14,15). Although image quality has been shown to be equivalent, this often comes at the cost of increased contrast volume and radiation exposure (14–16).
This study was performed to compare the diagnostic yield of traditional CTA and TRO protocols in the evaluation of acute chest pain across a variety of medical centers participating in the ACIC (Advanced Cardiovascular Imaging Consortium).
The ACIC was a statewide quality improvement initiative sponsored by Blue Cross Blue Shield/Blue Care Network of Michigan and included 53 hospitals and imaging centers performing coronary CTA (17). Participating centers received internal review board approval and included a waiver of consent. The ACIC database includes demographic characteristics, risk factors, symptoms, results of previous testing, and medical history. To ensure data accuracy, patients were interviewed at the point-of-service for symptoms and medical history. Pre-test likelihood of CAD is calculated using the Diamond-Forrester criteria. Nurses and/or technologists recorded information about medication use, vital signs, and procedural times. Information was also provided about the scanner model and manufacturer and protocol parameters. Effective radiation exposure and intravenous contrast doses were recorded. Coronary artery calcium scoring scans were performed only when specifically requested by the ordering physician.
Patients undergoing computed tomography (CT) scanning in the ED or as inpatients at these centers from 2007 to 2013 were included in this analysis. Outpatient studies were excluded.
Coronary CTA protocol and image interpretation
Coronary CTA was performed on various types of scanners available at each institution, ranging from 64- to 320-slice scanners, with scan techniques dictated by standard clinical protocols at each site (18). Multiple manufacturers (Siemens Healthcare [Erlangen, Germany], Philips Medical Systems [Eindhoven, the Netherlands], GE Healthcare [Milwaukee, Wisconsin], and Toshiba Medical Systems [Otawara, Japan]) with single- or dual-source systems were used. Beta-blockers were typically administered before scanning with a target heart rate as close to 60 beats/min as possible. Sublingual nitroglycerin was also given before scanning to patients who had no contraindication. Typical tube voltage was based on body mass index typically ranging from 100 to 120 kV, but as low as 80 kV on selected scans. Both prospective and retrospective scanning protocols were used (Table 1). Prospective includes both axial or “step-and-shoot” scanning and high-pitch spiral scanning. Topograms were obtained to determine scan range. Tube current, gantry rotation time, pitch, and collimation varied based on the protocol and manufacturer. Iodinated contrast infused during the scan varied based on scan length and time, patient habitus, and cardiac output. Details of bolus tracking/timing or number of contrast bolus phases were not specified. Protocols were generally determined based on the reading physician’s discretion; there were no pre-specified parameters across centers with regard to heart rate, arrhythmia, body habitus and other patient-related factors.
Coronary CTA scans were interpreted at each institution by cardiologists and/or radiologists with Level II (or III) training. Coronary stenoses were evaluated using a 16-segment model. Obstructive CAD was defined as coronary stenosis >50% in major epicardial vessels (left main, proximal, and mid segments of the left anterior descending, left circumflex, and right coronary arteries and the first and second diagonal and obtuse marginal branches). At centers where a cardiologist was the primary reader, all studies were read by a radiologist for noncardiac pathology.
The quality of each study was rated by the reading physician at every site on a scale of 1 to 4. Excellent (a score of 1) was defined as complete absence of motion artifacts, excellent signal-to-noise ratio, and clear delineation of vessel walls, with the ability to assess luminal stenosis as well as plaque characteristics. Good (a score of 2) was defined as nonlimiting motion artifacts, reduced signal-to-noise ratio, and/or calcifications present, with preserved ability to assess luminal stenosis as well as plaque characteristics. Adequate (a score of 3) was defined as reduced image quality due to any combination of noise, motion, poor contrast enhancement, or calcium that significantly impairs ease of interpretation, but image quality is sufficient to rule out significant stenosis. Nondiagnostic (a score of 4) was defined as reduced image quality that precludes adequate assessment of stenosis in the majority of vessels (18). Individual readers based their rating on clinical impression; sample images were not provided.
Estimation of radiation dose
Radiation doses were estimated by previously described methods (19,20). Each scanner provided a protocol summary containing the dose/length product for each image series, which integrated estimated absorbed radiation in the x, y, and z directions based on the CT dose index volume. The total dose for the entire CT examination was then used to derive the effective radiation dose using the summed dose/length product multiplied by the European Working Group for Guidelines on Quality Criteria in Computed Tomography conversion coefficient (kappa = 0.014 mSv/mGy·cm) (21).
The primary endpoint of this study was diagnostic yield: a composite of coronary artery diameter stenosis >50%, PE, and AD. The threshold of 50% luminal narrowing was used because this often prompts further invasive or noninvasive coronary evaluation. Secondary endpoints included radiation dose, contrast volume, and image quality.
