Keywords
aorta - aneurysm - dissection - risk prediction - genetics
Introduction
The journey to create a model of stratification for risk prediction of adverse outcomes
after the incidence of acute aortic dissection (AD) encompasses an array of factors.[1]
[2]
[3]
[4]
[5] The goal of such risk models lies in the ability to be incorporated into the decision-making
framework for treatment, as well as future care to be provided for the patients and
their families.
Numerous studies[1]
[2]
[3]
[4]
[5] have attempted to create a model to aid in the risk prediction post-AD, in hopes
of guiding the surgeon and team in management of their patients. Quantitative and
qualitative models have been devised, considering different factors, including varying
degrees of malperfusion, blood biochemistry, and perioperative stability, alongside
other factors.[1]
[2]
[3]
[4]
[5] Across these models, the immediate aim is to stratify the risk of operative and
30-day mortality. However, the Penn classification also stratifies the risk of later
mortalities up to 5 years.[6] Many of these studies are significantly hindered by intrinsic retrospective limitations
and the majority of predictive paradigms still require further validation.[1]
[2]
[3]
[4]
When we start to think about risk prediction following AD, how can we ignore the risks
that build up prior to the dissection itself? We must consider both syndromic and
nonsyndromic cases of AD, as well as familial versus sporadic within the latter data.[7] Risk prediction after the dissection has occurred builds on the very foundation
of risk of the initial dissection itself. The family history and known diseases linked
to syndromic dissection exert significant influences on the ability to make a prompt
diagnosis of AD. Delays in diagnosis and treatment influence the risk of adverse effects
postdissection.
This paper reviews the literature to not only critically analyze some of the current
models looking to stratify risk prediction postdissection, with their burdening limitations,
but also to bring into consideration genetic factors, the natural history of dissection,
and the role of prior screening. We explore the potential for these two different,
yet clearly related, aspects of risk prediction in AD.
The Need for Risk Models
A plethora of nonspecific symptoms, an insidious onset, with a lethality rate of 1
to 2% per hour, and a vast range of potential symptoms[8] are just a few of the hurdles challenging an efficient diagnosis of AD.[9] The concomitant events of carotid and cerebral malperfusion worsen the prognosis
not only due to poorer clinical conditions of the patients, but also because of the
heightened risk of misdiagnosis.[8]
[10] Any tool to support the efficacy of the decision-making and rapid risk stratification
for intervention is invaluable. Stratifying the risk of adverse outcomes postdissection
will guide the surgeon toward the optimum therapeutic choice.
Current Risk Models: Breaking the Code
Current Risk Models: Breaking the Code
Malperfusion
The impact of malperfusion, with its severe adverse implications, has been well-appreciated,
notably illustrated by the German Registry for Acute Aortic Dissection Type A (GERAADA)
analysis by Czerny et al[7] ([Fig. 1]) and the Penn classification.[5]
Fig. 1 Impact of different types of malperfusion as independent risk factors for early mortality.[7]
As per the model developed by Augoustide and colleagues, the Penn classification,
patients were stratified as 4 levels of ischemia. This classification was defined
by authors as follows: class Aa, presence of no ischemia; class Ab, localized ischemia;
class Ac, generalized ischemia; and class Abc, localized and general ischemia.[5] The prospective single-centered study analyzing 221 patients, demonstrated an 8.3-fold
increase in mortality (p = 0.0001) between patients with no ischemia and patients with any ischemia at all.
The validity of the Penn classification has been substantiated in various retrospective
studies[6]
[11]
[12] highlighting the significant impact malperfusion has on adverse postdissection outcomes.
[Table 1] illustrates the impact of malperfusion as an independent risk factor on short- and
long-term mortalities as reported across various studies.
