As David Newman will tell you, the additive prognostic and predictive value for stress tests is fairly weak. CT coronary angiograms are still a test looking for the correct population. Conventional coronary angiography is expensive, invasive, and clearly not appropriate for the massive population of low risk patients we evaluate.
So, how about a non-invasive test that combines three-dimensional anatomic coronary reconstruction with predictive flow dynamics to identify lesions resulting in ischemia? This test is CT coronary angiography combined with computed fractional flow reserve. And, it would be a beautiful thing if it were ready for primetime – but it’s not.
I’ve reviewed a previous trial of FFR(CT). This is a larger study, published in JAMA, of 285 patients with suspected CAD who underwent CCTA with FFR(CT), followed by conventional coronary angiography with invasive FFR measurement. Figure 1 summarizes the results relatively succinctly – but essentially, 56 of the 172 FFR(CT) patients with lesions calculated as ischemic were false positives. On the flip side, 67 of the 80 FFR(CT) patients with lesions calculated as non-ischemic were false negatives. The per-vessel performance of FFR(CT) basically added no additional diagnostic AUC to CT alone.
The study is sponsed by HeartFlow, and authored by several physicians disclosing conflicts of interest with diagnostic imaging manufacturers. Unsurprisingly, the authors try to spin the positive out of it in their conclusions and abstract.
“Diagnostic Accuracy of Fractional Flow Reserve From Anatomic CT Angiography“
http://jama.jamanetwork.com/article.aspx?articleid=1352969
Category: Radiology
Head to Head With Head CT Rules
The “headline” you’ll see from this article is that the Canadian Head CT Rule outperforms the New Orleans Criteria for radiographic imaging in minor head trauma. Specifically, it outperforms it in this prospective, observational cohort from several hospitals in Tunisia, consecutive patients with blunt trauma to the head and at least one symptom secondary to the head trauma.
The most striking thing about this article, however, remains the gruesome number of false positives generated by each of these head CT decision rules. While, obviously, the intent is to capture all the cases requiring neurosurgical intervention, the New Orleans Criteria could not rule out potential need for neurosurgical intervention in 1,180 out of 1,582. When the theoretical purpose of these rules is to prevent “scanning everyone”, we’re not getting much bang for our buck. The Canadian Head CT Rule was better – but still indicated a need for scan in 656 out of 1,582.
While the article focuses mostly on the need for neurosurgical intervention in GCS 15 patients, it’s interesting to see their “secondary outcomes” which did not need “intervention”. Only 34 total patients in their cohort required intervention – while they found 133 skull fractures, 41 subdurals, 45 epidurals, 69 subarachnoids/hemorrhagic contusions, and 1 case of pneumocephalus. The Canadian rule would have missed 11 of the 218 “clinically significant” findings, for a sensitivity of 95%. The article does not specific precisely which types of findings were missed, but, clearly, many of those may be argued to be not significant. Unfortunately, deriving a better rule based on a more liberal definition of “clinical significance” is likely to result in more missed interventions – but it’s still probably worth trying.
“Prediction Value of the Canadian CT Head Rule and the New Orleans Criteria for Positive Head CT Scan and Acute Neurosurgical Procedures in Minor Head Trauma: A Multicenter External Validation Study”
http://www.ncbi.nlm.nih.gov/pubmed/22251188
Excitement For/Failure of CCTA
The third of the big CT coronary angiography studies from the last year – and, yet again, this is positive for its primary endpoint.
However, that value of that endpoint is another matter – mean length of stay in the hospital. For the CCTA cohort, that mean was 23.2 hours and the “standard evaluation” was 30.8 hours. However, more illuminating – and further favoring CCTA – is that the median CCTA evaluation time was 8.6 hours compared with 26.7 hours in the “standard evaluation” group. Just like in the previous studies, CCTA is faster, and, for some patients, much, much faster.
But, as you can probably gather from that mean/median discrepancy, a substantial cohort in the CCTA group went on to have some pretty extensive downstream testing and prolonged hospital stays. This means, from a costs standpoint, the two strategies eventually even out. No significant safety differences were detected between the two strategies.
Now that we’ve seen the full results of ROMICAT II, CT-STAT, and ACRIN-PA, we have a pretty good idea of what this test does. If you must evaluate these low-risk chest pain patients with imaging of some sort, need to clear them out of your Emergency Department quickly, your cardiology team is excited to take on the false positives, and you’re unconcerned about the downstream harms – then CCTA is the test for you. If the potential harms, the poor specificity, and the non-functional nature of the test concerns you – then no one will fault you for dragging your feet.
The accompanying editorial gets it right – this is still a test looking for the correct application. However, we don’t just need a better test – we need a better consensus for whom we’re simply not going to test.
“Coronary CT Angiography versus Standard Evaluation in Acute Chest Pain”
Nephropathy Was As Common as PE after CTPA
It’s Jeff Kline Week at EMLitOfNote, with the second Carolinas paper this week – and, as a Patient Safety and Quality Fellow, I just can’t help but cite articles that deal with the consequences of otherwise well-meaning practice.