All statistical analyses were performed using version 9.3 of SAS for Windows (SAS Institute, Cary, North Carolina). The study population was divided into 2 groups based on the protocol: 1) coronary CTA; and 2) TRO. Using Pearson’s chi-square where the expected frequency was >5 or the Fisher exact test, the categorical variables were reported as numbers and percentages. Continuous variables were examined using logistic regression, with TRO as the dependent variable and reported as mean ± SD or median and 25th, 75th percentiles where appropriate. Backward elimination logistic regression analysis was performed to determine whether TRO was an independent predictor of diagnostic yield as well as to determine predictors of uninterpretable scans. The following significant variables were included: sex, age, body mass index >30 kg/m2, Framingham Risk Score, scanning mode, tube strength, radiation dose (mSv), heart rate (beats/min), contrast volume (ml), and history of atrial fibrillation, valve disease, chronic obstructive lung disease, or congestive heart failure.
A total of 12,834 patients from July 2007 to September 2013 at 53 institutions in Michigan met the study criteria. Of these, 11,279 (87.9%) were in the coronary CTA group and 1,555 (12.1%) were in the TRO group. Demographic characteristics were similar in both groups (Table 2). Those undergoing TRO were more often women (57.1% vs. 47.8%; p < 0.0001) with a lower CAD pre-test likelihood according to the Diamond-Forrester criteria and a lower frequency of tobacco use, family history of premature CAD, hypertension, and hyperlipidemia.
Composite diagnostic yield determined by the presence of obstructive CAD (stenosis >50%), PE, and/or AD was 17.4% for TRO and 18.3% for coronary CTA (p = 0.37) (Table 3). This was driven by obstructive CAD: 15.5% in TRO and 17.2% in coronary CTA (p = 0.93) and accounted for 88.5% and 93.6% of diagnoses, respectively. Diagnostic yields of PE and AD were 1.1% and 0.4% (p = 0.0004) and 1.7% and 1.1% (p = 0.046), respectively. Some patients had more than 1 diagnosis. A multivariable model for diagnostic yield adjusting for the differences between the TRO and coronary CTA subjects found that having a TRO scan was not predictive (odds ratio: 0.87, 95% confidence interval: 0.71 to 1.07) of diagnostic yield.
PE and AD were diagnosed more often on TRO than on coronary CTA in centers that routinely performed TRO and coronary CTA, with no statistically significant difference in the overall diagnostic yield. Uninterpretable studies were more often reported with TRO than with coronary CTA. Similarly, PE and AD were more often detected on TRO than on coronary CTA among ED patients. Among inpatients, however, there was no difference in the detection of PE and AD between TRO and coronary CTA (Table 3).
The median radiation dose was 49% higher for TRO than CTA: 9.1 mSv (interquartile range: 6.4 to 13.5) versus 6.2 mSv (interquartile range: 3.9 to 11.6) (p < 0.0001) (Table 4). For TRO, the median dose varied based on protocol and type of scanner; the lowest radiation dose was associated with high-pitch spiral scanning. Mean contrast volume was 27% higher for TRO than for coronary CTA: 113 ± 16 ml versus 89 ± 17 ml (p < 0.0001). Nondiagnostic or uninterpretable image quality was noted more frequently in TRO (9.4% vs. 6.5%; p < 0.0001) with similar reported reasons (Table 5). Results of the multivariable model found that the factors that contributed to poor image quality were higher heart rates, body mass index >30 kg/m2, valvular heart disease, congestive heart failure, higher contrast volume, and higher Framingham Risk Score (Table 6).
In this statewide study, the overall diagnostic yield was similar between TRO and coronary CTA. TRO was associated with a slightly higher yield of PE and AD, but also with higher nondiagnostic image quality and radiation and contrast doses. These data suggest that although TRO is feasible, it cannot be recommended in all patients presenting with acute chest symptoms. Clinical scenarios with the suggestion of increased risk of PE and/or AD in addition to CAD may present the best use for TRO.
Although the composite diagnostic yield in both groups were comparable, this was predominantly driven by obstructive CAD, which was 15 times that of PE and AD. Although PE and AD were infrequent, they were diagnosed more frequently in the TRO group, unlike a similar study comprising a smaller cohort of patients with similar risk profiles (14). It is important to note that in that previous study, there was a lower diagnostic yield for PE in the coronary CTA group (1.1% with TRO and 0.2% with coronary CTA; p = 0.05) with no AD in either group. Given similar patient risk factor profiles in the present study, these differences were most likely driven by sample size as well as varied protocols across centers. The diagnostic yield for PE in this study in TRO and coronary CTA was much lower than the 9% to 19% in dedicated PE protocols (22–24). However, in those studies, patients were selected for CT examinations based on clinical presentation and/or traditional risk stratification models. Although the benefits of risk stratification for PE with d-dimer, Wells criteria, and Geneva criteria are well described (25,26), their effects on the diagnostic yield of coronary CT (CTA or TRO) are yet to be demonstrated. The standardized use of risk stratification tools would likely improve the diagnostic yield and thus prevent unnecessary use of TRO.