Table 1
Mortality outcome across the degrees of malperfusion according to Penn's classification[5]
[6]
[11]
[12]
|
Mortality
|
Penn's classification
|
p-value
|
|
Aa
|
Ab
|
Ac
|
Abc
|
|
No ischemia
|
Localized ischemia
|
Generalized ischemia
|
Localized and generalized ischemia
|
|
Original derivation study
|
Augoustides et al[5] (2005)
|
All cause
|
3.1%
|
25.6%
|
17.6%
|
40%
|
0.0001
[a]
|
|
Intraoperative
|
N/A
|
N/A
|
N/A
|
N/A
|
N/A
|
|
In-hospital
|
N/A
|
N/A
|
N/A
|
N/A
|
N/A
|
|
5 years
|
N/A
|
N/A
|
N/A
|
N/A
|
N/A
|
|
Validation studies
|
Kimura et al[6] (2014)
|
All cause
|
N/A
|
N/A
|
N/A
|
N/A
|
N/A
|
|
Intraoperative
|
3.6%
|
5.9%
|
14%
|
15%
|
0.007
|
|
In-hospital
|
14%
|
24%
|
24%
|
44%
|
0.0007
|
|
5 years
|
N/A
|
N/A
|
N/A
|
N/A
|
N/A
|
|
Olsson et al[11] (2011)
|
All cause
|
N/A
|
N/A
|
N/A
|
N/A
|
N/A
|
|
Intraoperative
|
N/A
|
N/A
|
N/A
|
N/A
|
N/A
|
|
In-hospital
|
3%
|
6%
|
17%
|
22%
|
<0.01
|
|
5 years
|
15%
|
26%
|
22%
|
33%
|
Ab: p = 0.027
|
|
Abc: p < 0.001
|
|
Pisano et al[12] (2016)
|
All cause
|
N/A
|
N/A
|
N/A
|
N/A
|
N/A
|
|
Intraoperative
|
N/A
|
N/A
|
N/A
|
N/A
|
N/A
|
|
In-hospital
|
10.7%
|
Non-class Aa: 36.7%[b]
|
0.02
|
|
5 years
|
N/A
|
N/A
|
N/A
|
N/A
|
N/A
|
Note: N/A, not available.
a
p = 0.0001 for the comparison of no ischemia versus any ischemia.
b The reported in-hospital mortality by Pisano et al[12] is for collective Penn's class non-Aa versus Aa.
Kimura et al[6] have investigated early and late outcomes following intervention for acute TYPE
A aortic dissection (ATAAD), using the Penn risk modeling tool. The authors showed
significant (p < 0.01) differences in their results for in-hospital mortality across the different
Penn classes: 3 versus 22% for class Aa and Abc, respectively. Further analysis of
longer-term outcomes reported significantly lower survival at 5 years for the Penn
class Ac and Abc, compared with class A.[6]
Upon analysis of a larger group of 360 patients by Olsson et al,[11] the Penn class Abc was validated as a significant independent predictor and risk
factor, for intraoperative (p = 0.03) and in-hospital mortality (p = 0.02), respectively. Furthermore, Penn's class non-A (p = 0.014) was in itself an independent risk factor for in-hospital mortality, further
highlighting the grave impact of malperfusion.[11]
While the Penn classification is well validated, there are limitations to bear in
mind.[5] First, building on current validation studies in the literature, there remains a
need for prospective validation of the Penn findings in large populations.[5] It is important to consider the hindering limitations of retrospective studies,
arising from the lack of recognition of other potential risk factors and their impact
on postoperative complications.[11] Furthermore, there is ambiguity in how ischemia was described, a limitation building
from the initial Penn's study itself. Also, there is a need for consideration of the
vast diversity of clinical manifestations, of ATAAD.[6] Nonetheless some validating studies do provide support for the Penn findings. The
distribution of patients amongst classes was similar in validating studies. The significant
impact that malperfusion has on outcomes ([Fig. 1], [Table 1]) is confirmed.[6]
[11] However, there is still value to be added from consideration of other clinical factors,
as our review shall discuss.
Bringing in the Biochemistry
Ghoreishi et al[1] sought to develop a risk model, building on the premise of the impact of malperfusion,
to improve their predicting power for mortality following repair of ATAAD. In their
retrospective review at a single institution over a 13-year period, results from 269
patients were included.[1] Following multivariate analysis, they found that creatinine (p = 0.0008), lactic acid (p = 0.01), and liver malperfusion (p = 0.02) were significant risk factors for operative mortality. This risk model attained
a c-statistic of 0.75.