This small study followed 174 patients undergoing CTPA demonstrated a yield of 7% for PE. On the other hand, this same cohort demonstrated a yield of 14% for contrast-induced nephropathy – as defined by an increase in serum Cr of 0.5 mg/dL or >25%. Three of the 24 patients with CIN progressed to severe renal failure, two of whom died. The proportion of CIN and renal failure were similar to the outcomes observed in the additional 459 patients they followed for CT imaging on other contrast protocols.
So, the rate of CIN is not insignificant – particularly compared to the rate of diagnosis of PE at this institution. It seems to be suggested by this study, although not shown, that the relative risk of death conferred by receiving contrast and developing CIN might even exceed the number of adverse events that might have occurred from PE if left undiagnosed or untreated.
“Prospective Study of the Incidence of Contrast-induced Nephropathy Among Patients Evaluated for Pulmonary Embolism by Contrast-enhanced Computed Tomography”
http://www.ncbi.nlm.nih.gov/pubmed/22687176
A Little Proof of Harms from CTs
It is popular to worry about the harms of CT scans in small children. A retrospective Swedish study suggests decreased intelligence. And, our models based on nuclear weapon exposure data combined with dummy CT exposure suggest these scans are likely to result in an increased risk of malignancy.
This is another retrospective study in the National Health Service of Britain comparing malignancy outcomes with their exposure to CT in childhood. The scary headline: CT scan radiation triples the risk of leukemia and primary brain malignancy. Of course, triple the risk is essentially 1 additional case of leukemia and 1 additional case of primary brain malignancy in the first 10 years after exposure. So, this is potentially another study you can use to discuss the Number Needed to Harm with families when discussing the need for CT radiation in pediatric cases.
Now, whether articles like this trigger a wave of legal trolling for malignancies preceded by CT remains to be seen….
“Radiation exposure from CT scans in childhood and subsequent risk of leukaemia and brain tumours: a retrospective cohort study”
http://www.thelancet.com/journals/lancet/article/PIIS0140-6736(12)60815-0/abstract
Everyone Is On the Cardiac CT Bandwagon
The NEJM is on the wagon with their recent publication. Annals of EM has been publishing all the ROMICAT trials. And, not to be outdone, the American College of Cardiology is publishing the CT-STAT trial – a head to head comparison between coronary CT angiogram in the Emergency Department and stress perfusion imaging.
The endpoint of interest, however, is length of stay – and by association total index visit costs – rather than accuracy or safety. And, in this sense, it was successful. The primary difference in LOS was the length of time it took to perform the CT or stress test, which was approximately 4 hours quicker in the CT group. ED costs were also lower, somehow, presumably billing for an observation code while awaiting the stress test and results.
However, what the authors don’t include are the total downstream costs and time of additional testing after the Emergency Department visit. The stress test group had 34 abnormal or non-diagnostic scans, while the CT group had 64. 27 patients in the stress group underwent additional testing vs. 51 in the CT group – mostly stress tests that were subsequently normal – and none of these costs or times are included in their analysis. I imagine if these extra tests are included in their analysis, the cost difference shrinks or disappears.
It seems to be a trend to advertise more than CT angiography actually delivers.
Several authors are sponsored by Siemens.
“The CT-STAT (Coronary Computed Tomographic Angiography for Systematic Triage of Acute Chest Pain Patients to Treatment) Trial”
www.ncbi.nlm.nih.gov/pubmed/21939822
Defensive Medicine is Defensive
Sometimes, people order a CTA chest to evaluate for pulmonary embolism because they’ve used the available evidence to risk-stratify the patient for a pulmonary embolism, and it’s an important diagnosis to make. Sometimes, people order CTAs of the chest to evaluate for pulmonary embolism out of defensive practice, in order to avoid missing a pulmonary embolism.
There are some holes in this paper, considering how few patients in their cohort received the study intervention. However, the general statistical gist was is that physicians who indicated that defensive medicine played a role in their ordering decisions had a much lower yield on their CTA for PE. Conversely, elevated Wells/Geneva scores were associated with higher yield CTA. Positive d-Dimers and patient request were non-significantly positively associated with increased CTA yield.
Not precisely an earthshaking paper, but it does weakly reinforce what we probably already suspected – defensive medicine harms the patient and the healthcare system.
“Ordering CT pulmonary angiography to exclude pulmonary embolism: defense versus evidence in the emergency room”
www.ncbi.nlm.nih.gov/pubmed/22584801
Plain C-Spine Radiography in Children
In adults, the use of plain radiography has largely been replaced in the U.S. by computed tomography over concerns regarding missed injuries – and some literature even argues that, given the right clinical circumstances, even a normal CT scan is inadequate. But, in children, the harms of radiation exposure are greater, so pediatrics has been more hesitant to move to CT as the first imaging study of the cervical spine in blunt trauma.