It is important to note that even though the rates of PE and AD on coronary CTA (with limited field of view) were lower than in dedicated protocols, they were high for unsuspected but clinically important diagnoses. However, >97% of TRO studies showed neither AD nor PE. These discordant findings between scan intention and results highlight the fact that the diagnosis of acute chest pain can be challenging, particularly because clinical presentations are often atypical or unclear.
The majority of the overall studies (64%) were performed in the ED. There was a statistically significant difference in the diagnostic yield of both PE and AD with TRO in the ED, but not among inpatients. This could reflect alternate indications for ordering the study as an inpatient (for example, further evaluation of an equivocal stress test in the ED or clarification of an abnormal chest x-ray).
TRO was associated with higher median radiation doses and contrast volumes. Although the increased scan time and scan length required with TRO protocols are generally considered the main cause of an increased dose, the difference in scan length between coronary CTA and TRO in this study was only 3.6 cm. These data suggest “overscanning” on coronary CTA or “underscanning” on TRO at least at some sites; with rigid adherence to prescribed scan lengths, the difference in radiation dose could be conceivably higher. These data also demonstrate the overall ordering patterns of both protocols. Patients with higher pre-test likelihood of CAD underwent coronary CTA, whereas those with a lower likelihood underwent TRO. Thus, TRO included younger patients and more women, the specific populations in which radiation dose exposure must be limited (27). Additionally, the diagnostic yield among women was one-half of that in men. This finding highlights the fact that women more often present with atypical symptoms, but also received unnecessary radiation and contrast. Although the diagnosis of PE on TRO in the women who had the disease is a positive, the lack of such diagnoses in the majority of those scanned highlight the challenges and limitations of pre-test clinical identification of patients with potential life-threatening conditions.
There have been extensive, successful efforts by the ACIC to optimize protocols to decrease radiation dose (28). Our median radiation dose reflects both older and newer protocols. Most of the scans included in this study were performed using 64-slice scanners. Centers using dual-source scanners capable of high-pitch scanning reported <50% of the median radiation dose. Also, low-dose protocols for 320-row scanners can decrease radiation exposure to <3.5 mSv and total contrast volume to 60 ml without sacrificing image quality (29).
TRO was associated with a 45% higher frequency of nondiagnostic image quality compared with coronary CTA. Although the TRO scan type itself was not a significant predictor of nondiagnostic image quality on multivariate analysis, it is likely that this protocol is affected by the type of contrast bolus used as well as body habitus. Higher heart rates and ectopy (resulting in motion artifacts) and clinical conditions that may result in the patient’s inability to lie flat were found to be significant predictive factors for scan uninterpretability. These results further demonstrate the need for appropriate patient selection for TRO to avoid unnecessary radiation exposure from a nondiagnostic scan, leading to additional testing with their associated risks and costs.
This large multicenter study suggests that TRO provides a slightly increased diagnostic yield for PE and AD compared with coronary CTA. It is unclear whether this is related to liberal use of TRO. Nonetheless, patients undergoing TRO are more often low-risk women with an overall low yield on CT scanning. Thus, these results do not necessarily tip the balance in favor of TRO use in all patients with acute chest symptoms.
Currently, there are no established guidelines outlining pre-scan evaluation of indications for TRO. The effects of clinical evaluation and subsequent patient selection on diagnostic yield are unknown. The results of this study suggest that although there is a potential utility for TRO in the evaluation of acute chest pain, appropriate patient selection through established clinical algorithms is necessary. TRO must be used judiciously and not indiscriminately, particularly because nearly 1 in 10 TRO scans was nondiagnostic despite the higher radiation and contrast doses necessarily associated with it. TRO may be considered when clinical features and initial laboratory data (e.g., indeterminate troponin and elevated d-dimer) raise concern for: 1) obstructive CAD; and 2) PE or AD. Therefore, we propose an algorithm that divides patients presenting with acute symptoms into 3 risk groups: high, low to intermediate, and very low (noncardiac) (Figure 1). Both coronary CTA and TRO are best avoided in patients with known CAD; such patients may be best suited for stress testing with/without imaging.