Although the authors describe the value of their model, the limitations of their study
are acknowledged. The single-centered retrospective nature is one limitation, alongside
the need for analysis of long-term results as well as prospective external validation
on a larger scale still being required.[1]
International Registry for Acute Aortic Dissection Analysis and Perioperative Factors
With the goal of providing at bedside a tool that can aid the decision-making framework
for a surgeon considering an intervention, Rampoldi et al[2] created a risk prediction model for patients undergoing ATAAD repair using their
retrospective analysis of 682 patients from the International Registry for Acute Aortic
Dissection (IRAD) from 1996 to 2003. Their model encompassed not only preoperative
variables but also intraoperative variables as well.
Univariate analysis was performed to identify statistically significant clinical characteristics
which had an independent impact on operative mortality ([Fig. 2]).
Fig. 2
(A) Clinical perioperative risk factors used in a risk model devised by Rampoldi et al
on their analysis of patients from IRAD.[2]
(B) clinical preoperative parameters used in the risk model devised by Czerny et al on
their analysis of patients from GERAADA.[17] GERAADA, the German Registry for Acute Aortic Dissection Type A; IRAD, International
Registry for Acute Aortic Dissection.
The authors reported similar significant risk factors of operative mortality as those
of prior studies.[2]
[13]
[14] The authors point out that their findings should be generalizable, given data collection
from various institutions in six countries, with a vast range of clinical presentations
taken into consideration. However, their model considers only early mortality. There
needs to be further validation of these models, especially for longer-term results.[2]
Yu et al[15] analyzed the predictive value of the aforementioned model[2] in a retrospective analysis of a small group of 79 patients. They found their model
inadequate in accuracy for risk prediction of surgical outcomes.
A Scorecard Model
The attempt to create a preoperative “scorecard” for risk prediction of operative
mortality was pursued by Leontyev et al.[4] The independent risk factors included age, critical preoperative state, malperfusion
syndrome, and coronary artery disease. The authors assigned integer scores for each
risk factor depending on the extent of impact; visceral malperfusion would score 3,
whereas coronary malperfusion would score 1.[4]
With a sample size of 534 consecutive patients, over two institutes with similar patient
population characteristics, this model was based on a larger cohort compared with
past models.[2]
[16] Selection bias will always hinder retrospective studies and this is no different
for this study, further hampered by a lack of consideration of confounding variables
over the study period of almost two decades.
The German Registry for Acute Aortic Dissection Type A Analysis, Another Scoring Model
A recent multicenter experience of 2,537 patients analyzed from the GERAADA, attempted
risk prediction of 30-day mortality. Czerny et al[17] sought to devise a “scoring system” for risk prediction of this early outcome. The
final model devised incorporated a variety of clinical and preoperative parameters
following multivariate analysis ([Fig. 2]). This study also overlooked the natural history, genetics, and foundational risks
from before the dissection itself. The authors aimed for a “simple, effective tool”[17] for the prediction of only early mortality. The large sample size of patients analyzed
from over 50 centers confers power to this study. However, the prediction was found
only moderately accurate, with the area under the curve (AUC) of 0.725, lower than
The European System for Cardiac Operative Risk Evaluation. A key limitation of GERAADA
is the lack of detailed reporting regarding malperfusion and its significant impact
on prognosis. Furthermore, there was a lack of consideration for the morphological
nature of the tear-tailored surgical approach, which may or may not have been taken
in individual cases.[17]
The GERAADA score devised by Czerny et al[17] has recently been evaluated in a 10-year retrospective study of 371 patients operated
on for ATAAD. Luehr et al[18] aimed to investigate if the prediction attained using GERAADA score corresponded
with that of the authors' institution results. The authors reported their actual 30-day
mortality to be 15.7%, which had been comparable, with no significant difference (p = 0.776), to the predicted mortality of 15.1%. Several patient subgroups had mortality
rates with a greater raw difference (higher and lower levels) than those predicted
by the GERAADA score; however, these had not reached statistical significance. Luehr
et al reported that following multivariable analysis it had been age, resuscitation
prior to surgery, and other/unknown malperfusion that were significant independent
risk factors for 30-day mortality. The authors use of this risk prediction model attained
an AUC score of 0.673. We have illustrated in [Fig. 3], the varying AUC and c-statistic values reported from the respective risk models
in the literature. There remains a need for further evaluation of this risk stratification
tool in larger scale prospective studies. The authors expressed concerns of over-
or underprediction of mortality using the GERAADA score, especially for smaller subgroups
of less than 100 patients where there had been notable raw differences in actual vs
predicted rates, albeit not statistically significant.[18]
Fig. 3 A graph illustrating the AUC and c-statistic scores attained by various risk prediction
models in the literature.[1]
[2]
[4]
[11]
[17] AUC, area under the curve; ROC, receiver operating characteristics.