Unfortunately, this retrospective PECARN study of children with cervical spine injuries isn’t as helpful as one would hope. The authors identified 204 children, 58 of whom were aged less than 7 years, who sustained a CSI and had plain radiographs of the cervical spine performed. Of these patients, 127 patients had a definite injury on plain radiography. 41 additional patients had “possible” abnormalities. Then, 20 films were judged to be inadequate by technique. And, finally, there were 18 adequate radiographs with normal findings who subsequently had a CSI identified. The overall sensitivity, then, was 90% (CI 85-94%) – which compares very similarly to the sensitivity in adults from the 34,000 patients in the NEXUS study.
The authors note that most missed injuries fell into two general categories: they were either subtle and non-morbid, or the patients were altered/intubated/focal neurologic findings. It is probably still reasonable to start with screening plain-film radiography and use clinical judgment to determine when CT may be necessary, but if you’re looking for airtight evidence to guide your decision-making, CSI in children is too rare to generate that sort of data.
“Utility of Plain Radiographs in Detecting Traumatic Injuries of the Cervical Spine in Children”
CCTA Is A Bad Shortcut Around Bad Care
“Although an acute coronary syndrome is ultimately diagnosed in only 10 to 15% of patients who present with chest pain, the majority of these patients are admitted to hospitals, at an estimated cost of over $3 billion annually.”
This is, essentially, the statement of problem from the NEJM article, sponsored by Siemens, regarding the use of coronary CT angiograms in the Emergency Department on low-to-intermediate risk chest pain. They are clearly huge fans of CCTA up at the University of Pennsylvania, and I hate to think it has something to do with the parade of imaging technology companies and patent applications listed as disclosures by the authors.
In this study, the authors enrolled 1,392 patients with chest pain with the goal of testing the primary hypothesis that “patients without clinically significant coronary disease on CCTA (i.e., no coronary-artery stenosis ≥50%) would have a 30-day rate of cardiac death or myocardial infarction of less than 1%.”
Good news! They were right. Bad news: their entire enrolled cohort had a 30-day death or MI rate of only nearly 1%. It’s rather incredible, really, that they have this entire article in which they sing the praises of CCTA for identifying low-risk chest pain, when in reality, they gloss over the fact that they simply could have sent home every single patient in the study without doing a single additional test and only nearly 1% would have had death or MI within 30 days.
Going back to their essential statement of problem, it might be true that CCTA were valuable if they were looking to apply it in a population and practice environment in which we were actually hospitalizing patients with a 10-15% rule-in rate. However, the opposite of what these authors propose is the real truth – clinically identify all the low-risk chest pain and stop doing all these expensive tests! They claim it expedites discharge from the Emergency Department, which, in theory, saves money – but it isn’t! Despite 90% of their CCTA being negative for stenosis >50%, they still end up admitting half their CCTA cohort. Even the negative CCTA cohort, while their length of stay is reduced to 12 hours, still means they’re being placed in observation status and billed an additional separate observation code – which in many places is a protocolized chest pain observation unit run by the Emergency Department.
This is simply a bad solution to bad baseline practice patterns. The measurable benefit here isn’t to the patient, it’s to the malpractice risk of the physician, to Siemens and other sponsors of the study, and likely to the Emergency Departments whose billing increases for these short stay observation patients.
“CT Angiography for Safe Discharge of Patients with Possible Acute Coronary Syndromes”
http://www.nejm.org/doi/full/10.1056/NEJMoa1201163
Pan-Scan & Zero-Miss
In an interesting contrast to a prior article regarding over-scanning in trauma, this article takes a different perspective. While the authors do note that there is controversy regarding the impact of “pan scan” on survival, they are rather focused on the sensitivity/specificity of the “pan scan,” rather than the appropriateness.
This is a review of 982 consecutive patients undergoing “pan scan” in Germany. The indications for scanning were a set of “red flag” criteria, which included impaired patients, patients with obvious injuries, “suspicion of severe trauma” or “high risk mechanism”. The diagnostic reference standard was chart review by two reviewers of the electronic notes for any injuries missed on the initial scan.
The results are rather interesting in a couple ways: the prevalence of injuries per organ system is not terribly high, and the sensitivity of scanning was rather low. The highest prevalence of injuries for an organ system was 37%, for chest, followed by head and neck at 34%. However, the sensitivities range from a high of only 86.7% for chest down to a low of 79.6% for face – likely because dedicated fine cuts of the face were not part of their protocol. Regardless, with sensitivities in the mid-80s meant they missed almost one seventh of the total number of injuries. Of these 70 missed injuries, almost half required surgery or a critical intervention as treatment.
So, pan-scanning: expensive, low yield, yet still misses important injuries. The authors do not try to fully address whether their yield is reasonable or not, and wisely simply state further research is needed regarding triaging patients into groups likely to benefit from scanning.
“Accuracy of single-pass whole-body computed tomography for detection of injuries in patients with major blunt trauma.”
http://www.ncbi.nlm.nih.gov/pubmed/22392949