Among low- to intermediate-risk patients, clinical suspicion of PE must warrant additional evaluation of risk that includes the Wells criteria or Geneva criteria (e.g., history of deep vein thrombosis, malignancy, tachycardia, immobilization, age older than 65 years) and suggestive laboratory data (increased d-dimer) (25,26). If the suspicion of PE remains intermediate with such detailed evaluation along with that for CAD, TRO may be considered after meticulous patient preparation (Figure 2).
Similarly, intermediate suspicion of AD (1 high-risk feature in history, symptoms, or physical examination) along with continued suspicion of CAD may warrant TRO (Figure 3). A double rule out protocol may be considered in patients with an overlap in diagnosis of AD and CAD with inclusion of the abdominal aorta, without the extra contrast for opacification of the pulmonary circulation. More than 1 high-risk feature in the patient’s history (Marfan syndrome, aortic valve disease or recent aortic surgery, known aortic aneurysm), symptoms (abrupt onset of ripping/tearing/stabbing chest pain), and physical examination (pulse deficit, blood pressure difference in limbs, perfusion deficits, murmur of aortic insufficiency) would warrant triage to dedicated AD pathway for diagnosis and management (30).
The most important limitation of this study is its inability to distinguish the source of the increased diagnostic yield of TRO for PE and AD. There are 2 possibilities: 1) physicians appropriately selected patients for TRO, yielding a higher incidence of PE or AD; or 2) the TRO procedure has a higher sensitivity due to a contrast bolus appropriately timed for right-sided circulation. We were able to retrospectively evaluate 8 of 17 patients in the TRO group who had PE and found that each had filling defects that fell within the imaging window of standard coronary CTA. However, it is foreseeable that PE may have been missed on coronary CTA scans due to a limited field of view and/or varying protocols.
As a retrospective study, it is difficult to make definitive comparative conclusions. There are multiple participating centers using varied chest pain and imaging algorithms; thus, this study reflects varied practice patterns. Protocols and physician threshold for ordering coronary CTA vary within and across different institutions, with the potential for selection bias. It is important to note that TRO can often be used inappropriately as a screening tool to evaluate all chest pain. Data on varying reconstruction methods that could have contributed to image quality were not collected. There is no discrimination of studies duplicated from patients receiving care at multiple medical centers.
In this large registry, TRO was associated with slightly higher diagnostic yield for PE and AD compared with coronary CTA, particularly in the ED, and was used more often in younger women. Additionally, TRO is associated with significantly higher radiation and contrast doses compared with coronary CTA. Thus, although feasible, clinical criteria for TRO use and appropriateness must be further defined.
COMPETENCY IN PATIENT CARE AND PROCEDURAL SKILLS: In patients with a low to intermediate likelihood of CAD and PE/AD, TRO may result in higher diagnostic yield compared with coronary CTA. Due to the significantly higher radiation and contrast doses associated with TRO, it must be judiciously applied in clinical practice.
TRANSLATIONAL OUTLOOK: Although the diagnostic yield for PE and AD was higher in TRO compared with coronary CTA, it is associated with higher radiation and contrast doses as well as more frequent nondiagnostic image quality. Thus, patient selection and appropriate use of TRO must be further defined.
This study was sponsored and funded by the Blue Cross/Blue Shield/Blue Care network of Michigan (BCBSM), Detroit, Michigan. The authors have reported that they have no relationships relevant to the contents of this paper to disclose.
- Abbreviations and Acronyms
- aortic dissection
- coronary artery disease
- computed tomography angiography
- emergency department
- pulmonary embolism
- triple rule out
- Received October 14, 2014.
- Revision received February 10, 2015.
- Accepted February 12, 2015.
- American College of Cardiology Foundation
- ↵Bhuiya FA, Pitts SR, McCaig LF. Emergency department visits for chest pain and abdominal pain: United States, 1999-2008. NCHS Data Brief, no 43. Hyattsville, MD: National Center for Health Statistics, 2010.
- Heller G.V.,
- Stowers S.A.,
- Hendel R.C.,
- et al.
- Goldstein J.A.,
- Chinnaiyan K.M.,
- Abidov A.,
- et al.
- Hoffmann U.,
- Bamberg F.,
- Chae C.U.,
- et al.
- Hulten E.A.,
- Carbonaro S.,
- Petrillo S.P.,
- Mitchell J.D.,
- Villines T.C.
- Min J.K.,
- Shaw L.J.,
- Devereux R.B.,
- et al.
- Morin R.L.,
- Gerber T.C.,
- McCollough C.H.
- ↵(2007) The International Commission on Radiological Protection. The 2007 recommendations of the International Commission on Radiological Protection: ICRP publication 103. Ann ICRP 37:1–332.
- Chinnaiyan K.M.,
- Boura J.A.,
- Depetris A.,
- et al.
- Hiratzka L.F.,
- Bakris G.L.,
- Beckman J.A.,
- et al.