The Problem at Hand
We need a predictive tool that functions with ease, accuracy, and precision, despite
the inherent limitations that often hinder such studies. Currently, due to limitations,
we have only half a canvas.
Exploring the Natural History: The Missed Elements
Exploring the Natural History: The Missed Elements
The Upcoming Dissection, Does Size Matter?
A study reporting on the Yale database[19] analyzing results of 304 patients from 1985 to 2000 concluded that greater risks
of rupture and dissection are linked to the greater initial size of the aorta (p = 0.006). An aorta being >6 cm is reported as three times worse in comparison to
sizes of 4 to 4.9 cm. Davies et al[19] further reported that long-term survival was lower for patients with greater-sized
aneurysms (p = 0.0039).
Although the relatively large sample size along with long follow-up periods gives
strength to this study, which provided great insight into the natural history and
risk factors for rupture or dissection, patients followed in this single-centered
retrospective study were operated on electively based on criteria of aortic size,
hence not permitting prediction of postdissection outcomes. The reported yearly rates
may in fact be a representation of the lower limits of the more accurate rates, as
discussed by the authors.[19]
The Genetic Factors to Consider
Greater than 90% of thoracic aortic aneurysms are asymptomatic prior to dissection,
with less than half being diagnosed appropriately in emergency departments prior to
patient death.[20] Furthermore, considering that approximately a fifth of dissection patients have
a corresponding family history, the importance of investigating genetic links is amplified.[21]
[22]
The build-up to both syndromic and nonsyndromic onsets can potentially boil down to
a single mutated gene as the causative factor.[21] The syndromic side is better defined, resultant of our stronger understanding of
the interactions between aortic pathologies and connective tissue disorders such as
the Loeys–Dietz, Marfan, and Ehlers–Danlos syndromes.[21]
[23] Pathogenesis involves “dysfunction of the extracellular matrix, medial smooth muscle
cells, or TGF-β signaling.”[21] Recognition of these syndromic cases may require more aggressive and extensive replacement
approaches.[24]
[25]
It was initially thought that the rest of the patients without a family history, these
so-termed “sporadic” cases, were purely degenerative conditions; however, research
has suggested that there may well be an underlying genetic mechanism.[26] Even in patients with sporadic AD, pathogenic genetic variants found are in fact
common with those of syndromic features. One such example is a variant of the FBN1 gene, with its causative link in Marfan syndrome, which can also be a risk factor
for sporadic cases.[21] Guo et al[26] found that 9.3% of patients with sporadic thoracic AD in fact had pathogenic variants
in heritable genes. Genetic screening may well benefit not just the patient but also
their families, as further discussed by Ostberg et al.[21]
As recently as 2017, there were only 29 identified genes associated with the development
of AD.[27] In the latest updates from 2019, there are now 37 identified genes for this association.[28] These genes still only explain approximately 30% of familial nonsyndromic TAAD.[26]
[Fig. 4] illustrates the distribution of the aforementioned genes.[29]
Fig. 4 Frequency distribution of genetic defects in thoracic aortic aneurysm and dissection—related
genes as per the routine genetic testing program at the Yale Aortic Institute. Reproduced
with permission from Vinholo et al.[29]
Bearing these thoughts in mind and the progressive speed with which the “genetic dictionary”
will continue to evolve, alongside our evolving understanding of the genetic foundation
and natural history of AD will allow a new page to be written in the field of personalized
precision medicine.[28]
Family History and Prognosis
Genetic factors have an impact on the heritability and incidence of AD.[26]
[27]
[28] However, a key aspect is the impact that the presence of AD in the family history
may have on prognostic outcomes.
Chen et al[30] hypothesized an association between the presence of family history and the prognosis
of AD. In their nationwide study, a total of 93 AD patients with a family history
were matched with 894 control AD patients without any family history. The propensity-score
matching process encompassed: “age at diagnosis of AD, sex, comorbidities, and medication.”[30] The authors reported no significant difference between the groups for rates of in-hospital
mortality. There may have been higher incidences of root replacement procedures for
ATAAD patients that had a family history. It was found that for patients in the family
history group, there was a significantly higher risk of the patients having to receive
later aortic surgical intervention.[30]
Despite the strengths of this prospective (from the National Health Insurance Research
Database), we cannot ignore some key limitations. The potential misclassification
of diagnosis, lack of anatomic imaging in the database, and difficulty in generalizing
these conclusions based on results from the Taiwanese population are notable weaknesses.
Furthermore, there was a stark difference in the number of patients in the compared
groups.[30]
The Unfortunate Impact of Delays
A plethora of nonspecific symptoms, an insidious onset, and a vast range of symptoms
often lead to delay in diagnosis of AD,[8] especially distressing given the mortality rate of 1 to 2% per hour.[9] The concomitant events of carotid and cerebral malperfusion carry the burden of
a worse prognosis, not only due to poorer clinical conditions of the patients but
due to the heightened risk of misdiagnosis.[8]
[10]
Harris et al[31] reported from their analysis of data from IRAD that those patients who had undergone
previous cardiac surgery, were transferred from nontertiary centers of non-White race,
or female had significantly greater time taken from the presentation to diagnosis.
Although extending transport times, there remain significant advantages of transferring
patients to high-volume centers. Goldstone et al reported that there was in fact a
7.2% reduction in risk of operative mortality for patients transferred to high-volume
centers.[32] Our recent analysis of 249 patients, from the National Institute for Cardiovascular
Outcomes Research database, further concluded that greater levels of in-hospital mortality
are related to lower volume surgeons.[33]
Conclusion… Paving the Future
Conclusion… Paving the Future
We have discussed the need for consideration and appreciation of the significance
carried by factors both prior to and following AD and how a true risk prediction model
of adverse outcomes postdissection can only be developed with due respect for all
these factors surrounding AD.
A greater aortic size not only predicts the risk of dissection occurring but carries
with it a greater risk of worse long-term outcomes, amplifying its importance in the
decision-making framework.[19] There is recognition of the impact that connective tissue disorders have on the
management for AD; however, even sporadic cases of dissection can have pathogenic
variants of heritable genes such as FBN1.[21] Recognition of the many genes associated with AD eloquently highlights the importance
of considering the genetic foundation that is painting the picture and the need for
its incorporation in the decision-making framework.[28]
The Penn classification has been recognized as a valuable tool for risk stratification
for AD. Despite some significant limitations of the validating studies, there is strong
encouragement and verification of this classification as a risk model.[5]
[6]
[11]
[12] There are various other risk models, yet to be validated, proposed in the literature,
showing that fellow perioperative factors, blood biochemistry results, and comorbidities
all have a role in risk prediction ([Table 2]).
Table 2
Risk prediction of adverse outcomes after acute aortic dissection: key messages, points,
and conclusions
|
Key summary points
|
|
• Various studies have devised their own models for the stratification of risk prediction
in acute aortic dissection. These have significant inherent limitations, and majority
of the models lack large-scale prospective external validation
|
|
• The well-validated Penn classification demonstrates the incorporation of varying
levels of malperfusion as a significant factor in a risk prediction model
|
|
• Many of these risk prediction models following dissection overlook the influence
of predissection risk and patient characteristics
|
|
• Malperfusion, blood biochemistry, and clinical perioperative factors are significant
components for a risk prediction model; however, they make up one end of the spectrum.
|
|
• Impact of genetics in syndromic and nonsyndromic cases are fundamental elements.
Further appreciation for aortic size and the impact of delays in transport and diagnosis
are significant covariates for postdissection outcomes
|
|
• We must aim toward personalized precision medicine, which is only achieved by incorporating
all risk prediction components into the decision-making framework
|
We have discussed in our recent paper how risk profiling, optimum assessment, and
increased awareness for AD will facilitate diagnosis and efficient transition of patients
from ER to OR. This will further reduce the risk of postdissection mortality.[34]
This journey to develop a stratification tool for the risk of adverse outcomes once
the dissection has occurred builds atop the very foundation of risk prior to the dissection
itself. We must appreciate the classification of the dissection regarding syndromic
and nonsyndromic, with recognition of family history, understanding of the natural
history with the immense genetic impact, alongside the vital role those clinical characteristics
play. Only following the incorporation of all these various aspects can we begin to
put the puzzle together and have a true risk prediction model to consider for our
patients.
